diff --git a/.gitattributes b/.gitattributes
index a6344aac8c09253b3b630fb776ae94478aa0275b..c7b9ec682bf0348c48118671ad47db589404458a 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text
+img/langchain+chatglm.png filter=lfs diff=lfs merge=lfs -text
+img/official_wechat_mp_account.png filter=lfs diff=lfs merge=lfs -text
+knowledge_base/samples/content/llm/img/分布式训练技术原理-幕布图片-392521-261326.jpg filter=lfs diff=lfs merge=lfs -text
diff --git a/chains/llmchain_with_history.py b/chains/llmchain_with_history.py
new file mode 100644
index 0000000000000000000000000000000000000000..2a845086175086b8f00195212ec13c94a8ac707d
--- /dev/null
+++ b/chains/llmchain_with_history.py
@@ -0,0 +1,22 @@
+from server.utils import get_ChatOpenAI
+from configs.model_config import LLM_MODELS, TEMPERATURE
+from langchain.chains import LLMChain
+from langchain.prompts.chat import (
+ ChatPromptTemplate,
+ HumanMessagePromptTemplate,
+)
+
+model = get_ChatOpenAI(model_name=LLM_MODELS[0], temperature=TEMPERATURE)
+
+
+human_prompt = "{input}"
+human_message_template = HumanMessagePromptTemplate.from_template(human_prompt)
+
+chat_prompt = ChatPromptTemplate.from_messages(
+ [("human", "我们来玩成语接龙,我先来,生龙活虎"),
+ ("ai", "虎头虎脑"),
+ ("human", "{input}")])
+
+
+chain = LLMChain(prompt=chat_prompt, llm=model, verbose=True)
+print(chain({"input": "恼羞成怒"}))
\ No newline at end of file
diff --git a/common/__init__.py b/common/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/configs/__init__.py b/configs/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e4f5f1d049f1b9593ace154e20c0952e82f6d5ed
--- /dev/null
+++ b/configs/__init__.py
@@ -0,0 +1,8 @@
+from .basic_config import *
+from .model_config import *
+from .kb_config import *
+from .server_config import *
+from .prompt_config import *
+
+
+VERSION = "v0.2.10"
diff --git a/configs/basic_config.py.example b/configs/basic_config.py.example
new file mode 100644
index 0000000000000000000000000000000000000000..7b50365fb1d0e90da048954170aaf355cbe2d320
--- /dev/null
+++ b/configs/basic_config.py.example
@@ -0,0 +1,32 @@
+import logging
+import os
+import langchain
+import tempfile
+import shutil
+
+
+# 是否显示详细日志
+log_verbose = False
+langchain.verbose = False
+
+# 通常情况下不需要更改以下内容
+
+# 日志格式
+LOG_FORMAT = "%(asctime)s - %(filename)s[line:%(lineno)d] - %(levelname)s: %(message)s"
+logger = logging.getLogger()
+logger.setLevel(logging.INFO)
+logging.basicConfig(format=LOG_FORMAT)
+
+
+# 日志存储路径
+LOG_PATH = os.path.join(os.path.dirname(os.path.dirname(__file__)), "logs")
+if not os.path.exists(LOG_PATH):
+ os.mkdir(LOG_PATH)
+
+# 临时文件目录,主要用于文件对话
+BASE_TEMP_DIR = os.path.join(tempfile.gettempdir(), "chatchat")
+try:
+ shutil.rmtree(BASE_TEMP_DIR)
+except Exception:
+ pass
+os.makedirs(BASE_TEMP_DIR, exist_ok=True)
diff --git a/configs/kb_config.py.example b/configs/kb_config.py.example
new file mode 100644
index 0000000000000000000000000000000000000000..23e06bdcb8c305004f1c1c2d8c7db0301405bd4f
--- /dev/null
+++ b/configs/kb_config.py.example
@@ -0,0 +1,145 @@
+import os
+
+# 默认使用的知识库
+DEFAULT_KNOWLEDGE_BASE = "samples"
+
+# 默认向量库/全文检索引擎类型。可选:faiss, milvus(离线) & zilliz(在线), pgvector,全文检索引擎es
+DEFAULT_VS_TYPE = "faiss"
+
+# 缓存向量库数量(针对FAISS)
+CACHED_VS_NUM = 1
+
+# 缓存临时向量库数量(针对FAISS),用于文件对话
+CACHED_MEMO_VS_NUM = 10
+
+# 知识库中单段文本长度(不适用MarkdownHeaderTextSplitter)
+CHUNK_SIZE = 250
+
+# 知识库中相邻文本重合长度(不适用MarkdownHeaderTextSplitter)
+OVERLAP_SIZE = 50
+
+# 知识库匹配向量数量
+VECTOR_SEARCH_TOP_K = 3
+
+# 知识库匹配的距离阈值,一般取值范围在0-1之间,SCORE越小,距离越小从而相关度越高。
+# 但有用户报告遇到过匹配分值超过1的情况,为了兼容性默认设为1,在WEBUI中调整范围为0-2
+SCORE_THRESHOLD = 1.0
+
+# 默认搜索引擎。可选:bing, duckduckgo, metaphor
+DEFAULT_SEARCH_ENGINE = "duckduckgo"
+
+# 搜索引擎匹配结题数量
+SEARCH_ENGINE_TOP_K = 3
+
+
+# Bing 搜索必备变量
+# 使用 Bing 搜索需要使用 Bing Subscription Key,需要在azure port中申请试用bing search
+# 具体申请方式请见
+# https://learn.microsoft.com/en-us/bing/search-apis/bing-web-search/create-bing-search-service-resource
+# 使用python创建bing api 搜索实例详见:
+# https://learn.microsoft.com/en-us/bing/search-apis/bing-web-search/quickstarts/rest/python
+BING_SEARCH_URL = "https://api.bing.microsoft.com/v7.0/search"
+# 注意不是bing Webmaster Tools的api key,
+
+# 此外,如果是在服务器上,报Failed to establish a new connection: [Errno 110] Connection timed out
+# 是因为服务器加了防火墙,需要联系管理员加白名单,如果公司的服务器的话,就别想了GG
+BING_SUBSCRIPTION_KEY = ""
+
+# metaphor搜索需要KEY
+METAPHOR_API_KEY = ""
+
+# 心知天气 API KEY,用于天气Agent。申请:https://www.seniverse.com/
+SENIVERSE_API_KEY = ""
+
+# 是否开启中文标题加强,以及标题增强的相关配置
+# 通过增加标题判断,判断哪些文本为标题,并在metadata中进行标记;
+# 然后将文本与往上一级的标题进行拼合,实现文本信息的增强。
+ZH_TITLE_ENHANCE = False
+
+# PDF OCR 控制:只对宽高超过页面一定比例(图片宽/页面宽,图片高/页面高)的图片进行 OCR。
+# 这样可以避免 PDF 中一些小图片的干扰,提高非扫描版 PDF 处理速度
+PDF_OCR_THRESHOLD = (0.6, 0.6)
+
+# 每个知识库的初始化介绍,用于在初始化知识库时显示和Agent调用,没写则没有介绍,不会被Agent调用。
+KB_INFO = {
+ "知识库名称": "知识库介绍",
+ "samples": "关于本项目issue的解答",
+}
+
+
+# 通常情况下不需要更改以下内容
+
+# 知识库默认存储路径
+KB_ROOT_PATH = os.path.join(os.path.dirname(os.path.dirname(__file__)), "knowledge_base")
+if not os.path.exists(KB_ROOT_PATH):
+ os.mkdir(KB_ROOT_PATH)
+# 数据库默认存储路径。
+# 如果使用sqlite,可以直接修改DB_ROOT_PATH;如果使用其它数据库,请直接修改SQLALCHEMY_DATABASE_URI。
+DB_ROOT_PATH = os.path.join(KB_ROOT_PATH, "info.db")
+SQLALCHEMY_DATABASE_URI = f"sqlite:///{DB_ROOT_PATH}"
+
+# 可选向量库类型及对应配置
+kbs_config = {
+ "faiss": {
+ },
+ "milvus": {
+ "host": "127.0.0.1",
+ "port": "19530",
+ "user": "",
+ "password": "",
+ "secure": False,
+ },
+ "zilliz": {
+ "host": "in01-a7ce524e41e3935.ali-cn-hangzhou.vectordb.zilliz.com.cn",
+ "port": "19530",
+ "user": "",
+ "password": "",
+ "secure": True,
+ },
+ "pg": {
+ "connection_uri": "postgresql://postgres:postgres@127.0.0.1:5432/langchain_chatchat",
+ },
+
+ "es": {
+ "host": "127.0.0.1",
+ "port": "9200",
+ "index_name": "test_index",
+ "user": "",
+ "password": ""
+ },
+ "milvus_kwargs":{
+ "search_params":{"metric_type": "L2"}, #在此处增加search_params
+ "index_params":{"metric_type": "L2","index_type": "HNSW"} # 在此处增加index_params
+ }
+}
+
+# TextSplitter配置项,如果你不明白其中的含义,就不要修改。
+text_splitter_dict = {
+ "ChineseRecursiveTextSplitter": {
+ "source": "huggingface", # 选择tiktoken则使用openai的方法
+ "tokenizer_name_or_path": "",
+ },
+ "SpacyTextSplitter": {
+ "source": "huggingface",
+ "tokenizer_name_or_path": "gpt2",
+ },
+ "RecursiveCharacterTextSplitter": {
+ "source": "tiktoken",
+ "tokenizer_name_or_path": "cl100k_base",
+ },
+ "MarkdownHeaderTextSplitter": {
+ "headers_to_split_on":
+ [
+ ("#", "head1"),
+ ("##", "head2"),
+ ("###", "head3"),
+ ("####", "head4"),
+ ]
+ },
+}
+
+# TEXT_SPLITTER 名称
+TEXT_SPLITTER_NAME = "ChineseRecursiveTextSplitter"
+
+# Embedding模型定制词语的词表文件
+EMBEDDING_KEYWORD_FILE = "embedding_keywords.txt"
diff --git a/configs/model_config.py.example b/configs/model_config.py.example
new file mode 100644
index 0000000000000000000000000000000000000000..8746f098bfd9467819eb8d3dd6633d7dac49b899
--- /dev/null
+++ b/configs/model_config.py.example
@@ -0,0 +1,302 @@
+import os
+
+# 可以指定一个绝对路径,统一存放所有的Embedding和LLM模型。
+# 每个模型可以是一个单独的目录,也可以是某个目录下的二级子目录。
+# 如果模型目录名称和 MODEL_PATH 中的 key 或 value 相同,程序会自动检测加载,无需修改 MODEL_PATH 中的路径。
+MODEL_ROOT_PATH = ""
+
+# 选用的 Embedding 名称
+EMBEDDING_MODEL = "bge-large-zh-v1.5"
+
+# Embedding 模型运行设备。设为 "auto" 会自动检测(会有警告),也可手动设定为 "cuda","mps","cpu","xpu" 其中之一。
+EMBEDDING_DEVICE = "auto"
+
+# 选用的reranker模型
+RERANKER_MODEL = "bge-reranker-large"
+# 是否启用reranker模型
+USE_RERANKER = False
+RERANKER_MAX_LENGTH = 1024
+
+# 如果需要在 EMBEDDING_MODEL 中增加自定义的关键字时配置
+EMBEDDING_KEYWORD_FILE = "keywords.txt"
+EMBEDDING_MODEL_OUTPUT_PATH = "output"
+
+# 要运行的 LLM 名称,可以包括本地模型和在线模型。列表中本地模型将在启动项目时全部加载。
+# 列表中第一个模型将作为 API 和 WEBUI 的默认模型。
+# 在这里,我们使用目前主流的两个离线模型,其中,chatglm3-6b 为默认加载模型。
+# 如果你的显存不足,可使用 Qwen-1_8B-Chat, 该模型 FP16 仅需 3.8G显存。
+
+LLM_MODELS = ["chatglm3-6b", "zhipu-api", "openai-api"]
+Agent_MODEL = None
+
+# LLM 模型运行设备。设为"auto"会自动检测(会有警告),也可手动设定为 "cuda","mps","cpu","xpu" 其中之一。
+LLM_DEVICE = "auto"
+
+HISTORY_LEN = 3
+
+MAX_TOKENS = 2048
+
+TEMPERATURE = 0.7
+
+ONLINE_LLM_MODEL = {
+ "openai-api": {
+ "model_name": "gpt-4",
+ "api_base_url": "https://api.openai.com/v1",
+ "api_key": "",
+ "openai_proxy": "",
+ },
+
+ # 智谱AI API,具体注册及api key获取请前往 http://open.bigmodel.cn
+ "zhipu-api": {
+ "api_key": "",
+ "version": "glm-4",
+ "provider": "ChatGLMWorker",
+ },
+
+ # 具体注册及api key获取请前往 https://api.minimax.chat/
+ "minimax-api": {
+ "group_id": "",
+ "api_key": "",
+ "is_pro": False,
+ "provider": "MiniMaxWorker",
+ },
+
+ # 具体注册及api key获取请前往 https://xinghuo.xfyun.cn/
+ "xinghuo-api": {
+ "APPID": "",
+ "APISecret": "",
+ "api_key": "",
+ "version": "v3.0", # 你使用的讯飞星火大模型版本,可选包括 "v3.0", "v2.0", "v1.5"
+ "provider": "XingHuoWorker",
+ },
+
+ # 百度千帆 API,申请方式请参考 https://cloud.baidu.com/doc/WENXINWORKSHOP/s/4lilb2lpf
+ "qianfan-api": {
+ "version": "ERNIE-Bot", # 注意大小写。当前支持 "ERNIE-Bot" 或 "ERNIE-Bot-turbo", 更多的见官方文档。
+ "version_url": "", # 也可以不填写version,直接填写在千帆申请模型发布的API地址
+ "api_key": "",
+ "secret_key": "",
+ "provider": "QianFanWorker",
+ },
+
+ # 火山方舟 API,文档参考 https://www.volcengine.com/docs/82379
+ "fangzhou-api": {
+ "version": "chatglm-6b-model",
+ "version_url": "",
+ "api_key": "",
+ "secret_key": "",
+ "provider": "FangZhouWorker",
+ },
+
+ # 阿里云通义千问 API,文档参考 https://help.aliyun.com/zh/dashscope/developer-reference/api-details
+ "qwen-api": {
+ "version": "qwen-max",
+ "api_key": "",
+ "provider": "QwenWorker",
+ "embed_model": "text-embedding-v1" # embedding 模型名称
+ },
+
+ # 百川 API,申请方式请参考 https://www.baichuan-ai.com/home#api-enter
+ "baichuan-api": {
+ "version": "Baichuan2-53B",
+ "api_key": "",
+ "secret_key": "",
+ "provider": "BaiChuanWorker",
+ },
+
+ # Azure API
+ "azure-api": {
+ "deployment_name": "", # 部署容器的名字
+ "resource_name": "", # https://{resource_name}.openai.azure.com/openai/ 填写resource_name的部分,其他部分不要填写
+ "api_version": "", # API的版本,不是模型版本
+ "api_key": "",
+ "provider": "AzureWorker",
+ },
+
+ # 昆仑万维天工 API https://model-platform.tiangong.cn/
+ "tiangong-api": {
+ "version": "SkyChat-MegaVerse",
+ "api_key": "",
+ "secret_key": "",
+ "provider": "TianGongWorker",
+ },
+ # Gemini API https://makersuite.google.com/app/apikey
+ "gemini-api": {
+ "api_key": "",
+ "provider": "GeminiWorker",
+ }
+
+}
+
+# 在以下字典中修改属性值,以指定本地embedding模型存储位置。支持3种设置方法:
+# 1、将对应的值修改为模型绝对路径
+# 2、不修改此处的值(以 text2vec 为例):
+# 2.1 如果{MODEL_ROOT_PATH}下存在如下任一子目录:
+# - text2vec
+# - GanymedeNil/text2vec-large-chinese
+# - text2vec-large-chinese
+# 2.2 如果以上本地路径不存在,则使用huggingface模型
+
+MODEL_PATH = {
+ "embed_model": {
+ "ernie-tiny": "nghuyong/ernie-3.0-nano-zh",
+ "ernie-base": "nghuyong/ernie-3.0-base-zh",
+ "text2vec-base": "shibing624/text2vec-base-chinese",
+ "text2vec": "GanymedeNil/text2vec-large-chinese",
+ "text2vec-paraphrase": "shibing624/text2vec-base-chinese-paraphrase",
+ "text2vec-sentence": "shibing624/text2vec-base-chinese-sentence",
+ "text2vec-multilingual": "shibing624/text2vec-base-multilingual",
+ "text2vec-bge-large-chinese": "shibing624/text2vec-bge-large-chinese",
+ "m3e-small": "moka-ai/m3e-small",
+ "m3e-base": "moka-ai/m3e-base",
+ "m3e-large": "moka-ai/m3e-large",
+ "bge-small-zh": "BAAI/bge-small-zh",
+ "bge-base-zh": "BAAI/bge-base-zh",
+ "bge-large-zh": "BAAI/bge-large-zh",
+ "bge-large-zh-noinstruct": "BAAI/bge-large-zh-noinstruct",
+ "bge-base-zh-v1.5": "BAAI/bge-base-zh-v1.5",
+ "bge-large-zh-v1.5": "BAAI/bge-large-zh-v1.5",
+ "piccolo-base-zh": "sensenova/piccolo-base-zh",
+ "piccolo-large-zh": "sensenova/piccolo-large-zh",
+ "nlp_gte_sentence-embedding_chinese-large": "damo/nlp_gte_sentence-embedding_chinese-large",
+ "text-embedding-ada-002": "your OPENAI_API_KEY",
+ },
+
+ "llm_model": {
+ "chatglm2-6b": "THUDM/chatglm2-6b",
+ "chatglm2-6b-32k": "THUDM/chatglm2-6b-32k",
+ "chatglm3-6b": "THUDM/chatglm3-6b",
+ "chatglm3-6b-32k": "THUDM/chatglm3-6b-32k",
+
+ "Orion-14B-Chat": "OrionStarAI/Orion-14B-Chat",
+ "Orion-14B-Chat-Plugin": "OrionStarAI/Orion-14B-Chat-Plugin",
+ "Orion-14B-LongChat": "OrionStarAI/Orion-14B-LongChat",
+
+ "Llama-2-7b-chat-hf": "meta-llama/Llama-2-7b-chat-hf",
+ "Llama-2-13b-chat-hf": "meta-llama/Llama-2-13b-chat-hf",
+ "Llama-2-70b-chat-hf": "meta-llama/Llama-2-70b-chat-hf",
+
+ "Qwen-1_8B-Chat": "Qwen/Qwen-1_8B-Chat",
+ "Qwen-7B-Chat": "Qwen/Qwen-7B-Chat",
+ "Qwen-14B-Chat": "Qwen/Qwen-14B-Chat",
+ "Qwen-72B-Chat": "Qwen/Qwen-72B-Chat",
+
+ "baichuan-7b-chat": "baichuan-inc/Baichuan-7B-Chat",
+ "baichuan-13b-chat": "baichuan-inc/Baichuan-13B-Chat",
+ "baichuan2-7b-chat": "baichuan-inc/Baichuan2-7B-Chat",
+ "baichuan2-13b-chat": "baichuan-inc/Baichuan2-13B-Chat",
+
+ "internlm-7b": "internlm/internlm-7b",
+ "internlm-chat-7b": "internlm/internlm-chat-7b",
+ "internlm2-chat-7b": "internlm/internlm2-chat-7b",
+ "internlm2-chat-20b": "internlm/internlm2-chat-20b",
+
+ "BlueLM-7B-Chat": "vivo-ai/BlueLM-7B-Chat",
+ "BlueLM-7B-Chat-32k": "vivo-ai/BlueLM-7B-Chat-32k",
+
+ "Yi-34B-Chat": "https://huggingface.co/01-ai/Yi-34B-Chat",
+
+ "agentlm-7b": "THUDM/agentlm-7b",
+ "agentlm-13b": "THUDM/agentlm-13b",
+ "agentlm-70b": "THUDM/agentlm-70b",
+
+ "falcon-7b": "tiiuae/falcon-7b",
+ "falcon-40b": "tiiuae/falcon-40b",
+ "falcon-rw-7b": "tiiuae/falcon-rw-7b",
+
+ "aquila-7b": "BAAI/Aquila-7B",
+ "aquilachat-7b": "BAAI/AquilaChat-7B",
+ "open_llama_13b": "openlm-research/open_llama_13b",
+ "vicuna-13b-v1.5": "lmsys/vicuna-13b-v1.5",
+ "koala": "young-geng/koala",
+ "mpt-7b": "mosaicml/mpt-7b",
+ "mpt-7b-storywriter": "mosaicml/mpt-7b-storywriter",
+ "mpt-30b": "mosaicml/mpt-30b",
+ "opt-66b": "facebook/opt-66b",
+ "opt-iml-max-30b": "facebook/opt-iml-max-30b",
+ "gpt2": "gpt2",
+ "gpt2-xl": "gpt2-xl",
+ "gpt-j-6b": "EleutherAI/gpt-j-6b",
+ "gpt4all-j": "nomic-ai/gpt4all-j",
+ "gpt-neox-20b": "EleutherAI/gpt-neox-20b",
+ "pythia-12b": "EleutherAI/pythia-12b",
+ "oasst-sft-4-pythia-12b-epoch-3.5": "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5",
+ "dolly-v2-12b": "databricks/dolly-v2-12b",
+ "stablelm-tuned-alpha-7b": "stabilityai/stablelm-tuned-alpha-7b",
+ },
+
+ "reranker": {
+ "bge-reranker-large": "BAAI/bge-reranker-large",
+ "bge-reranker-base": "BAAI/bge-reranker-base",
+ }
+}
+
+# 通常情况下不需要更改以下内容
+
+# nltk 模型存储路径
+NLTK_DATA_PATH = os.path.join(os.path.dirname(os.path.dirname(__file__)), "nltk_data")
+
+# 使用VLLM可能导致模型推理能力下降,无法完成Agent任务
+VLLM_MODEL_DICT = {
+ "chatglm2-6b": "THUDM/chatglm2-6b",
+ "chatglm2-6b-32k": "THUDM/chatglm2-6b-32k",
+ "chatglm3-6b": "THUDM/chatglm3-6b",
+ "chatglm3-6b-32k": "THUDM/chatglm3-6b-32k",
+
+ "Llama-2-7b-chat-hf": "meta-llama/Llama-2-7b-chat-hf",
+ "Llama-2-13b-chat-hf": "meta-llama/Llama-2-13b-chat-hf",
+ "Llama-2-70b-chat-hf": "meta-llama/Llama-2-70b-chat-hf",
+
+ "Qwen-1_8B-Chat": "Qwen/Qwen-1_8B-Chat",
+ "Qwen-7B-Chat": "Qwen/Qwen-7B-Chat",
+ "Qwen-14B-Chat": "Qwen/Qwen-14B-Chat",
+ "Qwen-72B-Chat": "Qwen/Qwen-72B-Chat",
+
+ "baichuan-7b-chat": "baichuan-inc/Baichuan-7B-Chat",
+ "baichuan-13b-chat": "baichuan-inc/Baichuan-13B-Chat",
+ "baichuan2-7b-chat": "baichuan-inc/Baichuan-7B-Chat",
+ "baichuan2-13b-chat": "baichuan-inc/Baichuan-13B-Chat",
+
+ "BlueLM-7B-Chat": "vivo-ai/BlueLM-7B-Chat",
+ "BlueLM-7B-Chat-32k": "vivo-ai/BlueLM-7B-Chat-32k",
+
+ "internlm-7b": "internlm/internlm-7b",
+ "internlm-chat-7b": "internlm/internlm-chat-7b",
+ "internlm2-chat-7b": "internlm/Models/internlm2-chat-7b",
+ "internlm2-chat-20b": "internlm/Models/internlm2-chat-20b",
+
+ "aquila-7b": "BAAI/Aquila-7B",
+ "aquilachat-7b": "BAAI/AquilaChat-7B",
+
+ "falcon-7b": "tiiuae/falcon-7b",
+ "falcon-40b": "tiiuae/falcon-40b",
+ "falcon-rw-7b": "tiiuae/falcon-rw-7b",
+ "gpt2": "gpt2",
+ "gpt2-xl": "gpt2-xl",
+ "gpt-j-6b": "EleutherAI/gpt-j-6b",
+ "gpt4all-j": "nomic-ai/gpt4all-j",
+ "gpt-neox-20b": "EleutherAI/gpt-neox-20b",
+ "pythia-12b": "EleutherAI/pythia-12b",
+ "oasst-sft-4-pythia-12b-epoch-3.5": "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5",
+ "dolly-v2-12b": "databricks/dolly-v2-12b",
+ "stablelm-tuned-alpha-7b": "stabilityai/stablelm-tuned-alpha-7b",
+ "open_llama_13b": "openlm-research/open_llama_13b",
+ "vicuna-13b-v1.3": "lmsys/vicuna-13b-v1.3",
+ "koala": "young-geng/koala",
+ "mpt-7b": "mosaicml/mpt-7b",
+ "mpt-7b-storywriter": "mosaicml/mpt-7b-storywriter",
+ "mpt-30b": "mosaicml/mpt-30b",
+ "opt-66b": "facebook/opt-66b",
+ "opt-iml-max-30b": "facebook/opt-iml-max-30b",
+
+}
+
+SUPPORT_AGENT_MODEL = [
+ "openai-api", # GPT4 模型
+ "qwen-api", # Qwen Max模型
+ "zhipu-api", # 智谱AI GLM4模型
+ "Qwen", # 所有Qwen系列本地模型
+ "chatglm3-6b",
+ "internlm2-chat-20b",
+ "Orion-14B-Chat-Plugin",
+]
diff --git a/configs/prompt_config.py.example b/configs/prompt_config.py.example
new file mode 100644
index 0000000000000000000000000000000000000000..6fb6996c1cd32f4efa15122d433ed9de12a66702
--- /dev/null
+++ b/configs/prompt_config.py.example
@@ -0,0 +1,127 @@
+# prompt模板使用Jinja2语法,简单点就是用双大括号代替f-string的单大括号
+# 本配置文件支持热加载,修改prompt模板后无需重启服务。
+
+# LLM对话支持的变量:
+# - input: 用户输入内容
+
+# 知识库和搜索引擎对话支持的变量:
+# - context: 从检索结果拼接的知识文本
+# - question: 用户提出的问题
+
+# Agent对话支持的变量:
+
+# - tools: 可用的工具列表
+# - tool_names: 可用的工具名称列表
+# - history: 用户和Agent的对话历史
+# - input: 用户输入内容
+# - agent_scratchpad: Agent的思维记录
+
+PROMPT_TEMPLATES = {
+ "llm_chat": {
+ "default":
+ '{{ input }}',
+
+ "with_history":
+ 'The following is a friendly conversation between a human and an AI. '
+ 'The AI is talkative and provides lots of specific details from its context. '
+ 'If the AI does not know the answer to a question, it truthfully says it does not know.\n\n'
+ 'Current conversation:\n'
+ '{history}\n'
+ 'Human: {input}\n'
+ 'AI:',
+
+ "py":
+ '你是一个聪明的代码助手,请你给我写出简单的py代码。 \n'
+ '{{ input }}',
+ },
+
+
+ "knowledge_base_chat": {
+ "default":
+ '<指令>根据已知信息,简洁和专业的来回答问题。如果无法从中得到答案,请说 “根据已知信息无法回答该问题”,'
+ '不允许在答案中添加编造成分,答案请使用中文。 指令>\n'
+ '<已知信息>{{ context }}已知信息>\n'
+ '<问题>{{ question }}问题>\n',
+
+ "text":
+ '<指令>根据已知信息,简洁和专业的来回答问题。如果无法从中得到答案,请说 “根据已知信息无法回答该问题”,答案请使用中文。 指令>\n'
+ '<已知信息>{{ context }}已知信息>\n'
+ '<问题>{{ question }}问题>\n',
+
+ "empty": # 搜不到知识库的时候使用
+ '请你回答我的问题:\n'
+ '{{ question }}\n\n',
+ },
+
+
+ "search_engine_chat": {
+ "default":
+ '<指令>这是我搜索到的互联网信息,请你根据这些信息进行提取并有调理,简洁的回答问题。'
+ '如果无法从中得到答案,请说 “无法搜索到能回答问题的内容”。 指令>\n'
+ '<已知信息>{{ context }}已知信息>\n'
+ '<问题>{{ question }}问题>\n',
+
+ "search":
+ '<指令>根据已知信息,简洁和专业的来回答问题。如果无法从中得到答案,请说 “根据已知信息无法回答该问题”,答案请使用中文。 指令>\n'
+ '<已知信息>{{ context }}已知信息>\n'
+ '<问题>{{ question }}问题>\n',
+ },
+
+
+ "agent_chat": {
+ "default":
+ 'Answer the following questions as best you can. If it is in order, you can use some tools appropriately. '
+ 'You have access to the following tools:\n\n'
+ '{tools}\n\n'
+ 'Use the following format:\n'
+ 'Question: the input question you must answer1\n'
+ 'Thought: you should always think about what to do and what tools to use.\n'
+ 'Action: the action to take, should be one of [{tool_names}]\n'
+ 'Action Input: the input to the action\n'
+ 'Observation: the result of the action\n'
+ '... (this Thought/Action/Action Input/Observation can be repeated zero or more times)\n'
+ 'Thought: I now know the final answer\n'
+ 'Final Answer: the final answer to the original input question\n'
+ 'Begin!\n\n'
+ 'history: {history}\n\n'
+ 'Question: {input}\n\n'
+ 'Thought: {agent_scratchpad}\n',
+
+ "ChatGLM3":
+ 'You can answer using the tools, or answer directly using your knowledge without using the tools. '
+ 'Respond to the human as helpfully and accurately as possible.\n'
+ 'You have access to the following tools:\n'
+ '{tools}\n'
+ 'Use a json blob to specify a tool by providing an action key (tool name) '
+ 'and an action_input key (tool input).\n'
+ 'Valid "action" values: "Final Answer" or [{tool_names}]'
+ 'Provide only ONE action per $JSON_BLOB, as shown:\n\n'
+ '```\n'
+ '{{{{\n'
+ ' "action": $TOOL_NAME,\n'
+ ' "action_input": $INPUT\n'
+ '}}}}\n'
+ '```\n\n'
+ 'Follow this format:\n\n'
+ 'Question: input question to answer\n'
+ 'Thought: consider previous and subsequent steps\n'
+ 'Action:\n'
+ '```\n'
+ '$JSON_BLOB\n'
+ '```\n'
+ 'Observation: action result\n'
+ '... (repeat Thought/Action/Observation N times)\n'
+ 'Thought: I know what to respond\n'
+ 'Action:\n'
+ '```\n'
+ '{{{{\n'
+ ' "action": "Final Answer",\n'
+ ' "action_input": "Final response to human"\n'
+ '}}}}\n'
+ 'Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. '
+ 'Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.\n'
+ 'history: {history}\n\n'
+ 'Question: {input}\n\n'
+ 'Thought: {agent_scratchpad}',
+ }
+}
diff --git a/configs/server_config.py.example b/configs/server_config.py.example
new file mode 100644
index 0000000000000000000000000000000000000000..eea9c34dfa317e616523a247baf638470695f5b7
--- /dev/null
+++ b/configs/server_config.py.example
@@ -0,0 +1,137 @@
+import sys
+from configs.model_config import LLM_DEVICE
+
+# httpx 请求默认超时时间(秒)。如果加载模型或对话较慢,出现超时错误,可以适当加大该值。
+HTTPX_DEFAULT_TIMEOUT = 300.0
+
+# API 是否开启跨域,默认为False,如果需要开启,请设置为True
+# is open cross domain
+OPEN_CROSS_DOMAIN = False
+
+# 各服务器默认绑定host。如改为"0.0.0.0"需要修改下方所有XX_SERVER的host
+DEFAULT_BIND_HOST = "0.0.0.0" if sys.platform != "win32" else "127.0.0.1"
+
+# webui.py server
+WEBUI_SERVER = {
+ "host": DEFAULT_BIND_HOST,
+ "port": 8501,
+}
+
+# api.py server
+API_SERVER = {
+ "host": DEFAULT_BIND_HOST,
+ "port": 7861,
+}
+
+# fastchat openai_api server
+FSCHAT_OPENAI_API = {
+ "host": DEFAULT_BIND_HOST,
+ "port": 20000,
+}
+
+# fastchat model_worker server
+# 这些模型必须是在model_config.MODEL_PATH或ONLINE_MODEL中正确配置的。
+# 在启动startup.py时,可用通过`--model-name xxxx yyyy`指定模型,不指定则为LLM_MODELS
+FSCHAT_MODEL_WORKERS = {
+ # 所有模型共用的默认配置,可在模型专项配置中进行覆盖。
+ "default": {
+ "host": DEFAULT_BIND_HOST,
+ "port": 20002,
+ "device": LLM_DEVICE,
+ # False,'vllm',使用的推理加速框架,使用vllm如果出现HuggingFace通信问题,参见doc/FAQ
+ # vllm对一些模型支持还不成熟,暂时默认关闭
+ "infer_turbo": False,
+
+ # model_worker多卡加载需要配置的参数
+ # "gpus": None, # 使用的GPU,以str的格式指定,如"0,1",如失效请使用CUDA_VISIBLE_DEVICES="0,1"等形式指定
+ # "num_gpus": 1, # 使用GPU的数量
+ # "max_gpu_memory": "20GiB", # 每个GPU占用的最大显存
+
+ # 以下为model_worker非常用参数,可根据需要配置
+ # "load_8bit": False, # 开启8bit量化
+ # "cpu_offloading": None,
+ # "gptq_ckpt": None,
+ # "gptq_wbits": 16,
+ # "gptq_groupsize": -1,
+ # "gptq_act_order": False,
+ # "awq_ckpt": None,
+ # "awq_wbits": 16,
+ # "awq_groupsize": -1,
+ # "model_names": LLM_MODELS,
+ # "conv_template": None,
+ # "limit_worker_concurrency": 5,
+ # "stream_interval": 2,
+ # "no_register": False,
+ # "embed_in_truncate": False,
+
+ # 以下为vllm_worker配置参数,注意使用vllm必须有gpu,仅在Linux测试通过
+
+ # tokenizer = model_path # 如果tokenizer与model_path不一致在此处添加
+ # 'tokenizer_mode':'auto',
+ # 'trust_remote_code':True,
+ # 'download_dir':None,
+ # 'load_format':'auto',
+ # 'dtype':'auto',
+ # 'seed':0,
+ # 'worker_use_ray':False,
+ # 'pipeline_parallel_size':1,
+ # 'tensor_parallel_size':1,
+ # 'block_size':16,
+ # 'swap_space':4 , # GiB
+ # 'gpu_memory_utilization':0.90,
+ # 'max_num_batched_tokens':2560,
+ # 'max_num_seqs':256,
+ # 'disable_log_stats':False,
+ # 'conv_template':None,
+ # 'limit_worker_concurrency':5,
+ # 'no_register':False,
+ # 'num_gpus': 1
+ # 'engine_use_ray': False,
+ # 'disable_log_requests': False
+
+ },
+ "Qwen-1_8B-Chat": {
+ "device": "cpu",
+ },
+ "chatglm3-6b": {
+ "device": "cuda",
+ },
+
+ # 以下配置可以不用修改,在model_config中设置启动的模型
+ "zhipu-api": {
+ "port": 21001,
+ },
+ "minimax-api": {
+ "port": 21002,
+ },
+ "xinghuo-api": {
+ "port": 21003,
+ },
+ "qianfan-api": {
+ "port": 21004,
+ },
+ "fangzhou-api": {
+ "port": 21005,
+ },
+ "qwen-api": {
+ "port": 21006,
+ },
+ "baichuan-api": {
+ "port": 21007,
+ },
+ "azure-api": {
+ "port": 21008,
+ },
+ "tiangong-api": {
+ "port": 21009,
+ },
+ "gemini-api": {
+ "port": 21010,
+ },
+}
+
+FSCHAT_CONTROLLER = {
+ "host": DEFAULT_BIND_HOST,
+ "port": 20001,
+ "dispatch_method": "shortest_queue",
+}
diff --git "a/docs/ES\351\203\250\347\275\262\346\214\207\345\215\227.md" "b/docs/ES\351\203\250\347\275\262\346\214\207\345\215\227.md"
new file mode 100644
index 0000000000000000000000000000000000000000..f461582675dea2ec3fb261370299d867fdce7dd7
--- /dev/null
+++ "b/docs/ES\351\203\250\347\275\262\346\214\207\345\215\227.md"
@@ -0,0 +1,29 @@
+
+# 实现基于ES的数据插入、检索、删除、更新
+```shell
+author: 唐国梁Tommy
+e-mail: flytang186@qq.com
+
+如果遇到任何问题,可以与我联系,我这边部署后服务是没有问题的。
+```
+
+## 第1步:ES docker部署
+```shell
+docker network create elastic
+docker run -id --name elasticsearch --net elastic -p 9200:9200 -p 9300:9300 -e "discovery.type=single-node" -e "xpack.security.enabled=false" -e "xpack.security.http.ssl.enabled=false" -t docker.elastic.co/elasticsearch/elasticsearch:8.8.2
+```
+
+### 第2步:Kibana docker部署
+**注意:Kibana版本与ES保持一致**
+```shell
+docker pull docker.elastic.co/kibana/kibana:{version}
+docker run --name kibana --net elastic -p 5601:5601 docker.elastic.co/kibana/kibana:{version}
+```
+
+### 第3步:核心代码
+```shell
+1. 核心代码路径
+server/knowledge_base/kb_service/es_kb_service.py
+
+2. 需要在 configs/model_config.py 中 配置 ES参数(IP, PORT)等;
+```
\ No newline at end of file
diff --git a/document_loaders/FilteredCSVloader.py b/document_loaders/FilteredCSVloader.py
new file mode 100644
index 0000000000000000000000000000000000000000..d9ca508b9b6bde552e4ff72915c7bd087761cbfc
--- /dev/null
+++ b/document_loaders/FilteredCSVloader.py
@@ -0,0 +1,81 @@
+## 指定制定列的csv文件加载器
+
+from langchain.document_loaders import CSVLoader
+import csv
+from io import TextIOWrapper
+from typing import Dict, List, Optional
+from langchain.docstore.document import Document
+from langchain.document_loaders.helpers import detect_file_encodings
+
+
+class FilteredCSVLoader(CSVLoader):
+ def __init__(
+ self,
+ file_path: str,
+ columns_to_read: List[str],
+ source_column: Optional[str] = None,
+ metadata_columns: List[str] = [],
+ csv_args: Optional[Dict] = None,
+ encoding: Optional[str] = None,
+ autodetect_encoding: bool = False,
+ ):
+ super().__init__(
+ file_path=file_path,
+ source_column=source_column,
+ metadata_columns=metadata_columns,
+ csv_args=csv_args,
+ encoding=encoding,
+ autodetect_encoding=autodetect_encoding,
+ )
+ self.columns_to_read = columns_to_read
+
+ def load(self) -> List[Document]:
+ """Load data into document objects."""
+
+ docs = []
+ try:
+ with open(self.file_path, newline="", encoding=self.encoding) as csvfile:
+ docs = self.__read_file(csvfile)
+ except UnicodeDecodeError as e:
+ if self.autodetect_encoding:
+ detected_encodings = detect_file_encodings(self.file_path)
+ for encoding in detected_encodings:
+ try:
+ with open(
+ self.file_path, newline="", encoding=encoding.encoding
+ ) as csvfile:
+ docs = self.__read_file(csvfile)
+ break
+ except UnicodeDecodeError:
+ continue
+ else:
+ raise RuntimeError(f"Error loading {self.file_path}") from e
+ except Exception as e:
+ raise RuntimeError(f"Error loading {self.file_path}") from e
+
+ return docs
+
+ def __read_file(self, csvfile: TextIOWrapper) -> List[Document]:
+ docs = []
+ csv_reader = csv.DictReader(csvfile, **self.csv_args) # type: ignore
+ for i, row in enumerate(csv_reader):
+ if self.columns_to_read[0] in row:
+ content = row[self.columns_to_read[0]]
+ # Extract the source if available
+ source = (
+ row.get(self.source_column, None)
+ if self.source_column is not None
+ else self.file_path
+ )
+ metadata = {"source": source, "row": i}
+
+ for col in self.metadata_columns:
+ if col in row:
+ metadata[col] = row[col]
+
+ doc = Document(page_content=content, metadata=metadata)
+ docs.append(doc)
+ else:
+ raise ValueError(f"Column '{self.columns_to_read[0]}' not found in CSV file.")
+
+ return docs
diff --git a/document_loaders/__init__.py b/document_loaders/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..88cfeae81c3bf03ca16b4268eceb500321589c7c
--- /dev/null
+++ b/document_loaders/__init__.py
@@ -0,0 +1,4 @@
+from .mypdfloader import RapidOCRPDFLoader
+from .myimgloader import RapidOCRLoader
+from .mydocloader import RapidOCRDocLoader
+from .mypptloader import RapidOCRPPTLoader
diff --git a/document_loaders/mydocloader.py b/document_loaders/mydocloader.py
new file mode 100644
index 0000000000000000000000000000000000000000..7f5462a2e996175600d49895899eb8bc9dba4f49
--- /dev/null
+++ b/document_loaders/mydocloader.py
@@ -0,0 +1,71 @@
+from langchain.document_loaders.unstructured import UnstructuredFileLoader
+from typing import List
+import tqdm
+
+
+class RapidOCRDocLoader(UnstructuredFileLoader):
+ def _get_elements(self) -> List:
+ def doc2text(filepath):
+ from docx.table import _Cell, Table
+ from docx.oxml.table import CT_Tbl
+ from docx.oxml.text.paragraph import CT_P
+ from docx.text.paragraph import Paragraph
+ from docx import Document, ImagePart
+ from PIL import Image
+ from io import BytesIO
+ import numpy as np
+ from rapidocr_onnxruntime import RapidOCR
+ ocr = RapidOCR()
+ doc = Document(filepath)
+ resp = ""
+
+ def iter_block_items(parent):
+ from docx.document import Document
+ if isinstance(parent, Document):
+ parent_elm = parent.element.body
+ elif isinstance(parent, _Cell):
+ parent_elm = parent._tc
+ else:
+ raise ValueError("RapidOCRDocLoader parse fail")
+
+ for child in parent_elm.iterchildren():
+ if isinstance(child, CT_P):
+ yield Paragraph(child, parent)
+ elif isinstance(child, CT_Tbl):
+ yield Table(child, parent)
+
+ b_unit = tqdm.tqdm(total=len(doc.paragraphs)+len(doc.tables),
+ desc="RapidOCRDocLoader block index: 0")
+ for i, block in enumerate(iter_block_items(doc)):
+ b_unit.set_description(
+ "RapidOCRDocLoader block index: {}".format(i))
+ b_unit.refresh()
+ if isinstance(block, Paragraph):
+ resp += block.text.strip() + "\n"
+ images = block._element.xpath('.//pic:pic') # 获取所有图片
+ for image in images:
+ for img_id in image.xpath('.//a:blip/@r:embed'): # 获取图片id
+ part = doc.part.related_parts[img_id] # 根据图片id获取对应的图片
+ if isinstance(part, ImagePart):
+ image = Image.open(BytesIO(part._blob))
+ result, _ = ocr(np.array(image))
+ if result:
+ ocr_result = [line[1] for line in result]
+ resp += "\n".join(ocr_result)
+ elif isinstance(block, Table):
+ for row in block.rows:
+ for cell in row.cells:
+ for paragraph in cell.paragraphs:
+ resp += paragraph.text.strip() + "\n"
+ b_unit.update(1)
+ return resp
+
+ text = doc2text(self.file_path)
+ from unstructured.partition.text import partition_text
+ return partition_text(text=text, **self.unstructured_kwargs)
+
+
+if __name__ == '__main__':
+ loader = RapidOCRDocLoader(file_path="../tests/samples/ocr_test.docx")
+ docs = loader.load()
+ print(docs)
diff --git a/document_loaders/myimgloader.py b/document_loaders/myimgloader.py
new file mode 100644
index 0000000000000000000000000000000000000000..e09c61720da9c5cee6651bb6e9cdb988b73b71e6
--- /dev/null
+++ b/document_loaders/myimgloader.py
@@ -0,0 +1,25 @@
+from typing import List
+from langchain.document_loaders.unstructured import UnstructuredFileLoader
+from document_loaders.ocr import get_ocr
+
+
+class RapidOCRLoader(UnstructuredFileLoader):
+ def _get_elements(self) -> List:
+ def img2text(filepath):
+ resp = ""
+ ocr = get_ocr()
+ result, _ = ocr(filepath)
+ if result:
+ ocr_result = [line[1] for line in result]
+ resp += "\n".join(ocr_result)
+ return resp
+
+ text = img2text(self.file_path)
+ from unstructured.partition.text import partition_text
+ return partition_text(text=text, **self.unstructured_kwargs)
+
+
+if __name__ == "__main__":
+ loader = RapidOCRLoader(file_path="../tests/samples/ocr_test.jpg")
+ docs = loader.load()
+ print(docs)
diff --git a/document_loaders/mypdfloader.py b/document_loaders/mypdfloader.py
new file mode 100644
index 0000000000000000000000000000000000000000..faaf63dd4189ab59e9c020298c88eb0f4a916389
--- /dev/null
+++ b/document_loaders/mypdfloader.py
@@ -0,0 +1,51 @@
+from typing import List
+from langchain.document_loaders.unstructured import UnstructuredFileLoader
+from configs import PDF_OCR_THRESHOLD
+from document_loaders.ocr import get_ocr
+import tqdm
+
+
+class RapidOCRPDFLoader(UnstructuredFileLoader):
+ def _get_elements(self) -> List:
+ def pdf2text(filepath):
+ import fitz # pyMuPDF里面的fitz包,不要与pip install fitz混淆
+ import numpy as np
+ ocr = get_ocr()
+ doc = fitz.open(filepath)
+ resp = ""
+
+ b_unit = tqdm.tqdm(total=doc.page_count, desc="RapidOCRPDFLoader context page index: 0")
+ for i, page in enumerate(doc):
+ b_unit.set_description("RapidOCRPDFLoader context page index: {}".format(i))
+ b_unit.refresh()
+ text = page.get_text("")
+ resp += text + "\n"
+
+ img_list = page.get_image_info(xrefs=True)
+ for img in img_list:
+ if xref := img.get("xref"):
+ bbox = img["bbox"]
+ # 检查图片尺寸是否超过设定的阈值
+ if ((bbox[2] - bbox[0]) / (page.rect.width) < PDF_OCR_THRESHOLD[0]
+ or (bbox[3] - bbox[1]) / (page.rect.height) < PDF_OCR_THRESHOLD[1]):
+ continue
+ pix = fitz.Pixmap(doc, xref)
+ img_array = np.frombuffer(pix.samples, dtype=np.uint8).reshape(pix.height, pix.width, -1)
+ result, _ = ocr(img_array)
+ if result:
+ ocr_result = [line[1] for line in result]
+ resp += "\n".join(ocr_result)
+
+ # 更新进度
+ b_unit.update(1)
+ return resp
+
+ text = pdf2text(self.file_path)
+ from unstructured.partition.text import partition_text
+ return partition_text(text=text, **self.unstructured_kwargs)
+
+
+if __name__ == "__main__":
+ loader = RapidOCRPDFLoader(file_path="../tests/samples/ocr_test.pdf")
+ docs = loader.load()
+ print(docs)
diff --git a/document_loaders/mypptloader.py b/document_loaders/mypptloader.py
new file mode 100644
index 0000000000000000000000000000000000000000..f14d0728e8602cc649da3108397aed4da13fd661
--- /dev/null
+++ b/document_loaders/mypptloader.py
@@ -0,0 +1,59 @@
+from langchain.document_loaders.unstructured import UnstructuredFileLoader
+from typing import List
+import tqdm
+
+
+class RapidOCRPPTLoader(UnstructuredFileLoader):
+ def _get_elements(self) -> List:
+ def ppt2text(filepath):
+ from pptx import Presentation
+ from PIL import Image
+ import numpy as np
+ from io import BytesIO
+ from rapidocr_onnxruntime import RapidOCR
+ ocr = RapidOCR()
+ prs = Presentation(filepath)
+ resp = ""
+
+ def extract_text(shape):
+ nonlocal resp
+ if shape.has_text_frame:
+ resp += shape.text.strip() + "\n"
+ if shape.has_table:
+ for row in shape.table.rows:
+ for cell in row.cells:
+ for paragraph in cell.text_frame.paragraphs:
+ resp += paragraph.text.strip() + "\n"
+ if shape.shape_type == 13: # 13 表示图片
+ image = Image.open(BytesIO(shape.image.blob))
+ result, _ = ocr(np.array(image))
+ if result:
+ ocr_result = [line[1] for line in result]
+ resp += "\n".join(ocr_result)
+ elif shape.shape_type == 6: # 6 表示组合
+ for child_shape in shape.shapes:
+ extract_text(child_shape)
+
+ b_unit = tqdm.tqdm(total=len(prs.slides),
+ desc="RapidOCRPPTLoader slide index: 1")
+ # 遍历所有幻灯片
+ for slide_number, slide in enumerate(prs.slides, start=1):
+ b_unit.set_description(
+ "RapidOCRPPTLoader slide index: {}".format(slide_number))
+ b_unit.refresh()
+ sorted_shapes = sorted(slide.shapes,
+ key=lambda x: (x.top, x.left)) # 从上到下、从左到右遍历
+ for shape in sorted_shapes:
+ extract_text(shape)
+ b_unit.update(1)
+ return resp
+
+ text = ppt2text(self.file_path)
+ from unstructured.partition.text import partition_text
+ return partition_text(text=text, **self.unstructured_kwargs)
+
+
+if __name__ == '__main__':
+ loader = RapidOCRPPTLoader(file_path="../tests/samples/ocr_test.pptx")
+ docs = loader.load()
+ print(docs)
diff --git a/document_loaders/ocr.py b/document_loaders/ocr.py
new file mode 100644
index 0000000000000000000000000000000000000000..2b66dd35701a1cdaf0c037eae0ab2d50e55a7b04
--- /dev/null
+++ b/document_loaders/ocr.py
@@ -0,0 +1,18 @@
+from typing import TYPE_CHECKING
+
+
+if TYPE_CHECKING:
+ try:
+ from rapidocr_paddle import RapidOCR
+ except ImportError:
+ from rapidocr_onnxruntime import RapidOCR
+
+
+def get_ocr(use_cuda: bool = True) -> "RapidOCR":
+ try:
+ from rapidocr_paddle import RapidOCR
+ ocr = RapidOCR(det_use_cuda=use_cuda, cls_use_cuda=use_cuda, rec_use_cuda=use_cuda)
+ except ImportError:
+ from rapidocr_onnxruntime import RapidOCR
+ ocr = RapidOCR()
+ return ocr
diff --git a/embeddings/__init__.py b/embeddings/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/embeddings/add_embedding_keywords.py b/embeddings/add_embedding_keywords.py
new file mode 100644
index 0000000000000000000000000000000000000000..fe6a8233b53df4000ac45f657f8daf0c69b6d25c
--- /dev/null
+++ b/embeddings/add_embedding_keywords.py
@@ -0,0 +1,79 @@
+'''
+该功能是为了将关键词加入到embedding模型中,以便于在embedding模型中进行关键词的embedding
+该功能的实现是通过修改embedding模型的tokenizer来实现的
+该功能仅仅对EMBEDDING_MODEL参数对应的的模型有效,输出后的模型保存在原本模型
+感谢@CharlesJu1和@charlesyju的贡献提出了想法和最基础的PR
+
+保存的模型的位置位于原本嵌入模型的目录下,模型的名称为原模型名称+Merge_Keywords_时间戳
+'''
+import sys
+
+sys.path.append("..")
+import os
+import torch
+
+from datetime import datetime
+from configs import (
+ MODEL_PATH,
+ EMBEDDING_MODEL,
+ EMBEDDING_KEYWORD_FILE,
+)
+
+from safetensors.torch import save_model
+from sentence_transformers import SentenceTransformer
+from langchain_core._api import deprecated
+
+
+@deprecated(
+ since="0.3.0",
+ message="自定义关键词 Langchain-Chatchat 0.3.x 重写, 0.2.x中相关功能将废弃",
+ removal="0.3.0"
+ )
+def get_keyword_embedding(bert_model, tokenizer, key_words):
+ tokenizer_output = tokenizer(key_words, return_tensors="pt", padding=True, truncation=True)
+ input_ids = tokenizer_output['input_ids']
+ input_ids = input_ids[:, 1:-1]
+
+ keyword_embedding = bert_model.embeddings.word_embeddings(input_ids)
+ keyword_embedding = torch.mean(keyword_embedding, 1)
+ return keyword_embedding
+
+
+def add_keyword_to_model(model_name=EMBEDDING_MODEL, keyword_file: str = "", output_model_path: str = None):
+ key_words = []
+ with open(keyword_file, "r") as f:
+ for line in f:
+ key_words.append(line.strip())
+
+ st_model = SentenceTransformer(model_name)
+ key_words_len = len(key_words)
+ word_embedding_model = st_model._first_module()
+ bert_model = word_embedding_model.auto_model
+ tokenizer = word_embedding_model.tokenizer
+ key_words_embedding = get_keyword_embedding(bert_model, tokenizer, key_words)
+
+ embedding_weight = bert_model.embeddings.word_embeddings.weight
+ embedding_weight_len = len(embedding_weight)
+ tokenizer.add_tokens(key_words)
+ bert_model.resize_token_embeddings(len(tokenizer), pad_to_multiple_of=32)
+ embedding_weight = bert_model.embeddings.word_embeddings.weight
+ with torch.no_grad():
+ embedding_weight[embedding_weight_len:embedding_weight_len + key_words_len, :] = key_words_embedding
+
+ if output_model_path:
+ os.makedirs(output_model_path, exist_ok=True)
+ word_embedding_model.save(output_model_path)
+ safetensors_file = os.path.join(output_model_path, "model.safetensors")
+ metadata = {'format': 'pt'}
+ save_model(bert_model, safetensors_file, metadata)
+ print("save model to {}".format(output_model_path))
+
+
+def add_keyword_to_embedding_model(path: str = EMBEDDING_KEYWORD_FILE):
+ keyword_file = os.path.join(path)
+ model_name = MODEL_PATH["embed_model"][EMBEDDING_MODEL]
+ model_parent_directory = os.path.dirname(model_name)
+ current_time = datetime.now().strftime('%Y%m%d_%H%M%S')
+ output_model_name = "{}_Merge_Keywords_{}".format(EMBEDDING_MODEL, current_time)
+ output_model_path = os.path.join(model_parent_directory, output_model_name)
+ add_keyword_to_model(model_name, keyword_file, output_model_path)
diff --git a/embeddings/embedding_keywords.txt b/embeddings/embedding_keywords.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3822b992f9ec1b649b3a17a5d3166737aacc10b3
--- /dev/null
+++ b/embeddings/embedding_keywords.txt
@@ -0,0 +1,3 @@
+Langchain-Chatchat
+数据科学与大数据技术
+人工智能与先进计算
\ No newline at end of file
diff --git a/img/LLM_success.png b/img/LLM_success.png
new file mode 100644
index 0000000000000000000000000000000000000000..48bd2746ce3a0bfc5532bc31801154e579169eda
Binary files /dev/null and b/img/LLM_success.png differ
diff --git a/img/agent_continue.png b/img/agent_continue.png
new file mode 100644
index 0000000000000000000000000000000000000000..be02fd7dc798ab0bb51ea716975fb62a8d352bca
Binary files /dev/null and b/img/agent_continue.png differ
diff --git a/img/agent_success.png b/img/agent_success.png
new file mode 100644
index 0000000000000000000000000000000000000000..5c2e3104fa424da37e460aae0a12cd11de735753
Binary files /dev/null and b/img/agent_success.png differ
diff --git a/img/chatchat-qrcode.jpg b/img/chatchat-qrcode.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a16a6081c327d22d0e64ccbe5a196d20f7f0027c
Binary files /dev/null and b/img/chatchat-qrcode.jpg differ
diff --git a/img/chatchat_icon_blue_square_v2.png b/img/chatchat_icon_blue_square_v2.png
new file mode 100644
index 0000000000000000000000000000000000000000..5de8ee80b5f219484c90b0b4604cfd1e85b58e3d
Binary files /dev/null and b/img/chatchat_icon_blue_square_v2.png differ
diff --git a/img/docker_logs.png b/img/docker_logs.png
new file mode 100644
index 0000000000000000000000000000000000000000..038295822880d041d1e4619a1fc5f2c2bc0b37a3
Binary files /dev/null and b/img/docker_logs.png differ
diff --git a/img/fastapi_docs_026.png b/img/fastapi_docs_026.png
new file mode 100644
index 0000000000000000000000000000000000000000..92a2cf35d82ae8b7a6743888b97852e0fcf170ed
Binary files /dev/null and b/img/fastapi_docs_026.png differ
diff --git a/img/init_knowledge_base.jpg b/img/init_knowledge_base.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a031b7718c110bda23f4446cd2b57e7b0839c06a
Binary files /dev/null and b/img/init_knowledge_base.jpg differ
diff --git a/img/knowledge_base_success.jpg b/img/knowledge_base_success.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6639ee976ac341f9fb151840e5257b717679fb15
Binary files /dev/null and b/img/knowledge_base_success.jpg differ
diff --git a/img/langchain+chatglm.png b/img/langchain+chatglm.png
new file mode 100644
index 0000000000000000000000000000000000000000..ba10abfaa9b737f57ee02b6efaf1add5ab387074
--- /dev/null
+++ b/img/langchain+chatglm.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ae4af8281129ba13033d172ce0556baf2c5f4b07f1bcf50ec233082266208b5
+size 1120970
diff --git a/img/langchain+chatglm2.png b/img/langchain+chatglm2.png
new file mode 100644
index 0000000000000000000000000000000000000000..d98b8101be6b9d634911f96ba6bf47fe011649f7
Binary files /dev/null and b/img/langchain+chatglm2.png differ
diff --git a/img/logo-long-chatchat-trans-v2.png b/img/logo-long-chatchat-trans-v2.png
new file mode 100644
index 0000000000000000000000000000000000000000..7a689e8c2e6058a31006d45d20d814a62f019bfd
Binary files /dev/null and b/img/logo-long-chatchat-trans-v2.png differ
diff --git a/img/official_account_qr.png b/img/official_account_qr.png
new file mode 100644
index 0000000000000000000000000000000000000000..a16a6081c327d22d0e64ccbe5a196d20f7f0027c
Binary files /dev/null and b/img/official_account_qr.png differ
diff --git a/img/official_wechat_mp_account.png b/img/official_wechat_mp_account.png
new file mode 100644
index 0000000000000000000000000000000000000000..6c6509ad7d043ef2d7931470f3f8b9e276e90ed4
--- /dev/null
+++ b/img/official_wechat_mp_account.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:021285c88e22bf0976c5188c5717466fa10af23ada09d4210ccf88bc8df7516c
+size 4268298
diff --git a/img/partners/autodl.svg b/img/partners/autodl.svg
new file mode 100644
index 0000000000000000000000000000000000000000..38b46d31e8afe7b99abebbdc965cd793740cbe51
--- /dev/null
+++ b/img/partners/autodl.svg
@@ -0,0 +1,9 @@
+
diff --git a/img/partners/aws.svg b/img/partners/aws.svg
new file mode 100644
index 0000000000000000000000000000000000000000..7c29d2d0c056b17dc1bdcf9b1276923cf690e738
--- /dev/null
+++ b/img/partners/aws.svg
@@ -0,0 +1,9 @@
+
diff --git a/img/partners/chatglm.svg b/img/partners/chatglm.svg
new file mode 100644
index 0000000000000000000000000000000000000000..c9923316fc5c2a716660a7eb5b3032860bbac8ad
--- /dev/null
+++ b/img/partners/chatglm.svg
@@ -0,0 +1,55 @@
+
diff --git a/img/partners/zhenfund.svg b/img/partners/zhenfund.svg
new file mode 100644
index 0000000000000000000000000000000000000000..7cd7eb7682269d5ab71ae26ebca11e186f38201e
--- /dev/null
+++ b/img/partners/zhenfund.svg
@@ -0,0 +1,9 @@
+
diff --git a/img/qr_code_86.jpg b/img/qr_code_86.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f353a04df6fd71e2ebfe738ce00433196ca49247
Binary files /dev/null and b/img/qr_code_86.jpg differ
diff --git a/img/qr_code_87.jpg b/img/qr_code_87.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ce8fc77af66ed44cf2c62aeddedcf9f3936829ee
Binary files /dev/null and b/img/qr_code_87.jpg differ
diff --git a/img/qr_code_88.jpg b/img/qr_code_88.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f6d5ead110e662764bf144be86b79c0a7605d96e
Binary files /dev/null and b/img/qr_code_88.jpg differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-124076-270516.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-124076-270516.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..066a89164f87a71c0f28d424d7d7a688a4cca58a
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-124076-270516.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-20096-279847.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-20096-279847.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..7c7c460a09fb30de4cb6cca3e199d65921973c01
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-20096-279847.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-220157-552735.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-220157-552735.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..287c6f6de12ee9d1e7c899168c11e102e2691d1d
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-220157-552735.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-36114-765327.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-36114-765327.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..c2927f4eb4f594a8b4d85116a57023baa523079b
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-36114-765327.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-392521-261326.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-392521-261326.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..a420a7b42c6a5312fb10794535a325800febd54d
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-392521-261326.jpg"
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:434aeea6c4491658ff7f7555060f708bd326d0ecf6fa62d7ca261a6ec845817a
+size 1089669
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-42284-124759.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-42284-124759.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..fec5b774683a90bcf2cd204a47369498a725ef49
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-42284-124759.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-57107-679259.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-57107-679259.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..9c0cddbbf94a9df24fb02f72d459463ab2c04600
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-57107-679259.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-618350-869132.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-618350-869132.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..2e20c3c5f503d3173cebfebd63aaf7e9e1f0e9fb
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-618350-869132.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-838373-426344.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-838373-426344.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..8fe62541d60153265d82f03c275f041099dedfee
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-838373-426344.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-906937-836104.jpg" "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-906937-836104.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..b2ad3ca537338b2bd36e750eaf19d77f00ed88b7
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-906937-836104.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-108319-429731.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-108319-429731.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..246bbab612aff163e784027b94b078f9d3417dae
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-108319-429731.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-580318-260070.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-580318-260070.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..fb9b36afcfe378186dc532567ecf827f4f2bfa50
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-580318-260070.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-793118-735987.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-793118-735987.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..4a9e505a9ab016d2d3da68a070a3b66db3b2fda2
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-793118-735987.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-918388-323086.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-918388-323086.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..d4c7d9d666956458aae04370b03f592c0caeae0c
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-918388-323086.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-19929-302935.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-19929-302935.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..3d0d0eb2b18b9380818017606025f3a3c2e000e8
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-19929-302935.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-299768-254064.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-299768-254064.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..096469aa4438fecf8c7a2e67d72250933dba2da2
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-299768-254064.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-454007-940199.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-454007-940199.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..13d85aa752c3af7717568e430efd0d50663654b3
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-454007-940199.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-628857-182232.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-628857-182232.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..548efc7bc40959912b55ed6db8067df083c70cfd
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-628857-182232.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-729151-372321.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-729151-372321.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..7a49477973b79e122f5e13a4c5039960097c4d42
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-729151-372321.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-81470-404273.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-81470-404273.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..8dfb2a6638679aeb04b4cde15f4885d97d7ae784
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-81470-404273.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-17565-176537.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-17565-176537.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..227de2c7ac43100a33ea920250c06b34563b1fb5
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-17565-176537.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-349153-657791.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-349153-657791.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..300029ad91b5e1e60fadd878cd1901f74e9028ed
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-349153-657791.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-350029-666381.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-350029-666381.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..09ef4094c2789c7dff3d4287f1674fc6d426ac25
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-350029-666381.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-759487-923925.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-759487-923925.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..6cf0cd6a3d89c82d26664e6c54c82964479ec523
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-759487-923925.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-805089-731888.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-805089-731888.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..a4adfa142e4df8ee8fa3365f0a521fd5e9f387ad
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-805089-731888.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-95996-523276.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-95996-523276.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..a48089b0d1c7c0da7b79da341096024f5378fda4
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206-\345\271\225\345\270\203\345\233\276\347\211\207-95996-523276.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-276446-401476.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-276446-401476.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..2cb96319cab50a06e5437c5896996ee7b1ebca66
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-276446-401476.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-380552-579242.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-380552-579242.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..d993b1bdce5e1e215be12362f52264c27e6c7d6e
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-380552-579242.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-590671-36787.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-590671-36787.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..cf6b095bdae68e0ddb587d58a782d3f3cd373080
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-590671-36787.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-699343-219844.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-699343-219844.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..22a61ccf13955e4be625feb9bf134f9c407515f4
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-699343-219844.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-789705-122117.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-789705-122117.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..7556834a551ee9bc77d94a1e18ab494c41d07705
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-789705-122117.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-923924-83386.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-923924-83386.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..d5160470b4637f3b29999df08785df59b7d7c1b8
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-923924-83386.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-930255-616209.jpg" "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-930255-616209.jpg"
new file mode 100644
index 0000000000000000000000000000000000000000..3c43116df5090bf4fdcc181b54490f83217ce0c1
Binary files /dev/null and "b/knowledge_base/samples/content/llm/img/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245-\345\271\225\345\270\203\345\233\276\347\211\207-930255-616209.jpg" differ
diff --git "a/knowledge_base/samples/content/llm/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206.md" "b/knowledge_base/samples/content/llm/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206.md"
new file mode 100644
index 0000000000000000000000000000000000000000..9d31f9495e165c975c5d33b2ffee984823458e67
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/\345\210\206\345\270\203\345\274\217\350\256\255\347\273\203\346\212\200\346\234\257\345\216\237\347\220\206.md"
@@ -0,0 +1,74 @@
+# 分布式训练技术原理
+- 数据并行
+ - FSDP
+ - FSDP算法是由来自DeepSpeed的ZeroRedundancyOptimizer技术驱动的,但经过修改的设计和实现与PyTorch的其他组件保持一致。FSDP将模型实例分解为更小的单元,然后将每个单元内的所有参数扁平化和分片。分片参数在计算前按需通信和恢复,计算结束后立即丢弃。这种方法确保FSDP每次只需要实现一个单元的参数,这大大降低了峰值内存消耗。(数据并行+Parameter切分)
+ - DDP
+ - DistributedDataParallel (DDP), **在每个设备上维护一个模型副本,并通过向后传递的集体AllReduce操作同步梯度,从而确保在训练期间跨副本的模型一致性** 。为了加快训练速度, **DDP将梯度通信与向后计算重叠** ,促进在不同资源上并发执行工作负载。
+ - ZeRO
+ - Model state
+ - Optimizer->ZeRO1
+ - 将optimizer state分成若干份,每块GPU上各自维护一份
+ - 每块GPU上存一份完整的参数W,做完一轮foward和backward后,各得一份梯度,对梯度做一次 **AllReduce(reduce-scatter + all-gather)** , **得到完整的梯度G,由于每块GPU上只保管部分optimizer states,因此只能将相应的W进行更新,对W做一次All-Gather**
+ - Gradient+Optimzer->ZeRO2
+ - 每个GPU维护一块梯度
+ - 每块GPU上存一份完整的参数W,做完一轮foward和backward后, **算得一份完整的梯度,对梯度做一次Reduce-Scatter,保证每个GPU上所维持的那块梯度是聚合梯度,每块GPU用自己对应的O和G去更新相应的W。更新完毕后,每块GPU维持了一块更新完毕的W。同理,对W做一次All-Gather,将别的GPU算好的W同步到自己这来**
+ - Parameter+Gradient+Optimizer->ZeRO3
+ - 每个GPU维护一块模型状态
+ - 每块GPU上只保存部分参数W,做forward时,对W做一次 **All-Gather** ,取回分布在别的GPU上的W,得到一份完整的W, **forward做完,立刻把不是自己维护的W抛弃,做backward时,对W做一次All-Gather,取回完整的W,backward做完,立刻把不是自己维护的W抛弃. 做完backward,算得一份完整的梯度G,对G做一次Reduce-Scatter,从别的GPU上聚合自己维护的那部分梯度,聚合操作结束后,立刻把不是自己维护的G抛弃。用自己维护的O和G,更新W。由于只维护部分W,因此无需再对W做任何AllReduce操作**
+ - Residual state
+ - activation->Partitioned Activation Checkpointing
+ - 每块GPU上只维护部分的activation,需要时再从别的地方聚合过来就行。需要注意的是,activation对显存的占用一般会远高于模型本身,通讯量也是巨大的
+ - temporary buffer->Constant Size Buffer
+ - 提升带宽利用率。当GPU数量上升,GPU间的通讯次数也上升,每次的通讯量可能下降(但总通讯量不会变)。数据切片小了,就不能很好利用带宽了。所以这个buffer起到了积攒数据的作用:等数据积攒到一定大小,再进行通讯。
+ - 使得存储大小可控。在每次通讯前,积攒的存储大小是常量,是已知可控的。更方便使用者对训练中的存储消耗和通讯时间进行预估
+ - unusable fragment->Memory Defragmentation
+ - 对碎片化的存储空间进行重新整合,整出连续的存储空间。防止出现总存储足够,但连续存储不够而引起的存储请求fail
+ - offload
+ - ZeRO-Offload
+ - **forward和backward计算量高** ,因此和它们相关的部分,例如参数W(fp16),activation,就全放入GPU
+ - **update的部分计算量低** ,因此和它相关的部分,全部放入CPU中。例如W(fp32),optimizer states(fp32)和gradients(fp16)等
+ - ZeRO-Offload 分为 Offload Strategy 和 Offload Schedule 两部分,前者解决如何在 GPU 和 CPU 间划分模型的问题,后者解决如何调度计算和通信的问题
+ - ZeRO-Infinity
+ - 一是将offload和 ZeRO 的结合从 ZeRO-2 延伸到了 ZeRO-3,解决了模型参数受限于单张 GPU 内存的问题
+ - 二是解决了 ZeRO-Offload 在训练 batch size 较小的时候效率较低的问题
+ - 三是除 CPU 内存外,进一步尝试利用 NVMe 的空间
+- 模型并行
+ - tensor-wise parallelism
+ - MLP切分
+ - 对第一个线性层按列切分,对第二个线性层按行切分
+ - ![图片](./img/分布式训练技术原理-幕布图片-36114-765327.jpg)
+ - ![图片](./img/分布式训练技术原理-幕布图片-392521-261326.jpg)
+ - ![图片](./img/分布式训练技术原理-幕布图片-57107-679259.jpg)
+ - self-attention切分
+ - attention的多头计算天然适合tensor并行,因为每个头上都可以独立计算最后再将结果concat起来,从而 **可以把每个头的参数放到一块GPU上**
+ - 对线性层, **按照“行切割”** 。切割的方式和MLP层基本一致,其forward与backward原理也一致
+ - 输入层Embedding切分
+ - 对positional embedding来说,max_s本身不会太长,因此每个GPU上都拷贝一份,对显存的压力也不会太大
+ - 将word embedding拆分到不同GPU上,每块GPU维护一分部词表。当输入X去GPU上查找时,能找到的词,就正常返回词向量,找到不到就把词向量中的全部全素都置0。按此方式查找完毕后,每块GPU上的数据做一次AllReduce,就能得到最终的输入。
+ - ![图片](./img/分布式训练技术原理-幕布图片-220157-552735.jpg)
+ - 输出层Embedding切分
+ - **输入层和输出层共用一个word embeding**
+ - **当模型的输入层到输入层都在一块GPU上时(即流水线并行深度=1),我们不必担心这点(实践中大部分用Megatron做并行的项目也是这么做的)。但若模型输入层和输出层在不同的GPU上时,我们就要保证在权重更新前,两块GPU上的word embedding梯度做了一次AllReduce** 。
+ - ![图片](./img/分布式训练技术原理-幕布图片-42284-124759.jpg)
+ - cross-entroy
+ - ![图片](./img/分布式训练技术原理-幕布图片-124076-270516.jpg)
+ - ![图片](./img/分布式训练技术原理-幕布图片-838373-426344.jpg)
+ - [pipeline paralelism]("https://zhuanlan.zhihu.com/p/629637468")
+ - GPipe
+ - PipeDream
+ - 1F1B
+ - 每个 GPU 以交替的方式执行每个 micro batch 的正向和反向过程,以尽早释放其占用的显存,进而减少显存占用
+ - ![图片](./img/分布式训练技术原理-幕布图片-20096-279847.jpg)
+ - 1F1B 并不能减少 bubble time, **为了进一步减少 bubble time,Megatron 又提出了 interleaved 1F1B 模式** 。也就是原本每个 GPU 负责连续 4 个层的计算,现在变成负责连续两个层的计算,只有原来的一半,从而 bubble time 也变成了原来的一半,即把一个设备上连续的层划分为若干不连续的层,负责的数量不变,但顺序变了。
+ - ![图片](./img/分布式训练技术原理-幕布图片-618350-869132.jpg)
+ - DAPPLE
+ - ![图片](./img/分布式训练技术原理-幕布图片-906937-836104.jpg)
+ - layer-wise parallelism
+ - sequence parallelism
+ - Sequence 并行的好处在于不会增加通信量,并且可以大大减少显存占用
+ - Layer-norm 和 Dropout 沿着序列的维度是独立的,因此可以按照 Sequence 维度进行拆分
+ - 使用了 Sequence 并行之后,对于超大规模的模型而言,其实显存占用量还是很大的。因此,Megatron 又引入了激活重计算技术,找到一些计算量很少但显存占用很大的算子,比如 Attention 里的 Softmax、Dropout 等算子,对这些算子进行激活重计算就可以显著减少显存,并且计算开销增加不大
+- MoE
+ - 核心思想:将大模型拆分成多个小模型。每个样本只需要激活部分专家模型进行计算,从而大大节省计算资源。 **MoE 的基本思路是通过宽度换取深度,因为模型深度越深,计算层数越多,进而推理时间越长**
+ - Hard Gate MoE
+ - Sparse MoE
\ No newline at end of file
diff --git "a/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206.md" "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206.md"
new file mode 100644
index 0000000000000000000000000000000000000000..66fc253ba7c4b5a5e7551a9c56290ec49c9324b3
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\345\272\224\347\224\250\346\212\200\346\234\257\345\216\237\347\220\206.md"
@@ -0,0 +1,97 @@
+# 大模型应用技术原理
+- RAG
+ - 向量数据库 [对比]("https://www.jianshu.com/p/43cc19426113")
+ - 选型标准
+ - 开源vs.闭源vs. 源码可见
+ - 客户端/SDK语言
+ - 托管方式
+ - self-hosted/on-premise
+ - redis,pgvector,milvus
+ - managed/cloud-native
+ - zilliz,pinecone
+ - embeded+cloud-native
+ - chroma,lanceDB
+ - self-hosted+cloud-native
+ - vald,drant,weaviate,vspa,elasticsearch
+ - 索引方法
+ - 算法
+ - Flat
+ - Tree-based
+ - Annoy(Approximate Nearest Neighbors Oh Yeah)
+ - KD-Tree
+ - Trinary Projection Trees
+ - IVF
+ - IVF
+ - IVMF(Inverted Multi-index File)
+ - Graph-based
+ - HNSW
+ - NSG
+ - Vamana(DiskANN)
+ - ![图片](./img/大模型应用技术原理-幕布图片-793118-735987.jpg)
+ - ![图片](./img/大模型应用技术原理-幕布图片-580318-260070.jpg)
+ - Hashing-based
+ - LSH
+ - Spherical Hashing
+ - Spectral Hashing
+ - 量化
+ - PQ(Product Quantization)
+ - PQ 将特征空间分解为多个低维子空间的笛卡尔乘积,然后单独地对每一个子空间进行量化
+ - SQ(Scalar Quantization)
+ - SQ是将每一个维度量化成指定位数的一个数
+ - 主流方案
+ - professional
+ - weaviate
+ - 1. 文档丰富,容易上手
+ - 2. 提供混合索引
+ - 3. 支持自托管+云原生
+ - 4.支持python,js,ts,go,java等客户端
+ - 5. 支持HNSW,HNSW-PQ,DisANN等索引
+ - chroma
+ - LanceDB
+ - pinecone
+ - 1. 完全云原生,非常容易上手
+ - 2. 自建复合索引
+ - faiss
+ - 1.来自 Meta AI(原 Facebook Research)的开源项目
+ - 2.同时支持cpu和GPU两种设备
+ - 3. 支持C++,python, go等客户端
+ - 4. 支持常见的索引方式,如IVF,HNSW,支持PQ量化
+ - 5. in-memory运行
+ - 6. self-hosted
+ - milvus
+ - 1. 通过代理、负载均衡器、消息代理、Kafka和Kubernetes的组合实现了高度可扩展性,这使得整个系统变得非常复杂和资源密集
+ - 2. 截至2023年,它是唯一一个提供可工作的DiskANN实现的主要供应商
+ - 3. 支持在向量相似度检索过程中进行标量字段过滤,实现混合查询
+ - 4. 采用 **存储与计算分离** 的架构设计
+ - 5. 提供python,juava,go,node.js等语言SDK,也提供milvus lite等in-momery运行
+ - 6. 提供了图形界面客户端
+ - traiditional
+ - ES
+ - redis
+ - pgvector
+ - Embedding模型
+ - bi-encoder
+ - cross-encoder
+ - 【可选】文本检索引擎
+ - ElasticSearch
+ - OpenSearch
+ - 【可选】图数据库
+ - 检索
+ - 向量检索
+ - 关键字检索
+ - BM25
+ - NL2Cypher
+ - NL2SQL
+ - RAG增强
+ - Self-RAG
+ - 框架
+ - 自反思检索增强生成(Self-RAG, Self-Reflective Retrieval-Augmented Generation)。这是一个新框架,它不仅可以根据需要自适应地检索段落(即:模型可以判断是否有必要进行检索增强),还引入了名为反思令牌(reflection tokens)的特殊令牌,使LM在推理阶段可控。
+ - ![图片](./img/大模型应用技术原理-幕布图片-108319-429731.jpg)
+ - ![图片](./img/大模型应用技术原理-幕布图片-918388-323086.jpg)
+ - 训练
+ - 首先,训练评论家,使用检索器检索到的段落以及反思令牌增强指令-输出数据,然后,使用标准的下一个 token 预测目标来训练生成器 LM,以学习生成 自然延续(continuations)以及特殊 tokens (用来检索或批评其自己的生成内容).
+ - 推理
+ - 它可以适应性地使用检索令牌进行检索,因此模型可以自发判断是不是有必要进行检索。它引入了多种细粒度的批评令牌,这些令牌用于评估生成内容的各个方面的质量。在生成过程中,作者使用期望的批评令牌概率的线性插值进行segment级的beam search,以在每一个时间步骤中确定最佳的K个续写方案
+- Agent
+ - function call
+ - ToolFormer
\ No newline at end of file
diff --git "a/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\345\256\236\346\210\230\344\270\216\345\272\224\347\224\250.md" "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\345\256\236\346\210\230\344\270\216\345\272\224\347\224\250.md"
new file mode 100644
index 0000000000000000000000000000000000000000..cf3ac14a5307b9f824b812a97f812137d3ee3956
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\345\256\236\346\210\230\344\270\216\345\272\224\347\224\250.md"
@@ -0,0 +1,44 @@
+# 大模型技术栈-实战与应用
+- 训练框架
+ - deepspeed
+ - megatron-lm
+ - colossal-ai
+ - trlx
+- 推理框架
+ - triton
+ - vllm
+ - text-generation-inference
+ - lit-llama
+ - lightllm
+ - TensorRT-LLM(原FasterTransformer)
+ - fastllm
+ - inferllm
+ - llama-cpp
+ - openPPL-LLM
+- 压缩框架
+ - bitsandbytes
+ - auto-gptq
+ - deepspeed
+- embedding框架
+ - sentence-transformer
+ - FlagEmbedding
+- 向量数据库 [向量数据库对比]("https://www.jianshu.com/p/43cc19426113")
+ - faiss
+ - pgvector
+ - milvus
+ - pinecone
+ - weaviate
+ - LanceDB
+ - Chroma
+- 应用框架
+ - Auto-GPT
+ - langchain
+ - llama-index
+ - quivr
+- python前端
+ - streamlit
+ - gradio
+- python API工具
+ - FastAPI+uvicorn
+ - flask
+ - Django
\ No newline at end of file
diff --git "a/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206.md" "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206.md"
new file mode 100644
index 0000000000000000000000000000000000000000..f28e21a4f419cc506b37d0739822a8585b6d4ee7
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\212\200\346\234\257\346\240\210-\347\256\227\346\263\225\344\270\216\345\216\237\347\220\206.md"
@@ -0,0 +1,301 @@
+# 大模型技术栈-算法与原理
+- 1. tokenizer方法
+ - word-level
+ - char-level
+ - subword-level
+ - BPE
+ - WordPiece
+ - UniLM
+ - SentencePiece
+ - ByteBPE
+- 2. position encoding
+ - 绝对位置编码
+ - ROPE
+ - AliBi
+ - 相对位置编码
+ - Transformer-XL
+ - T5/TUPE
+ - DeBERTa
+ - 其他位置编码
+- 3. 注意力机制
+ - 稀疏注意力
+ - flash-attention
+ -
+- 4. 分布式训练
+ - 数据并行
+ - FSDP
+ - DDP
+ - ZeRO
+ - Model state
+ - Optimizer->ZeRO1
+ - 将optimizer state分成若干份,每块GPU上各自维护一份
+ - 每块GPU上存一份完整的参数W,做完一轮foward和backward后,各得一份梯度,对梯度做一次 **AllReduce(reduce-scatter + all-gather)** , **得到完整的梯度G,由于每块GPU上只保管部分optimizer states,因此只能将相应的W进行更新,对W做一次All-Gather**
+ - Gradient+Optimzer->ZeRO2
+ - 每个GPU维护一块梯度
+ - 每块GPU上存一份完整的参数W,做完一轮foward和backward后, **算得一份完整的梯度,对梯度做一次Reduce-Scatter,保证每个GPU上所维持的那块梯度是聚合梯度,每块GPU用自己对应的O和G去更新相应的W。更新完毕后,每块GPU维持了一块更新完毕的W。同理,对W做一次All-Gather,将别的GPU算好的W同步到自己这来**
+ - Parameter+Gradient+Optimizer->ZeRO3
+ - 每个GPU维护一块模型状态
+ - 每块GPU上只保存部分参数W,做forward时,对W做一次 **All-Gather** ,取回分布在别的GPU上的W,得到一份完整的W, **forward做完,立刻把不是自己维护的W抛弃,做backward时,对W做一次All-Gather,取回完整的W,backward做完,立刻把不是自己维护的W抛弃. 做完backward,算得一份完整的梯度G,对G做一次Reduce-Scatter,从别的GPU上聚合自己维护的那部分梯度,聚合操作结束后,立刻把不是自己维护的G抛弃。用自己维护的O和G,更新W。由于只维护部分W,因此无需再对W做任何AllReduce操作**
+ - Residual state
+ - activation->Partitioned Activation Checkpointing
+ - 每块GPU上只维护部分的activation,需要时再从别的地方聚合过来就行。需要注意的是,activation对显存的占用一般会远高于模型本身,通讯量也是巨大的
+ - temporary buffer->Constant Size Buffer
+ - 提升带宽利用率。当GPU数量上升,GPU间的通讯次数也上升,每次的通讯量可能下降(但总通讯量不会变)。数据切片小了,就不能很好利用带宽了。所以这个buffer起到了积攒数据的作用:等数据积攒到一定大小,再进行通讯。
+ - 使得存储大小可控。在每次通讯前,积攒的存储大小是常量,是已知可控的。更方便使用者对训练中的存储消耗和通讯时间进行预估
+ - unusable fragment->Memory Defragmentation
+ - 对碎片化的存储空间进行重新整合,整出连续的存储空间。防止出现总存储足够,但连续存储不够而引起的存储请求fail
+ - offload
+ - ZeRO-Offload
+ - **forward和backward计算量高** ,因此和它们相关的部分,例如参数W(fp16),activation,就全放入GPU
+ - **update的部分计算量低** ,因此和它相关的部分,全部放入CPU中。例如W(fp32),optimizer states(fp32)和gradients(fp16)等
+ - ZeRO-Offload 分为 Offload Strategy 和 Offload Schedule 两部分,前者解决如何在 GPU 和 CPU 间划分模型的问题,后者解决如何调度计算和通信的问题
+ - ZeRO-Infinity
+ - 一是将offload和 ZeRO 的结合从 ZeRO-2 延伸到了 ZeRO-3,解决了模型参数受限于单张 GPU 内存的问题
+ - 二是解决了 ZeRO-Offload 在训练 batch size 较小的时候效率较低的问题
+ - 三是除 CPU 内存外,进一步尝试利用 NVMe 的空间
+ - 模型并行
+ - tensor-wise parallelism
+ - Megatron-LM
+ - 流水线并行
+ - GPipe
+ - PipeDream
+ - layer-wise parallelism
+ - sequence parallelism
+- 5. PEFT
+ - Lora类
+ - LoRA
+ - 用两个低秩矩阵替代待更新的权重矩阵的增量
+ - QLoRA
+ - 4 bit NormalFloat(NF4) 量化和双量化
+ - 引入了分页优化器,以防止梯度检查点期间的内存峰值
+ - AdaLoRA
+ - 用奇异值分解P \ Gamma Q代替AB,根据loss梯度评估对角线上值进行重要性评分,根据评分动态分配参数预算给权重矩阵
+ - AdaLoRA将关键的增量矩阵分配高秩以捕捉更精细和任务特定的信息,而将较不重要的矩阵的秩降低,以防止过拟合并节省计算预算。
+ - 以奇异值分解的形式对增量更新进行参数化,并根据重要性指标裁剪掉不重要的奇异值,同时保留奇异向量。
+ - 在训练损失中添加了额外的惩罚项,以规范奇异矩阵P和Q的正交性,从而避免SVD的大量计算并稳定训练
+ - IA3
+ - 通过学习向量来对激活层加权进行缩放
+ - 学习到的向量被注入到attention和feedforward模块中
+ - ReLoRA
+ - **ReLoRA在合并和重新启动期间可以对优化器进行部分重置** ,并在随后的预热中过程中将学习率设置为0。 **具体来说,作者提出了一种锯齿状学习率调度算法**
+ - 出发点:通过不断叠加LoRA训练过程来达到更好的训练效果, **首先需要对LoRA过程进行重新启动,想要对已经完成的LoRA过程重新启动并不容易,这需要对优化器进行精细的调整,如果调整不到位,会导致模型在重启后立即与之前的优化方向出现分歧**
+ - Prompt类
+ - prompt tuning
+ - 在输入层加一个embedding层
+ - P-tuning
+ - 在输入层加一个embedding和一个LSTM或MLP
+ - prefix tuning
+ - 在每一层加入一个embedding和一个MLP
+ - P-tuning v2
+ - 在每一层都加一个embedding层
+ - Adapter类
+ - Adapter Tuning
+ - 针对每一个Transformer层,增加了两个Adapter结构(分别是多头注意力的投影之后和第二个feed-forward层之后)
+ - Adapter Fusion
+ - 在 Adapter 的基础上进行优化,通过将学习过程分为两阶段来提升下游任务表现
+ - 知识提取阶段:在不同任务下引入各自的Adapter模块,用于学习特定任务的信息。
+ - 知识组合阶段:将预训练模型参数与特定于任务的Adapter参数固定,引入新参数(AdapterFusion)来学习组合多个Adapter中的知识,以提高模型在目标任务中的表现
+ - Adapter Drop
+ - 在不影响任务性能的情况下,对Adapter动态高效的移除,尽可能的减少模型的参数量,提高模型在反向传播(训练)和正向传播(推理)时的效率
+ - 其他
+ - BitFit
+ - 疏的微调方法,它训练时只更新bias的参数或者部分bias参数
+ - 混合式
+ - MAM Adapter
+ - 用 FFN 层的并行Adapter和软提示的组合
+ - UniPELT
+ - 门控被实现为线性层,通过GP参数控制Prefix-tuning方法的开关,GL控制LoRA方法的开关,GA控制Adapter方法的开关
+- 6. 压缩
+ - 剪枝
+ - OBD(Optimal Brain Damage)
+ - 利用二阶导数信息度量模型参数的显著性,剪掉影响小的参数降低模型复杂度提高泛化能力
+ - ![图片](./img/大模型技术栈-算法与原理-幕布图片-628857-182232.jpg)
+ - OBS(Optimal Brain Surgeon )
+ - OBD粗暴的只考虑海森矩阵对角线元素。OBS考虑海森矩阵全局信息,由此也获得参数相互之间的影响。
+ - OBC(OPTIMAL BRAIN COMPRESSION )
+ - OBS对整个神经网络进行剪枝,OBC对神经网络模型分层剪枝或者量化
+ - ExactOBS
+ - 参数更新和代价评估不需要使用整个海森矩阵,仅使用和剪枝参数所在行相关的 d_col\time d_col大小的海森矩阵。
+ - 量化 ![图片](./img/大模型技术栈-算法与原理-幕布图片-454007-940199.jpg)
+ - GPTQ
+ - 1.是对OBC的改进
+ - 2. 取消了贪心算法,采用固定位置优化
+ - 3. 分组量化,并行加速
+ - ![图片](./img/大模型技术栈-算法与原理-幕布图片-729151-372321.jpg)
+ - SpQR
+ - 核心思想:参数的对模型的重要程度,存在极强的不均衡性。1%的参数,可能主导的量化过程中损失的性能,假如我们在量化中保护这1%的参数,就能极大程度保护模型的性能不受影响
+ - 2. 对于每一层,它使用一个小的输入数据集X,用来计算单个参数w_ij被量化前后造成的的误差s_ij. 有了s_ij之后再取top 1%的参数认为它们是重要参数进行保护。
+ - 在挑选出参数之后,SqQR使用一个稀疏矩阵来单独保存这些参数,令这些重要参数的精度仍为fp16。
+ - SqQR在实验中还观察到重要参数往往以行或者列聚集,因此提出使用更小的group_size比如8或16,而非GPTQ中常用的128
+ - AWQ
+ - 1. AWS是在smoothquant的基础上提出来的
+ - 2. AWQ针对channel维度来寻找重要参数,依据是输入X以及这个参数本身W的绝对大小
+ - 3.方式是寻找一个缩放比例s,在参数量化之前W乘以这个比例,计算时输入X除以这个比例,以减小误差
+ - 4. 把s分成两个值S_x和S_w相乘,我们需要W越大s越小,X越大,s越大
+ - ![图片](./img/大模型技术栈-算法与原理-幕布图片-299768-254064.jpg)
+ - OBC(OPTIMAL BRAIN COMPRESSION )
+ - ![图片](./img/大模型技术栈-算法与原理-幕布图片-19929-302935.jpg)
+ - SmoothQuant
+ - 1. 当模型规模更大时,单个token的值变化范围较大,难以量化,相比之下 weight 的变化范围较小,即 weight 较易量化,而 activation 较难量化
+ - 2. SmoothQuant 核心思想是引入一个超参,减小激活值的变化范围,增大权重的变化范围,从而均衡两者的量化难度
+ - 3. 得到smooth变换之后的 activation 和 weight 矩阵,可以再采用 per-token 或 per-tensor 的量化方式,
+ - ![图片](./img/大模型技术栈-算法与原理-幕布图片-81470-404273.jpg)
+ - LLM.int8
+ - 采用混合精度分解的量化方法:将包含了Emergent Features的几个维度从矩阵中分离出来,对其做高精度的矩阵乘法;其余部分进行量化
+ - ZeroQuant
+ - 1. 对权重使用分组量化,对激活使用token量化
+ - 2. 开发了高度优化的推理后端,消除了量化/反量化运算符昂贵的计算成本,在现代GPU硬件上实现INT8 Tensor内核的延迟加速
+ - 3. 提出了一种用于INT4/INT8混合精度量化的新的逐层知识蒸馏方法(LKD),其中神经网络通过蒸馏逐层量化,迭代最小,甚至不访问原始训练数据
+ - 分类学
+ - 对称量化vs非对称量化
+ - 量化是否均衡,原点是否为0
+ - 动态量化vs静态量化
+ - 输入的缩放因子计算方法不同
+ - 静态量化的模型在使用前有“calibrate”的过程(校准缩放因子),量化模型的缩放因子会根据输入数据的分布进行调整
+ - Weights量化vsActivation量化
+ - feature map(fm)就是每一层网络的输入tensor,featuremap量化就是我们常说的激活量化
+ - per-token vs. per-layer/per-tensor vs. per channel vs. per group vs
+ - **per-token quantization** :激活每个token对应的tensor共享量化系数
+ - **per-tensor quantization** : 对一整个tensor设置简单的量化集合
+ - **per-channel quantization** : 对权重的每个输出通道设置一个量化集合,但实际中feature 还是整个 tensor 共用一个 scale 和 zeropoint,但每个 kernel 会单独统计一个 scale 和 zeropoint(注意是每个 kernel,而不是 kernel 的每个 channel)
+ - **group-wise quantization** : 把多个channel合在一起用一组量化系数
+ - 蒸馏(layer reduction)
+- 7. 推理
+ - 7.1 吞吐量与显存优化
+ - PagedAttention
+ - Qunatized KV Cache
+ - MQA/GQA
+ - FlashAttention
+ - 7.2 算子融合
+ - 7.3 延迟优化
+ - No Padding优化
+ - 7.4 调度优化
+ - Dynamic Batching
+ - Async Servering
+ - Inflight Batching
+ - 7.5 量化
+ - 7.6 模型并行
+ - tensor paralellism
+ - 7.7 请求优化
+ - rpc
+ - grpc
+ - http
+- 8. 应用
+ - RAG
+ - Agent
+- 9. embedding模型
+ - 分类学
+ - 对称vs. 非对称 vs. 混合
+ - 对称 query:qestion, text:text
+ - sentence-T5
+ - 非对称:query:text
+ - GTR
+ - 混合
+ - Instructor
+ - 对比学习+对比学习 vs. 自编码+对比学习
+ - 对比学习+对比学习
+ - sentence-T5
+ - GTR
+ - E5
+ - 自编码+对比学习
+ - bge
+ - retromae
+ - bert-based vs. GPT-based
+ - bert-based
+ - LLM-based
+ - PromptEOL+CSE+LLM
+ - Bert-CLS,Bert-mean
+ - 双向decoder-encoder的Transformer
+ - T5 series
+ - Sentence-T5
+ - T5-encoder+mean pooling
+ - 无标注对比学习+有标注对比学习的两阶段训练
+ - Jina
+ - 以T5为基本架构
+ - 去重、语言过滤、一致性过来
+ - **采用了并行化方法在多个数据集上进行训练** ,但设计了一个约束条件:每个训练批次(batch)仅包含来自单一数据集的样本
+ - 三元组训练:enchor,entainment, contraversive
+ - GTR
+ - 与sentence-T5结构相同
+ - 将finetune的数据集从NLI换成检索相关的,并且利用百度的rocketqa来获得hard negative
+ - 对比学习改成双向对比学习(每个batch里有两个对比学习损失,第一个损失是以query为中心去构建正负样本,第二个损失是以positive document为中心去构建正负样本)
+ - simcse
+ - 无监督Simcse
+ - 对于同一条语句,在训练中使用两次不同的dropout mask,把两次dropout后的句子对视为一组互为正例的样本对,即相似句子对
+ - "不相似句子对"通过采样同一批(batch)内的其余句子即可
+ - 有监督simcse
+ - 采用NLI有监督数据集做对比学习训练,NLI,即自然语言推理,其任务是判断两句话之间的关系,可能的关系有entailment (相近), contradiction (矛盾)或neutral (中立)。
+ - entailment sentence pair作为正例, contradiction sentence pair作为hard negative样本
+ - 衍生算法
+ - Esimcse
+ - ESimCSE选择在句子中随机重复一些单词作为正样本,解决模型倾向于判断相同或相似长度的句子在表达上更相近的问题
+ - 维护了一个队列,重用前面紧接的mini-batch的编码嵌入来扩展负对,并使用了动量编码器
+ - CoSENT
+ - 在正负样本的基础上,基于circle loss进一步引入排序
+ - SNCSE
+ - 针对模型「无法区分文本相似度和语义相似度,更偏向具有相似文本,而不考虑实际语义差异」的问题,提出了一种「显式添加否定词从而生成软负样本」结合「双向边际损失」的方案。
+ - EASE
+ - 强调实体在句向量表征中的重要性。在数据层面,使用正、负实体代替正负样本。
+ - CLAIF
+ - 针对训练过程中缺乏细粒度的监督信号, 即没有考虑到正样本对之间的相似性差异,引入来自LLM的AI反馈,构造具有不同相似度的样本对,并对这些样本对给出细粒度的相似度分数作为监督信号,帮助文本表示的学习。
+ - Instructor
+ - 1. 以GTR为基底模型,经过进一步的“instruction tuning”得到
+ - 2. 将模型输入改成Task Instuction+[X]([X]代表具体的文本输入)
+ - E5
+ - E5提出了一个预训练数据过滤的方案consistency-based filter
+ - 以Bert为基座的embedding模型
+ - 在模型输入侧加入了Prefix("query:"跟“paragraph:”),从而让模型知道文本的类型,跟Instructor的instruction类似
+ - BGE
+ - 基于RetroMAE方案
+ - BGE在finetune阶段针对检索任务需要加入特定的Prefix(只在query侧加"Represent this sentence for searching relevant passages:")
+ - RetroMAE
+ - 包括一个以Bert为基底的Encoder跟一个只有一层的Decoder
+ - Encoder端以30%的比例对原文本进行mask,最终得到最后一层[CLS]位置的向量表征作为句向量
+ - Decoder端则以50%的比例对原文本进行mask,联合Encoder端的句向量,对原本进行重建
+ - PromptBert
+ - 以Bert为基底,通过选择合适的prompt(“This sentence:"[X]" means [MASK] ”),然后以最后一层[MASK]位置的向量表征作为句向量,即便不经过额外的finetune也能取得令人惊艳的效果
+ - PromptEOL+CLS+LLM
+ - 语言模型使用的是OPT跟LLaMA
+ - 构建了另一个新的prompt,"This sentence:”[X]” means in one word:",以下一个生成token的隐层状态作为text embedding
+ - 还引入了in-context learning,为每个语言模型找到了一个最佳的demonstration,从而指导语言模型生成更符合要求的text embedding
+ - 为了进一步提升性能,可以采用对比学习的方式作进一步的finetune
+- 10. 上下文扩展
+ - Alibi
+ - log(n)注意力缩放
+ - window attention
+ - RoPE改进
+ - Interpolation
+ - Position Interpolation
+ - 线性插值
+ - Extrapolation
+ - NTK感知缩放RoPE
+ - dynamic缩放RoPE
+ - consistent of Dynamically Scaled RoPE
+ - 混合
+ - Rectified RoPE
+ - **N** aive **B** ayes-based **C** ontext **E** xtension
+ - 只需要修改一下解码函数中的logits构建方式
+ - 即插即用、模型无关、无须微调、线性效率、实现简单
+ - NBCE的一大缺点是无序性,即无法识别Context的输入顺序,这在续写故事等场景可能表现欠佳
+- 11. Prompt Engineering
+ - **Chain of Thought**
+ - Let’s Think step by step
+ - **Self-Consistency**
+ - Few-shot + {question} 用几个相似的具有推导步骤的例子
+ - **Auto-CoT**
+ - Few-shot + {question} +Chain of Thought相似的具有推导步骤的例子+{问题}+给出具体思考过程。
+ - **Generation Knowledge**
+ - 以事实+知识的方式组织样例,再最后提问,要求给出解释和答案
+ - **Automatic Prompt Engineer**
+ - Let's work this out in a step by step way to be sure we have the right answer
+ - **OPRO**
+ - “Take a deep breath and think step by step.”
+ - Optimization by PROmpting (OPRO)总体架构:最开始输入meta-prompt,这个初始的meta-prompt基本上只是对优化任务进行了描述(也会有few-shot example)。输入后LLM便会生成一个solution,这个solution由objective function评估并打分。(solution, score)组合成一对添加到meta-prompt中,如此完成一个循环。多次循环后取分数最高的solution作为优化结果。
+ - meta-prompt分为两部分,问题描述和优化轨迹,问题描述就是用自然语言描述想要优化的问题,比如“generate a new instruction that achieves a higher accuracy”。而优化弹道(Optimization trajectory)则是指之前提到的(solution, score)对,即之前生成的解决方案和对应的分数,可以看作优化的“日志”。但是要注意这个弹道不是按 排的,而是按照打分升序排的。因为之前的研究也发现,越靠后的样例对输出的影响越大,所以把分数高的排在后面有利于LLM向其学习。 [时间顺序]("https://so.csdn.net/so/search?q=%E6%97%B6%E9%97%B4%E9%A1%BA%E5%BA%8F&spm=1001.2101.3001.7020")
+ - **Tree of Thought**
+ - f“给定当前的推理状态:‘{state_text}’,生成{k}条连贯的思想来实现推理过程:”
+ - f“鉴于当前的推理状态:‘{state_text}’,根据其实现 {initial_prompt} 的潜力悲观地将其值评估为 0 到 1 之间的浮点数”
+ - 利用树的遍历算法(BFS, DFS, MC,BF,A*),搜索最佳答案。
+ - **Graph of Thought**
+ - 创新点是将大模型生成的信息建模为一个图,节点是 “LLM的思想“,边是这些思想的依赖关系。这种方法能够将任意 LLM 思想,组合,提取出这个网络的思想本质。
+ - **出发点** :人类的思维在解决问题时,不是只会链式思维或者尝试不同的链(TOT),而是在脑中构建一个复杂的思维网络。人类在思考时会沿着一个链式的推理,回溯,再尝试一个新的方向,并把之前的链的优点保留,缺点剔除,与当前探索的链的方向结合生成一个新的解决方案
\ No newline at end of file
diff --git "a/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206.md" "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206.md"
new file mode 100644
index 0000000000000000000000000000000000000000..bf7ef12a3bf2830ff37474c22f69ac22f11f426b
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\214\207\344\273\244\345\257\271\351\275\220\350\256\255\347\273\203\345\216\237\347\220\206.md"
@@ -0,0 +1,30 @@
+# 大模型指令对齐训练原理
+- RLHF
+ - SFT
+ - RM
+ - PPO
+- AIHF-based
+ - RLAIF
+ - 核心在于通过AI 模型监督其他 AI 模型,即在SFT阶段,从初始模型中采样,然后生成自我批评和修正,然后根据修正后的反应微调原始模型。在 RL 阶段,从微调模型中采样,使用一个模型来评估生成的样本,并从这个 AI 偏好数据集训练一个偏好模型。然后使用偏好模型作为奖励信号对 RL 进行训练
+ - ![图片](./img/大模型指令对齐训练原理-幕布图片-17565-176537.jpg)
+ - ![图片](./img/大模型指令对齐训练原理-幕布图片-95996-523276.jpg)
+ - ![图片](./img/大模型指令对齐训练原理-幕布图片-349153-657791.jpg)
+ - RRHF
+ - RRHF( **R** ank **R** esponse from **H** uman **F** eedback) 不需要强化学习,可以利用不同语言模型生成的回复,包括 ChatGPT、GPT-4 或当前的训练模型。RRHF通过对回复进行评分,并通过排名损失来使回复与人类偏好对齐。RRHF 通过通过排名损失使评分与人类的偏好(或者代理的奖励模型)对齐。RRHF 训练好的模型可以同时作为生成语言模型和奖励模型使用。
+ - ![图片](./img/大模型指令对齐训练原理-幕布图片-805089-731888.jpg)
+- SFT-only
+ - LIMA
+ - LIMA(Less Is More for Alignment) 即浅层对齐假说,即一 **个模型的知识和能力几乎完全是在预训练中学习的,而对齐则是教会它与用户交互时如何选择子分布** 。如果假说正确,对齐主要有关于学习方式,那么该假说的一个推论是,人们可以用相当少的样本充分调整预训练的语言模型。因此, **该工作假设,对齐可以是一个简单的过程,模型学习与用户互动的风格或格式,以揭示在预训练中已经获得的知识和能力。**
+ - LTD Instruction Tuning
+ - ![图片](./img/大模型指令对齐训练原理-幕布图片-759487-923925.jpg)
+- Reward-only
+ - DPO
+ - DPO(Direct Preference Optimization) 提出了一种使用二进制交叉熵目标来精确优化LLM的方法,以替代基于 RL HF 的优化目标,从而大大简化偏好学习 pipeline。也就是说,完全可以直接优化语言模型以实现人类的偏好,而不需要明确的奖励模型或强化学习。
+ - DPO 也依赖于理论上的偏好模型(如 Bradley-Terry 模型),以此衡量给定的奖励函数与经验偏好数据的吻合程度。然而,现有的方法使用偏好模型定义偏好损失来训练奖励模型,然后训练优化所学奖励模型的策略,而 DPO 使用变量的变化来直接定义偏好损失作为策略的一个函数。鉴于人类对模型响应的偏好数据集,DPO 因此可以使用一个简单的二进制交叉熵目标来优化策略,而不需要明确地学习奖励函数或在训练期间从策略中采样。
+ - RAFT
+ - ![图片](./img/大模型指令对齐训练原理-幕布图片-350029-666381.jpg)
+- 参考文献
+ - [反思RLHF]("https://mp.weixin.qq.com/s/e3E_XsZTiNMNYqzzi6Pbjw")
+ - [RLHF笔记]("https://mathpretty.com/16017.html")
+ - [hf-blog]("https://huggingface.co/blog/zh/rlhf")
+ - ** [RLHF代码详解]("https://zhuanlan.zhihu.com/p/624589622")
\ No newline at end of file
diff --git "a/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245.md" "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245.md"
new file mode 100644
index 0000000000000000000000000000000000000000..6d925a9abcdbc3d8c226b0af2c259475739ce68c
--- /dev/null
+++ "b/knowledge_base/samples/content/llm/\345\244\247\346\250\241\345\236\213\346\216\250\347\220\206\344\274\230\345\214\226\347\255\226\347\225\245.md"
@@ -0,0 +1,63 @@
+# 大模型推理优化策略
+- 7.1 显存优化
+ - [PagedAttention]("https://zhuanlan.zhihu.com/p/638468472")
+ - KV cache,其具有以下特点:1. 显存占用大,14b级别的模型,每个token需要约0.7M-1M的显存;2. 动态变化:KV 缓存的大小取决于序列长度,这是高度可变和不可预测的。因此,这对有效管理 KV cache 挑战较大。该研究发现,由于碎片化和过度保留,现有系统浪费了 60% - 80% 的显存。
+ - 为了解决这个问题,该研究引入了 PagedAttention,这是一种受操作系统中虚拟内存和分页经典思想启发的注意力算法。与传统的注意力算法不同,PagedAttention 允许在非连续的内存空间中存储连续的 key 和 value 。具体来说,PagedAttention 将每个序列的 KV cache 划分为块,每个块包含固定数量 token 的键和值。在注意力计算期间,PagedAttention 内核可以有效地识别和获取这些块。因为块在内存中不需要连续,因而可以用一种更加灵活的方式管理 key 和 value ,就像在操作系统的虚拟内存中一样:可以将块视为页面,将 token 视为字节,将序列视为进程。序列的连续逻辑块通过块表映射到非连续物理块中。物理块在生成新 token 时按需分配。在 PagedAttention 中,内存浪费只会发生在序列的最后一个块中。这使得在实践中可以实现接近最佳的内存使用,仅浪费不到 4%。
+ - PagedAttention 还有另一个关键优势 —— 高效的内存共享。例如在并行采样中,多个输出序列是由同一个 prompt 生成的。在这种情况下,prompt 的计算和内存可以在输出序列中共享。PagedAttention 自然地通过其块表格来启动内存共享。与进程共享物理页面的方式类似,PagedAttention 中的不同序列可以通过将它们的逻辑块映射到同一个物理块的方式来共享块。为了确保安全共享,PagedAttention 会对物理块的引用计数进行跟踪,并实现写时复制(Copy-on-Write)机制。PageAttention 的内存共享大大减少了复杂采样算法的内存开销,例如并行采样和集束搜索的内存使用量降低了 55%。这可以转化为高达 2.2 倍的吞吐量提升。
+ - continuous batching
+ - CUDA kernel优化
+ - Qunatized KV Cache
+ - MQA/GQA
+ - 核心思想是检索kv-cache的数量,以少量kv-cache对应多个query ![图片](./img/大模型推理优化策略-幕布图片-699343-219844.jpg)
+ - ![图片](./img/大模型推理优化策略-幕布图片-930255-616209.jpg)
+ - [FlashAttention]("https://zhuanlan.zhihu.com/p/638468472")
+ - 解释1:记I为模型的计算强度I,单位FLOP/byte代表模型进行单位byte数据交互可实现的操作数,则I*带宽beta即模型的计算性能,单位为FLOP/s。令I_max=计算平台算力/计算平台带宽,当模型的计算强度I小于平台的理论计算强度I_max,模型的计算能力P即I*beta,当模型的计算强度大于I_max,则模型的计算性能P等于平台的算力。故若模型的计算强度小,则瓶颈在带宽,若模型的计算强度大,则瓶颈在算力。为提高计算性能,需提高计算强度,即每bytes数据交互的操作数。
+ - 解释2:记N=每次操作要求的FLOP,单位FLOP/OP;pi=平台的算力,单位FLOP/s;beta=内存带宽,单位byte/s;P=实际实现计算速度,单位为FLOP/s;优化目标为O=P/N每秒钟实现的运算次数,单位为OP/s.由于N固定,故优化目标转而为P,P=min{beta_r*I_max=beta_r*pi/beta,pi},故优化目标转而为beta,即改变内存访问策略,实现beta最大化。
+ - 注意力操作中,S和P的计算空间复杂度都是O(N^2),此外,scale,mask,softmax,dropout都是带宽约束操作。 ![图片](./img/大模型推理优化策略-幕布图片-380552-579242.jpg)
+ - ![图片](./img/大模型推理优化策略-幕布图片-789705-122117.jpg)
+ - 可以看出,O(N^2)空间复杂度的矩阵计算对HBM的读写是主要的内存瓶颈,因此主要优化点是:1. 在不访问整个输入的情况下计算softmax;2. 不为反向传播存储大的中间attention矩阵。FlashAttention提出两种方法来分步解决上述问题:tiling,recomputation.tiling - 注意力计算被重新构造,将输入分割成块,并通过在输入块上进行多次传递来递增地执行softmax操作。recomputation - 存储来自前向的 softmax 归一化因子,以便在反向中快速重新计算芯片上的 attention,这比从HBM读取中间矩阵的标准注意力方法更快。由于重新计算,这确实导致FLOPs增加,但是由于大量减少HBM访问,FlashAttention运行速度更快。该算法背后的主要思想是分割输入,将它们从慢速HBM加载到快速SRAM,然后计算这些块的 attention 输出。在将每个块的输出相加之前,将其按正确的归一化因子进行缩放,从而得到正确的结果。
+ - ![图片](./img/大模型推理优化策略-幕布图片-590671-36787.jpg)
+ - ![图片](./img/大模型推理优化策略-幕布图片-276446-401476.jpg)
+ - 参考文献
+ - [推理优化]("https://zhuanlan.zhihu.com/p/656485997") [推理优化]("https://zhuanlan.zhihu.com/p/656485997")
+- 7.2 算子融合
+- 7.3 延迟优化
+ - No Padding优化
+- 7.4 调度优化
+ - Dynamic Batching
+ - 批次大小固定不变,无法随计算资源负载动态变化,导致 GPU 资源利用率低
+ - 通过维护一个作业队列实现,在 batch 维度动态插入新序列
+ - Async Servering
+ - Tokenize / Detokenize 过程在 CPU 上执行,期间 GPU 处于空闲状态
+ - 多线程异步,流水线 overlap 实现降低时延
+ - Inflight Batching/continuous batching
+ - 同批次序列推理时,存在“气泡”,导致 GPU 资源利用率低
+ - 由 batch 粒度的调度细化为 step 级别的调度,在时间轴方向动态插入新序列
+- 7.5 量化
+ - GPTQ
+ - AWQ
+ - SmoothQuant
+ - SpQR
+- 7.6 模型并行
+ - tensor paralellism加速+降显存
+ - pipeline paralellism减少显存
+- 7.7 请求优化
+ - [网络通信]("https://article.juejin.cn/post/7226374741064892453")
+ - rpc
+ - grpc
+ - http
+ - [响应模式]("https://blog.csdn.net/weixin_44999716/article/details/128436984")
+ - SSE
+ - 轮询
+ - 长轮询comet
+ - WebSocket
+- 7.8 采样/解码
+ - speculative decoding
+ - ![图片](./img/大模型推理优化策略-幕布图片-923924-83386.jpg)
+ - Blockwise *Parallel* *Decoding*
+ - Medusa
+ - SOT - Parallel Decoding
+ - SpecInfer
+ - StreamingLLM
+- 参考资料
+ - [最佳实践]("https://mp.weixin.qq.com/s/nJLrE9Dzj1mmTeQDiqNdzA")
\ No newline at end of file
diff --git a/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.csv b/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b1ca2ac9d64563278ec19323efafe70382820d5f
--- /dev/null
+++ b/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.csv
@@ -0,0 +1,173 @@
+,title,file,url,detail,id
+0,加油~以及一些建议,2023-03-31.0002,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/2,加油,我认为你的方向是对的。,0
+1,当前的运行环境是什么,windows还是Linux,2023-04-01.0003,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/3,当前的运行环境是什么,windows还是Linux,python是什么版本?,1
+2,请问这是在CLM基础上运行吗?,2023-04-01.0004,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/4,请问是不是需要本地安装好clm并正常运行的情况下,再按文中的步骤执行才能运行起来?,2
+3,[复现问题] 构造 prompt 时从知识库中提取的文字乱码,2023-04-01.0005,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/5,hi,我在尝试复现 README 中的效果,也使用了 ChatGLM-6B 的 README 作为输入文本,但发现从知识库中提取的文字是乱码,导致构造的 prompt 不可用。想了解如何解决这个问题。,3
+4,后面能否加入上下文对话功能?,2023-04-02.0006,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/6,目前的get_wiki_agent_answer函数中已经实现了历史消息传递的功能,后面我再确认一下是否有langchain中model调用过程中是否传递了chat_history。,4
+5,请问:纯cpu可以吗?,2023-04-03.0007,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/7,很酷的实现,极大地开拓了我的眼界!很顺利的在gpu机器上运行了,5
+6,运行报错:AttributeError: 'NoneType' object has no attribute 'message_types_by_name',2023-04-03.0008,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/8,报错:,6
+7,运行环境:GPU需要多大的?,2023-04-03.0009,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/9,如果按照THUDM/ChatGLM-6B的说法,使用的GPU大小应该在13GB左右,但运行脚本后,占用了24GB还不够。,7
+8,请问本地知识的格式是什么?,2023-04-03.0010,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/10,已测试格式包括docx、md文件中的文本信息,具体格式可以参考 [langchain文档](https://python.langchain.com/en/latest/modules/indexes/document_loaders/examples/unstructured_file.html?highlight=pdf#),8
+9,24G的显存还是爆掉了,是否支持双卡运行,2023-04-03.0011,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/11,RuntimeError: CUDA out of memory. Tried to allocate 96.00 MiB (GPU 0; 23.70 GiB total capacity; 22.18 GiB already allocated; 12.75 MiB free; 22.18 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF,9
+10,你怎么知道embeddings方式和模型训练时候的方式是一样的?,2023-04-03.0012,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/12,embedding和LLM的方式不用一致,embedding能够解决语义检索的需求就行。这个项目里用到embedding是在对本地知识建立索引和对问句转换成向量的过程。,10
+11,是否能提供本地知识文件的格式?,2023-04-04.0013,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/13,是否能提供本地知识文件的格式?,11
+12,是否可以像清华原版跑在8G一以下的卡?,2023-04-04.0016,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/16,是否可以像清华原版跑在8G一以下的卡?我的8G卡爆显存了🤣🤣🤣,12
+13,请教一下langchain协调使用向量库和chatGLM工作的,2023-04-05.0018,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/18,代码里面这段是创建问答模型的,会接入ChatGLM和本地语料的向量库,langchain回答的时候是怎么个优先顺序?先搜向量库,没有再找chatglm么? 还是什么机制?,13
+14,在mac m2max上抛出了ValueError: 150001 is not in list这个异常,2023-04-05.0019,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/19,我把chatglm_llm.py加载模型的代码改成如下,14
+15,程序运行后一直卡住,2023-04-05.0020,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/20,感谢作者的付出,不过本人在运行时出现了问题,请大家帮助。,15
+16,问一下chat_history的逻辑,2023-04-06.0022,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/22,感谢开源。,16
+17,为什么每次运行都会loading checkpoint,2023-04-06.0023,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/23,我把这个embeding模型下载到本地后,无法正常启动。,17
+18,本地知识文件能否上传一些示例?,2023-04-06.0025,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/25,如题,怎么构造知识文件,效果更好?能否提供一个样例,18
+19,What version of you are using?,2023-04-06.0026,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/26,"Hi Panda, I saw the `pip install -r requirements` command in README, and want to confirm you are using python2 or python3? because my pip and pip3 version are all is 22.3.",19
+20,有兴趣交流本项目应用的朋友可以加一下微信群,2023-04-07.0027,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/27,![IMG_1630](https://user-images.githubusercontent.com/5668498/230533162-8b9bfcdd-249c-4efe-b066-4f9ba2ce9f23.jpeg),20
+21,本地知识越多,回答时检索的时间是否会越长,2023-04-07.0029,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/29,是的 因为需要进行向量匹配检索,21
+22,爲啥最後還是報錯 哭。。,2023-04-07.0030,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/30,Failed to import transformers.models.t5.configuration_t5 because of the following error (look up to see,22
+23,对话到第二次的时候就报错UnicodeDecodeError: 'utf-8' codec can't decode,2023-04-07.0031,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/31,对话第一次是没问题的,模型返回输出后又给到请输入你的问题,我再输入问题就报错,23
+24,用的in4的量化版本,推理的时候显示需要申请10Gb的显存,2023-04-07.0033,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/33,"File ""/root/.cache/huggingface/modules/transformers_modules/chatglm-6b-int4-qe/modeling_chatglm.py"", line 581, in forward",24
+25,使用colab运行,python3.9,提示包导入有问题,2023-04-07.0034,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/34,"from ._util import is_directory, is_path",25
+26,运行失败,Loading checkpoint未达到100%被kill了,请问下是什么原因?,2023-04-07.0035,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/35,日志如下:,26
+27,弄了个交流群,自己弄好多细节不会,大家技术讨论 加connection-image 我来拉你,2023-04-08.0036,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/36,自己搞好多不清楚的,一起来弄吧。。准备搞个部署问题的解决文档出来,27
+28,Error using the new version with langchain,2023-04-09.0043,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/43,Error with the new changes:,28
+29,程序报错torch.cuda.OutOfMemoryError如何解决?,2023-04-10.0044,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/44,报错详细信息如下:,29
+30,qa的训练数据格式是如何设置的,2023-04-10.0045,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/45,本项目不是使用微调的方式,所以并不涉及到训练过程。,30
+31,The FileType.UNK file type is not supported in partition. 解决办法,2023-04-10.0046,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/46,ValueError: Invalid file /home/yawu/Documents/langchain-ChatGLM-master/data. The FileType.UNK file type is not supported in partition.,31
+32,如何读取多个txt文档?,2023-04-10.0047,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/47,如题,请教一下如何读取多个txt文档?示例代码中只给了读一个文档的案例,这个input我换成string之后也只能指定一个文档,无法用通配符指定多个文档,也无法传入多个文件路径的列表。,32
+33,nltk package unable to either download or load local nltk_data folder,2023-04-10.0049,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/49,I'm running this project on an offline Windows Server environment so I download the Punkt and averaged_perceptron_tagger tokenizer in this directory:,33
+34,requirements.txt中需要指定langchain版本,2023-04-11.0055,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/55,langchain版本0.116下无法引入RetrievalQA,需要指定更高版本(0.136版本下无问题),34
+35,Demo演示无法给出输出内容,2023-04-12.0059,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/59,你好,测试了项目自带新闻稿示例和自行上传的一个文本,可以加载进去,但是无法给出答案,请问属于什么情况,如何解决,谢谢。PS: 1、今天早上刚下载全部代码;2、硬件服务器满足要求;3、按操作说明正常操作。,35
+36,群人数过多无法进群,求帮忙拉进群,2023-04-12.0061,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/61,您好,您的群人数超过了200人,目前无法通过二维码加群,请问您方便加我微信拉我进群吗?万分感谢,36
+37,群人数已满,求大佬拉入群,2023-04-12.0062,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/62,已在README中更新拉群二维码,37
+38,requirements中langchain版本错误,2023-04-12.0065,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/65,langchain版本应该是0.0.12而不是0.0.120,38
+39,Linux : Searchd in,2023-04-13.0068,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/68,import nltk,39
+40,No sentence-transformers model found,2023-04-13.0069,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/69,加载不了这个模型,错误原因是找不到这个模型,但是路径是配置好了的,40
+41,Error loading punkt: ",58
+59,为啥放到方法调用会出错,这个怎么处理?,2023-04-20.0150,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/150,```python,59
+60,No sentence-transformers model found with name C:\Users\Administrator/.cache\torch\sentence_transformers\GanymedeNil_text2vec-large-chinese. Creating a new one with MEAN pooling.,2023-04-21.0154,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/154,卡在这块很久是正常现象吗,60
+61,微信群需要邀请才能加入,2023-04-21.0155,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/155,RT,给个个人联系方式白,61
+62,No sentence-transformers model found with name GanymedeNil/text2vec-large-chinese. Creating a new one with MEAN pooling,2023-04-21.0156,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/156,ls GanymedeNil/text2vec-large-chinese,62
+63,embedding会加载两次,2023-04-23.0159,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/159,你好,为什么要这样设置呢,这样会加载两次呀。,63
+64,扫二维码加的那个群,群成员满了进不去了,2023-04-23.0160,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/160,如题,64
+65,执行python3 cli_demo.py 报错AttributeError: 'NoneType' object has no attribute 'chat',2023-04-24.0163,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/163,"刚开始怀疑是内存不足问题,换成int4,int4-qe也不行,有人知道是什么原因吗",65
+66,匹配得分,2023-04-24.0167,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/167,在示例cli_demo.py中返回的匹配文本没有对应的score,可以加上这个feature吗,66
+67,大佬有计划往web_ui.py加入打字机功能吗,2023-04-25.0170,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/170,目前在载入了知识库后,单张V100 32G在回答垂直领域的问题时也需要20S以上,没有打字机逐字输出的使用体验还是比较煎熬的....,67
+68,Is it possible to use a verctorDB for the embedings?,2023-04-25.0171,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/171,"when I play, I have to load the local data again and again when to start. I wonder if it is possible to use",68
+69,请问通过lora训练官方模型得到的微调模型文件该如何加载?,2023-04-25.0173,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/173,通过lora训练的方式得到以下文件:,69
+70,from langchain.chains import RetrievalQA的代码在哪里?,2023-04-25.0174,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/174,local_doc_qa.py,70
+71,哪里有knowledge_based_chatglm.py文件?怎么找不到了??是被替换成cli_demo.py文件了吗?,2023-04-26.0175,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/175,哪里有knowledge_based_chatglm.py文件?怎么找不到了??是被替换成cli_demo.py文件了吗?,71
+72,AttributeError: 'Chatbot' object has no attribute 'value',2023-04-26.0177,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/177,Traceback (most recent call last):,72
+73,控制台调api.py报警告,2023-04-26.0178,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/178,"you must pass the application as an import string to enable ""reload"" or ""workers""",73
+74,如何加入群聊,2023-04-27.0183,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/183,微信群超过200人了,需要邀请,如何加入呢?,74
+75,如何将Chatglm和本地知识相结合,2023-04-27.0185,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/185,您好,我想请教一下怎么才能让知识库匹配到的文本和chatglm生成的相结合,而不是说如果没搜索到,就说根据已知信息无法回答该问题,谢谢,75
+76,一点建议,2023-04-27.0189,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/189,1.weiui的get_vector_store方法里面添加一个判断以兼容gradio版本导致的上传异常,76
+77,windows环境下,按照教程,配置好conda环境,git完项目,修改完模型路径相关内容后,运行demo报错缺少,2023-04-28.0194,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/194,报错代码如下:,77
+78,ValueError: too many values to unpack (expected 2),2023-04-28.0198,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/198,"When i tried to use the non-streaming, `ValueError: too many values to unpack (expected 2)` error came out.",78
+79,加载doc后覆盖原本知识,2023-04-28.0201,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/201,加载较大量级的私有知识库后,原本的知识会被覆盖,79
+80,自定义知识库回答效果很差,2023-04-28.0203,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/203,"请问加了自定义知识库知识库,回答效果很差,是因为数据量太小的原因么",80
+81,python310下,安装pycocotools失败,提示低版本cython,实际已安装高版本,2023-04-29.0208,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/208,RT,纯离线环境安装,依赖安装的十分艰难,最后碰到pycocotools,始终无法安装上,求教方法!,81
+82,[FEATURE] 支持 RWKV 模型(目前已有 pip package & rwkv.cpp 等等),2023-05-01.0216,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/216,您好,我是 RWKV 的作者,介绍见:https://zhuanlan.zhihu.com/p/626083366,82
+83,[BUG] 为啥主机/服务器不联网不能正常启动服务?,2023-05-02.0220,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/220,**问题描述 / Problem Description**,83
+84,[BUG] 简洁阐述问题 / Concise description of the issue,2023-05-03.0222,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/222,**local variable 'torch' referenced before assignment**,84
+85,不支持txt文件的中文输入,2023-05-04.0235,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/235,"vs_path, _ = local_doc_qa.init_knowledge_vector_store(filepath)",85
+86,文件均未成功加载,请检查依赖包或替换为其他文件再次上传。 文件未成功加载,请重新上传文件,2023-05-05.0237,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/237,请大佬帮忙解决,谢谢!,86
+87,[BUG] 使用多卡时chatglm模型加载两次,2023-05-05.0241,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/241,chatglm_llm.py文件下第129行先加载了一次chatglm模型,第143行又加载了一次,87
+88,[BUG] similarity_search_with_score_by_vector函数返回多个doc时的score结果错误,2023-05-06.0252,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/252,**问题描述 / Problem Description**,88
+89,可以再建一个交流群吗,这个群满了进不去。,2023-05-06.0255,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/255,上午应该已经在readme里更新过了,如果不能添加可能是网页缓存问题,可以试试看直接扫描img/qr_code_12.jpg,89
+90,请问这是什么错误哇?KeyError: 'serialized_input',2023-05-06.0257,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/257,运行“python webui.py” 后这是什么错误?怎么解决啊?,90
+91,修改哪里的代码,可以再cpu上跑?,2023-05-06.0258,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/258,**问题描述 / Problem Description**,91
+92,ModuleNotFoundError: No module named 'modelscope',2023-05-07.0266,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/266,安装这个,92
+93,加载lora微调模型时,lora参数加载成功,但显示模型未成功加载?,2023-05-08.0270,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/270,什么原因呀?,93
+94,[BUG] 运行webui.py报错:name 'EMBEDDING_DEVICE' is not defined,2023-05-08.0274,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/274,解决了,我修改model_config时候把这个变量改错了,94
+95,基于ptuning训练完成,新老模型都进行了加载,但是只有新的,2023-05-08.0280,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/280,licitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.,95
+96,[BUG] 使用chatyuan模型时,对话Error,has no attribute 'stream_chat',2023-05-08.0282,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/282,**问题描述 / Problem Description**,96
+97,chaglm调用过程中 _call提示有一个 stop,2023-05-09.0286,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/286,**功能描述 / Feature Description**,97
+98,Logger._log() got an unexpected keyword argument 'end',2023-05-10.0295,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/295,使用cli_demo的时候,加载一个普通txt文件,输入问题后,报错:“TypeError: Logger._log() got an unexpected keyword argument 'end'”,98
+99,[BUG] 请问可以解释下这个FAISS.similarity_search_with_score_by_vector = similarity_search_with_score_by_vector的目的吗,2023-05-10.0296,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/296,我不太明白这个库自己写的similarity_search_with_score_by_vector方法做的事情,因为langchain原版的similarity_search_with_score_by_vector只是search faiss之后把返回的topk句子组合起来。我觉得原版理解起来没什么问题,但是这个库里自己写的我就没太看明白多做了什么其他的事情,因为没有注释。,99
+100,[BUG] Windows下上传中文文件名文件,faiss无法生成向量数据库文件,2023-05-11.0318,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/318,**问题描述 / Problem Description**,100
+101,cli_demo中的流式输出能否接着前一答案输出?,2023-05-11.0320,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/320,现有流式输出结果样式为:,101
+102,内网部署时网页无法加载,能否增加离线静态资源,2023-05-12.0326,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/326,内网部署时网页无法加载,能否增加离线静态资源,102
+103,我想把文件字符的编码格式改为encoding='utf-8'在哪修改呢,因为会有ascii codec can't decode byte报错,2023-05-14.0360,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/360,上传中文的txt文件时报错,编码格式为utf-8,103
+104,Batches的进度条是在哪里设置的?能否关闭显示?,2023-05-15.0366,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/366,"使用cli_demo.py进行命令行测试时,每句回答前都有个Batches的进度条",104
+105,ImportError: dlopen: cannot load any more object with static TLS or Segmentation fault,2023-05-15.0368,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/368,**问题描述 / Problem Description**,105
+106,读取PDF时报错,2023-05-16.0373,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/373,在Colab上执行cli_demo.py时,在路径文件夹里放了pdf文件,在加载的过程中会显示错误,然后无法加载PDF文件,106
+107,[BUG] webui报错 InvalidURL,2023-05-16.0375,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/375,python 版本:3.8.16,107
+108,[FEATURE] 如果让回答不包含出处,应该怎么处理,2023-05-16.0380,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/380,**功能描述 / Feature Description**,108
+109,加载PDF文件时,出现 unsupported colorspace for 'png',2023-05-16.0381,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/381,**问题描述 / Problem Description**,109
+110,'ascii' codec can't encode characters in position 14-44: ordinal not in range(128) 经典bug,2023-05-16.0382,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/382,添加了知识库之后进行对话,之后再新增知识库就会出现这个问题。,110
+111,微信群人数超过200了,扫码进不去了,群主可以再创建一个新群吗,2023-05-17.0391,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/391,**功能描述 / Feature Description**,111
+112,TypeError: 'ListDocsResponse' object is not subscriptable,2023-05-17.0393,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/393,应该是用remain_docs.code和remain_docs.data吧?吗?,112
+113,[BUG] 加载chatglm模型报错:'NoneType' object has no attribute 'message_types_by_name',2023-05-17.0398,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/398,**问题描述 / Problem Description**,113
+114,[BUG] 执行 python webui.py 没有报错,但是ui界面提示 Something went wrong Expecting value: line 1 column 1 (char 0,2023-05-18.0399,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/399,**环境配置**,114
+115,启动后调用api接口正常,过一会就不断的爆出 Since the angle classifier is not initialized,2023-05-18.0404,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/404,**问题描述 / Problem Description**,115
+116,[BUG] write_check_file方法中,open函数未指定编码,2023-05-18.0408,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/408,"def write_check_file(filepath, docs):",116
+117,导入的PDF中存在图片,有大概率出现 “unsupported colorspace for 'png'”异常,2023-05-18.0409,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/409,"pix = fitz.Pixmap(doc, img[0])",117
+118,请问流程图是用什么软件画的,2023-05-18.0410,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/410,draw.io,118
+119,mac 加载模型失败,2023-05-19.0417,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/417,Explicitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.,119
+120,使用GPU本地运行知识库问答,提问第一个问题出现异常。,2023-05-20.0419,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/419,配置文件model_config.py为:,120
+121,想加入讨论群,2023-05-20.0420,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/420,OK,121
+122,有没有直接调用LLM的API,目前只有知识库的API?,2023-05-22.0426,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/426,-------------------------------------------------------------------------------,122
+123,上传文件后出现 ERROR __init__() got an unexpected keyword argument 'autodetect_encoding',2023-05-22.0428,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/428,"上传文件后出现这个问题:ERROR 2023-05-22 11:46:19,568-1d: __init__() got an unexpected keyword argument 'autodetect_encoding'",123
+124,想问下README中用到的流程图用什么软件画的,2023-05-22.0431,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/431,**功能描述 / Feature Description**,124
+125,No matching distribution found for langchain==0.0.174,2023-05-23.0436,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/436,ERROR: Could not find a version that satisfies the requirement langchain==0.0.174 ,125
+126,[FEATURE] bing是必须的么?,2023-05-23.0437,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/437,从这个[脚步](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/configs/model_config.py#L129)里面发现需要申请bing api,如果不申请,纯用模型推理不可吗?,126
+127,同一台环境下部署了5.22号更新的langchain-chatglm v0.1.13和之前的版本,回复速度明显变慢,2023-05-23.0442,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/442,新langchain-chatglm v0.1.13版本速度很慢,127
+128,Error reported during startup,2023-05-23.0443,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/443,Traceback (most recent call last):,128
+129,"ValueError: not enough values to unpack (expected 2, got 1)on of the issue",2023-05-24.0449,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/449,"File "".cache\huggingface\modules\transformers_modules\chatglm-6b-int4\modeling_chatglm.py"", line 1280, in chat",129
+130,[BUG] API部署,流式输出的函数,少了个question,2023-05-24.0451,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/451,**问题描述 / Problem Description**,130
+131,项目结构的简洁性保持,2023-05-24.0454,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/454,**功能描述 / Feature Description**,131
+132,项目群扫码进不去了,2023-05-24.0455,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/455,项目群扫码进不去了,是否可以加一下微信拉我进群,谢谢!微信号:daniel-0527,132
+133,请求拉我入群讨论,海硕一枚,专注于LLM等相关技术,2023-05-24.0461,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/461,**功能描述 / Feature Description**,133
+134,[BUG] chatglm-6b模型报错OSError: Error no file named pytorch_model.bin found in directory /chatGLM/model/model-6b,2023-05-26.0474,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/474,**1、简述:**,134
+135,现在本项目交流群二维码扫描不进去了,需要群主通过,2023-05-27.0478,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/478,现在本项目交流群二维码扫描不进去了,需要群主通过,135
+136,RuntimeError: Only Tensors of floating point and complex dtype can require gradients,2023-05-28.0483,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/483,刚更新了最新版本:,136
+137,"RuntimeError: ""LayerNormKernelImpl"" not implemented for 'Half'",2023-05-28.0484,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/484,"已经解决了 params 只用两个参数 {'trust_remote_code': True, 'torch_dtype': torch.float16}",137
+138,[BUG] 文件未成功加载,请重新上传文件,2023-05-31.0504,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/504,webui.py,138
+139,[BUG] bug 17 ,pdf和pdf为啥还不一样呢?为啥有的pdf能识别?有的pdf识别不了呢?,2023-05-31.0506,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/506,bug 17 ,pdf和pdf为啥还不一样呢?为啥有的pdf能识别?有的pdf识别不了呢?,139
+140,[FEATURE] 简洁阐述功能 / Concise description of the feature,2023-05-31.0513,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/513,**功能描述 / Feature Description**,140
+141,[BUG] webui.py 加载chatglm-6b-int4 失败,2023-06-02.0524,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/524,**问题描述 / Problem Description**,141
+142,[BUG] webui.py 加载chatglm-6b模型异常,2023-06-02.0525,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/525,**问题描述 / Problem Description**,142
+143,增加对chatgpt的embedding和api调用的支持,2023-06-02.0531,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/531,能否支持openai的embedding api和对话的api?,143
+144,[FEATURE] 调整模型下载的位置,2023-06-02.0537,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/537,模型默认下载到 $HOME/.cache/huggingface/,当 C 盘空间不足时无法完成模型的下载。configs/model_config.py 中也没有调整模型位置的参数。,144
+145,[BUG] langchain=0.0.174 出错,2023-06-04.0543,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/543,**问题描述 / Problem Description**,145
+146,[BUG] 更新后加载本地模型路径不正确,2023-06-05.0545,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/545,**问题描述 / Problem Description**,146
+147,SystemError: 8bit 模型需要 CUDA 支持,或者改用量化后模型!,2023-06-06.0550,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/550,"docker 部署后,启动docker,过会儿容器会自动退出,logs报错 SystemError: 8bit 模型需要 CUDA 支持,或者改用量化后模型! [NVIDIA Container Toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) 也已经安装了",147
+148,[BUG] 上传知识库超过1M报错,2023-06-06.0556,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/556,**问题描述 / Problem Description**,148
+149,打开跨域访问后仍然报错,不能请求,2023-06-06.0560,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/560,报错信息:,149
+150,dialogue_answering 里面的代码是不是没有用到?,没有看到调用,2023-06-07.0571,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/571,dialogue_answering 是干啥的,150
+151,[BUG] 响应速度极慢,应从哪里入手优化?48C/128G/8卡,2023-06-07.0573,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/573,运行环境:ubuntu20.04,151
+152,纯CPU环境下运行cli_demo时报错,提示找不到nvcuda.dll,2023-06-08.0576,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/576,本地部署环境是纯CPU,之前的版本在纯CPU环境下能正常运行,但上传本地知识库经常出现encode问题。今天重新git项目后,运行时出现如下问题,请问该如何解决。,152
+153,如何加载本地的embedding模型(text2vec-large-chinese模型文件),2023-06-08.0582,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/582,"因为需要离线部署,所以要把模型放到本地,我修改了chains/local_doc_qa.py中的HuggingFaceEmbeddings(),在其中加了一个cache_folder的参数,保证下载的文件在cache_folder中,model_name是text2vec-large-chinese。如cache_folder='/home/xx/model/text2vec-large-chinese', model_name='text2vec-large-chinese',这样仍然需要联网下载报错,请问大佬如何解决该问题?",153
+154,ChatGLM-6B 在另外服务器安装好了,请问如何修改model.cofnig.py 来使用它的接口呢??,2023-06-09.0588,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/588,我本来想在这加一个api base url 但是运行web.py 发现 还是会去连huggingface 下载模型,154
+155,[BUG] raise partially initialized module 'charset_normalizer' has no attribute 'md__mypyc' when call interface `upload_file`,2023-06-10.0591,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/591,**问题描述 / Problem Description**,155
+156,[BUG] raise OSError: [Errno 101] Network is unreachable when call interface upload_file and upload .pdf files,2023-06-10.0592,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/592,**问题描述 / Problem Description**,156
+157,如果直接用vicuna作为基座大模型,需要修改的地方有哪些?,2023-06-12.0596,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/596,vicuna模型有直接转换好的没有?也就是llama转换之后的vicuna。,157
+158,[BUG] 通过cli.py调用api时抛出AttributeError: 'NoneType' object has no attribute 'get'错误,2023-06-12.0598,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/598,通过`python cli.py start api --ip localhost --port 8001` 命令调用api时,抛出:,158
+159,[BUG] 通过cli.py调用api时直接报错`langchain-ChatGLM: error: unrecognized arguments: start cli`,2023-06-12.0601,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/601,通过python cli.py start cli启动cli_demo时,报错:,159
+160,[BUG] error: unrecognized arguments: --model-dir conf/models/,2023-06-12.0602,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/602,关键字参数修改了吗?有没有文档啊?大佬,160
+161,[BUG] 上传文件全部失败,2023-06-12.0603,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/603,ERROR: Exception in ASGI application,161
+162,[BUG] config 使用 chatyuan 无法启动,2023-06-12.0604,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/604,"""chatyuan"": {",162
+163,使用fashchat api之后,后台报错APIError 如图所示,2023-06-12.0606,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/606,我按照https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/fastchat.md,163
+164,[BUG] 启用上下文关联,每次embedding搜索到的内容都会比前一次多一段,2023-06-13.0613,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/613,**问题描述 / Problem Description**,164
+165,local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法,其父类FAISS和VectorStore中也只有from_texts方法[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-14.0619,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/619,local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法,其父类FAISS和VectorStore中也只有from_texts方法,165
+166,[BUG] TypeError: similarity_search_with_score_by_vector() got an unexpected keyword argument 'filter',2023-06-14.0624,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/624,**问题描述 / Problem Description**,166
+167,please delete this issue,2023-06-15.0633,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/633,"sorry, incorrect submission. Please remove this issue!",167
+168,[BUG] vue前端镜像构建失败,2023-06-15.0635,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/635,**问题描述 / Problem Description**,168
+169,ChatGLM-6B模型能否回答英文问题?,2023-06-15.0640,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/640,大佬,请问一下,如果本地知识文档是英文,ChatGLM-6B模型能否回答英文问题?不能的话,有没有替代的模型推荐,期待你的回复,谢谢,169
+170,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-16.0644,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/644,**问题描述 / Problem Description**,170
+171,KeyError: 3224,2023-06-16.0645,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/645,```,171
diff --git a/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.jsonl b/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..fd2040e151f67fc60bc6f921fe9b357aa4c458ae
--- /dev/null
+++ b/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.jsonl
@@ -0,0 +1,172 @@
+{"title": "加油~以及一些建议", "file": "2023-03-31.0002", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/2", "detail": "加油,我认为你的方向是对的。", "id": 0}
+{"title": "当前的运行环境是什么,windows还是Linux", "file": "2023-04-01.0003", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/3", "detail": "当前的运行环境是什么,windows还是Linux,python是什么版本?", "id": 1}
+{"title": "请问这是在CLM基础上运行吗?", "file": "2023-04-01.0004", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/4", "detail": "请问是不是需要本地安装好clm并正常运行的情况下,再按文中的步骤执行才能运行起来?", "id": 2}
+{"title": "[复现问题] 构造 prompt 时从知识库中提取的文字乱码", "file": "2023-04-01.0005", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/5", "detail": "hi,我在尝试复现 README 中的效果,也使用了 ChatGLM-6B 的 README 作为输入文本,但发现从知识库中提取的文字是乱码,导致构造的 prompt 不可用。想了解如何解决这个问题。", "id": 3}
+{"title": "后面能否加入上下文对话功能?", "file": "2023-04-02.0006", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/6", "detail": "目前的get_wiki_agent_answer函数中已经实现了历史消息传递的功能,后面我再确认一下是否有langchain中model调用过程中是否传递了chat_history。", "id": 4}
+{"title": "请问:纯cpu可以吗?", "file": "2023-04-03.0007", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/7", "detail": "很酷的实现,极大地开拓了我的眼界!很顺利的在gpu机器上运行了", "id": 5}
+{"title": "运行报错:AttributeError: 'NoneType' object has no attribute 'message_types_by_name'", "file": "2023-04-03.0008", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/8", "detail": "报错:", "id": 6}
+{"title": "运行环境:GPU需要多大的?", "file": "2023-04-03.0009", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/9", "detail": "如果按照THUDM/ChatGLM-6B的说法,使用的GPU大小应该在13GB左右,但运行脚本后,占用了24GB还不够。", "id": 7}
+{"title": "请问本地知识的格式是什么?", "file": "2023-04-03.0010", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/10", "detail": "已测试格式包括docx、md文件中的文本信息,具体格式可以参考 [langchain文档](https://python.langchain.com/en/latest/modules/indexes/document_loaders/examples/unstructured_file.html?highlight=pdf#)", "id": 8}
+{"title": "24G的显存还是爆掉了,是否支持双卡运行", "file": "2023-04-03.0011", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/11", "detail": "RuntimeError: CUDA out of memory. Tried to allocate 96.00 MiB (GPU 0; 23.70 GiB total capacity; 22.18 GiB already allocated; 12.75 MiB free; 22.18 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF", "id": 9}
+{"title": "你怎么知道embeddings方式和模型训练时候的方式是一样的?", "file": "2023-04-03.0012", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/12", "detail": "embedding和LLM的方式不用一致,embedding能够解决语义检索的需求就行。这个项目里用到embedding是在对本地知识建立索引和对问句转换成向量的过程。", "id": 10}
+{"title": "是否能提供本地知识文件的格式?", "file": "2023-04-04.0013", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/13", "detail": "是否能提供本地知识文件的格式?", "id": 11}
+{"title": "是否可以像清华原版跑在8G一以下的卡?", "file": "2023-04-04.0016", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/16", "detail": "是否可以像清华原版跑在8G一以下的卡?我的8G卡爆显存了🤣🤣🤣", "id": 12}
+{"title": "请教一下langchain协调使用向量库和chatGLM工作的", "file": "2023-04-05.0018", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/18", "detail": "代码里面这段是创建问答模型的,会接入ChatGLM和本地语料的向量库,langchain回答的时候是怎么个优先顺序?先搜向量库,没有再找chatglm么? 还是什么机制?", "id": 13}
+{"title": "在mac m2max上抛出了ValueError: 150001 is not in list这个异常", "file": "2023-04-05.0019", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/19", "detail": "我把chatglm_llm.py加载模型的代码改成如下", "id": 14}
+{"title": "程序运行后一直卡住", "file": "2023-04-05.0020", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/20", "detail": "感谢作者的付出,不过本人在运行时出现了问题,请大家帮助。", "id": 15}
+{"title": "问一下chat_history的逻辑", "file": "2023-04-06.0022", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/22", "detail": "感谢开源。", "id": 16}
+{"title": "为什么每次运行都会loading checkpoint", "file": "2023-04-06.0023", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/23", "detail": "我把这个embeding模型下载到本地后,无法正常启动。", "id": 17}
+{"title": "本地知识文件能否上传一些示例?", "file": "2023-04-06.0025", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/25", "detail": "如题,怎么构造知识文件,效果更好?能否提供一个样例", "id": 18}
+{"title": "What version of you are using?", "file": "2023-04-06.0026", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/26", "detail": "Hi Panda, I saw the `pip install -r requirements` command in README, and want to confirm you are using python2 or python3? because my pip and pip3 version are all is 22.3.", "id": 19}
+{"title": "有兴趣交流本项目应用的朋友可以加一下微信群", "file": "2023-04-07.0027", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/27", "detail": "![IMG_1630](https://user-images.githubusercontent.com/5668498/230533162-8b9bfcdd-249c-4efe-b066-4f9ba2ce9f23.jpeg)", "id": 20}
+{"title": "本地知识越多,回答时检索的时间是否会越长", "file": "2023-04-07.0029", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/29", "detail": "是的 因为需要进行向量匹配检索", "id": 21}
+{"title": "爲啥最後還是報錯 哭。。", "file": "2023-04-07.0030", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/30", "detail": "Failed to import transformers.models.t5.configuration_t5 because of the following error (look up to see", "id": 22}
+{"title": "对话到第二次的时候就报错UnicodeDecodeError: 'utf-8' codec can't decode", "file": "2023-04-07.0031", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/31", "detail": "对话第一次是没问题的,模型返回输出后又给到请输入你的问题,我再输入问题就报错", "id": 23}
+{"title": "用的in4的量化版本,推理的时候显示需要申请10Gb的显存", "file": "2023-04-07.0033", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/33", "detail": "File \"/root/.cache/huggingface/modules/transformers_modules/chatglm-6b-int4-qe/modeling_chatglm.py\", line 581, in forward", "id": 24}
+{"title": "使用colab运行,python3.9,提示包导入有问题", "file": "2023-04-07.0034", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/34", "detail": "from ._util import is_directory, is_path", "id": 25}
+{"title": "运行失败,Loading checkpoint未达到100%被kill了,请问下是什么原因?", "file": "2023-04-07.0035", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/35", "detail": "日志如下:", "id": 26}
+{"title": "弄了个交流群,自己弄好多细节不会,大家技术讨论 加connection-image 我来拉你", "file": "2023-04-08.0036", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/36", "detail": "自己搞好多不清楚的,一起来弄吧。。准备搞个部署问题的解决文档出来", "id": 27}
+{"title": "Error using the new version with langchain", "file": "2023-04-09.0043", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/43", "detail": "Error with the new changes:", "id": 28}
+{"title": "程序报错torch.cuda.OutOfMemoryError如何解决?", "file": "2023-04-10.0044", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/44", "detail": "报错详细信息如下:", "id": 29}
+{"title": "qa的训练数据格式是如何设置的", "file": "2023-04-10.0045", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/45", "detail": "本项目不是使用微调的方式,所以并不涉及到训练过程。", "id": 30}
+{"title": "The FileType.UNK file type is not supported in partition. 解决办法", "file": "2023-04-10.0046", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/46", "detail": "ValueError: Invalid file /home/yawu/Documents/langchain-ChatGLM-master/data. The FileType.UNK file type is not supported in partition.", "id": 31}
+{"title": "如何读取多个txt文档?", "file": "2023-04-10.0047", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/47", "detail": "如题,请教一下如何读取多个txt文档?示例代码中只给了读一个文档的案例,这个input我换成string之后也只能指定一个文档,无法用通配符指定多个文档,也无法传入多个文件路径的列表。", "id": 32}
+{"title": "nltk package unable to either download or load local nltk_data folder", "file": "2023-04-10.0049", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/49", "detail": "I'm running this project on an offline Windows Server environment so I download the Punkt and averaged_perceptron_tagger tokenizer in this directory:", "id": 33}
+{"title": "requirements.txt中需要指定langchain版本", "file": "2023-04-11.0055", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/55", "detail": "langchain版本0.116下无法引入RetrievalQA,需要指定更高版本(0.136版本下无问题)", "id": 34}
+{"title": "Demo演示无法给出输出内容", "file": "2023-04-12.0059", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/59", "detail": "你好,测试了项目自带新闻稿示例和自行上传的一个文本,可以加载进去,但是无法给出答案,请问属于什么情况,如何解决,谢谢。PS: 1、今天早上刚下载全部代码;2、硬件服务器满足要求;3、按操作说明正常操作。", "id": 35}
+{"title": "群人数过多无法进群,求帮忙拉进群", "file": "2023-04-12.0061", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/61", "detail": "您好,您的群人数超过了200人,目前无法通过二维码加群,请问您方便加我微信拉我进群吗?万分感谢", "id": 36}
+{"title": "群人数已满,求大佬拉入群", "file": "2023-04-12.0062", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/62", "detail": "已在README中更新拉群二维码", "id": 37}
+{"title": "requirements中langchain版本错误", "file": "2023-04-12.0065", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/65", "detail": "langchain版本应该是0.0.12而不是0.0.120", "id": 38}
+{"title": "Linux : Searchd in", "file": "2023-04-13.0068", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/68", "detail": "import nltk", "id": 39}
+{"title": "No sentence-transformers model found", "file": "2023-04-13.0069", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/69", "detail": "加载不了这个模型,错误原因是找不到这个模型,但是路径是配置好了的", "id": 40}
+{"title": "Error loading punkt: ", "id": 58}
+{"title": "为啥放到方法调用会出错,这个怎么处理?", "file": "2023-04-20.0150", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/150", "detail": "```python", "id": 59}
+{"title": "No sentence-transformers model found with name C:\\Users\\Administrator/.cache\\torch\\sentence_transformers\\GanymedeNil_text2vec-large-chinese. Creating a new one with MEAN pooling.", "file": "2023-04-21.0154", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/154", "detail": "卡在这块很久是正常现象吗", "id": 60}
+{"title": "微信群需要邀请才能加入", "file": "2023-04-21.0155", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/155", "detail": "RT,给个个人联系方式白", "id": 61}
+{"title": "No sentence-transformers model found with name GanymedeNil/text2vec-large-chinese. Creating a new one with MEAN pooling", "file": "2023-04-21.0156", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/156", "detail": "ls GanymedeNil/text2vec-large-chinese", "id": 62}
+{"title": "embedding会加载两次", "file": "2023-04-23.0159", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/159", "detail": "你好,为什么要这样设置呢,这样会加载两次呀。", "id": 63}
+{"title": "扫二维码加的那个群,群成员满了进不去了", "file": "2023-04-23.0160", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/160", "detail": "如题", "id": 64}
+{"title": "执行python3 cli_demo.py 报错AttributeError: 'NoneType' object has no attribute 'chat'", "file": "2023-04-24.0163", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/163", "detail": "刚开始怀疑是内存不足问题,换成int4,int4-qe也不行,有人知道是什么原因吗", "id": 65}
+{"title": "匹配得分", "file": "2023-04-24.0167", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/167", "detail": "在示例cli_demo.py中返回的匹配文本没有对应的score,可以加上这个feature吗", "id": 66}
+{"title": "大佬有计划往web_ui.py加入打字机功能吗", "file": "2023-04-25.0170", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/170", "detail": "目前在载入了知识库后,单张V100 32G在回答垂直领域的问题时也需要20S以上,没有打字机逐字输出的使用体验还是比较煎熬的....", "id": 67}
+{"title": "Is it possible to use a verctorDB for the embedings?", "file": "2023-04-25.0171", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/171", "detail": "when I play, I have to load the local data again and again when to start. I wonder if it is possible to use", "id": 68}
+{"title": "请问通过lora训练官方模型得到的微调模型文件该如何加载?", "file": "2023-04-25.0173", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/173", "detail": "通过lora训练的方式得到以下文件:", "id": 69}
+{"title": "from langchain.chains import RetrievalQA的代码在哪里?", "file": "2023-04-25.0174", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/174", "detail": "local_doc_qa.py", "id": 70}
+{"title": "哪里有knowledge_based_chatglm.py文件?怎么找不到了??是被替换成cli_demo.py文件了吗?", "file": "2023-04-26.0175", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/175", "detail": "哪里有knowledge_based_chatglm.py文件?怎么找不到了??是被替换成cli_demo.py文件了吗?", "id": 71}
+{"title": "AttributeError: 'Chatbot' object has no attribute 'value'", "file": "2023-04-26.0177", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/177", "detail": "Traceback (most recent call last):", "id": 72}
+{"title": "控制台调api.py报警告", "file": "2023-04-26.0178", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/178", "detail": "you must pass the application as an import string to enable \"reload\" or \"workers\"", "id": 73}
+{"title": "如何加入群聊", "file": "2023-04-27.0183", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/183", "detail": "微信群超过200人了,需要邀请,如何加入呢?", "id": 74}
+{"title": "如何将Chatglm和本地知识相结合", "file": "2023-04-27.0185", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/185", "detail": "您好,我想请教一下怎么才能让知识库匹配到的文本和chatglm生成的相结合,而不是说如果没搜索到,就说根据已知信息无法回答该问题,谢谢", "id": 75}
+{"title": "一点建议", "file": "2023-04-27.0189", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/189", "detail": "1.weiui的get_vector_store方法里面添加一个判断以兼容gradio版本导致的上传异常", "id": 76}
+{"title": "windows环境下,按照教程,配置好conda环境,git完项目,修改完模型路径相关内容后,运行demo报错缺少", "file": "2023-04-28.0194", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/194", "detail": "报错代码如下:", "id": 77}
+{"title": "ValueError: too many values to unpack (expected 2)", "file": "2023-04-28.0198", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/198", "detail": "When i tried to use the non-streaming, `ValueError: too many values to unpack (expected 2)` error came out.", "id": 78}
+{"title": "加载doc后覆盖原本知识", "file": "2023-04-28.0201", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/201", "detail": "加载较大量级的私有知识库后,原本的知识会被覆盖", "id": 79}
+{"title": "自定义知识库回答效果很差", "file": "2023-04-28.0203", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/203", "detail": "请问加了自定义知识库知识库,回答效果很差,是因为数据量太小的原因么", "id": 80}
+{"title": "python310下,安装pycocotools失败,提示低版本cython,实际已安装高版本", "file": "2023-04-29.0208", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/208", "detail": "RT,纯离线环境安装,依赖安装的十分艰难,最后碰到pycocotools,始终无法安装上,求教方法!", "id": 81}
+{"title": "[FEATURE] 支持 RWKV 模型(目前已有 pip package & rwkv.cpp 等等)", "file": "2023-05-01.0216", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/216", "detail": "您好,我是 RWKV 的作者,介绍见:https://zhuanlan.zhihu.com/p/626083366", "id": 82}
+{"title": "[BUG] 为啥主机/服务器不联网不能正常启动服务?", "file": "2023-05-02.0220", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/220", "detail": "**问题描述 / Problem Description**", "id": 83}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-05-03.0222", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/222", "detail": "**local variable 'torch' referenced before assignment**", "id": 84}
+{"title": "不支持txt文件的中文输入", "file": "2023-05-04.0235", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/235", "detail": "vs_path, _ = local_doc_qa.init_knowledge_vector_store(filepath)", "id": 85}
+{"title": "文件均未成功加载,请检查依赖包或替换为其他文件再次上传。 文件未成功加载,请重新上传文件", "file": "2023-05-05.0237", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/237", "detail": "请大佬帮忙解决,谢谢!", "id": 86}
+{"title": "[BUG] 使用多卡时chatglm模型加载两次", "file": "2023-05-05.0241", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/241", "detail": "chatglm_llm.py文件下第129行先加载了一次chatglm模型,第143行又加载了一次", "id": 87}
+{"title": "[BUG] similarity_search_with_score_by_vector函数返回多个doc时的score结果错误", "file": "2023-05-06.0252", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/252", "detail": "**问题描述 / Problem Description**", "id": 88}
+{"title": "可以再建一个交流群吗,这个群满了进不去。", "file": "2023-05-06.0255", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/255", "detail": "上午应该已经在readme里更新过了,如果不能添加可能是网页缓存问题,可以试试看直接扫描img/qr_code_12.jpg", "id": 89}
+{"title": "请问这是什么错误哇?KeyError: 'serialized_input'", "file": "2023-05-06.0257", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/257", "detail": "运行“python webui.py” 后这是什么错误?怎么解决啊?", "id": 90}
+{"title": "修改哪里的代码,可以再cpu上跑?", "file": "2023-05-06.0258", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/258", "detail": "**问题描述 / Problem Description**", "id": 91}
+{"title": "ModuleNotFoundError: No module named 'modelscope'", "file": "2023-05-07.0266", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/266", "detail": "安装这个", "id": 92}
+{"title": "加载lora微调模型时,lora参数加载成功,但显示模型未成功加载?", "file": "2023-05-08.0270", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/270", "detail": "什么原因呀?", "id": 93}
+{"title": "[BUG] 运行webui.py报错:name 'EMBEDDING_DEVICE' is not defined", "file": "2023-05-08.0274", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/274", "detail": "解决了,我修改model_config时候把这个变量改错了", "id": 94}
+{"title": "基于ptuning训练完成,新老模型都进行了加载,但是只有新的", "file": "2023-05-08.0280", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/280", "detail": "licitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.", "id": 95}
+{"title": "[BUG] 使用chatyuan模型时,对话Error,has no attribute 'stream_chat'", "file": "2023-05-08.0282", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/282", "detail": "**问题描述 / Problem Description**", "id": 96}
+{"title": "chaglm调用过程中 _call提示有一个 stop", "file": "2023-05-09.0286", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/286", "detail": "**功能描述 / Feature Description**", "id": 97}
+{"title": "Logger._log() got an unexpected keyword argument 'end'", "file": "2023-05-10.0295", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/295", "detail": "使用cli_demo的时候,加载一个普通txt文件,输入问题后,报错:“TypeError: Logger._log() got an unexpected keyword argument 'end'”", "id": 98}
+{"title": "[BUG] 请问可以解释下这个FAISS.similarity_search_with_score_by_vector = similarity_search_with_score_by_vector的目的吗", "file": "2023-05-10.0296", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/296", "detail": "我不太明白这个库自己写的similarity_search_with_score_by_vector方法做的事情,因为langchain原版的similarity_search_with_score_by_vector只是search faiss之后把返回的topk句子组合起来。我觉得原版理解起来没什么问题,但是这个库里自己写的我就没太看明白多做了什么其他的事情,因为没有注释。", "id": 99}
+{"title": "[BUG] Windows下上传中文文件名文件,faiss无法生成向量数据库文件", "file": "2023-05-11.0318", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/318", "detail": "**问题描述 / Problem Description**", "id": 100}
+{"title": "cli_demo中的流式输出能否接着前一答案输出?", "file": "2023-05-11.0320", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/320", "detail": "现有流式输出结果样式为:", "id": 101}
+{"title": "内网部署时网页无法加载,能否增加离线静态资源", "file": "2023-05-12.0326", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/326", "detail": "内网部署时网页无法加载,能否增加离线静态资源", "id": 102}
+{"title": "我想把文件字符的编码格式改为encoding='utf-8'在哪修改呢,因为会有ascii codec can't decode byte报错", "file": "2023-05-14.0360", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/360", "detail": "上传中文的txt文件时报错,编码格式为utf-8", "id": 103}
+{"title": "Batches的进度条是在哪里设置的?能否关闭显示?", "file": "2023-05-15.0366", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/366", "detail": "使用cli_demo.py进行命令行测试时,每句回答前都有个Batches的进度条", "id": 104}
+{"title": "ImportError: dlopen: cannot load any more object with static TLS or Segmentation fault", "file": "2023-05-15.0368", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/368", "detail": "**问题描述 / Problem Description**", "id": 105}
+{"title": "读取PDF时报错", "file": "2023-05-16.0373", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/373", "detail": "在Colab上执行cli_demo.py时,在路径文件夹里放了pdf文件,在加载的过程中会显示错误,然后无法加载PDF文件", "id": 106}
+{"title": "[BUG] webui报错 InvalidURL", "file": "2023-05-16.0375", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/375", "detail": "python 版本:3.8.16", "id": 107}
+{"title": "[FEATURE] 如果让回答不包含出处,应该怎么处理", "file": "2023-05-16.0380", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/380", "detail": "**功能描述 / Feature Description**", "id": 108}
+{"title": "加载PDF文件时,出现 unsupported colorspace for 'png'", "file": "2023-05-16.0381", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/381", "detail": "**问题描述 / Problem Description**", "id": 109}
+{"title": "'ascii' codec can't encode characters in position 14-44: ordinal not in range(128) 经典bug", "file": "2023-05-16.0382", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/382", "detail": "添加了知识库之后进行对话,之后再新增知识库就会出现这个问题。", "id": 110}
+{"title": "微信群人数超过200了,扫码进不去了,群主可以再创建一个新群吗", "file": "2023-05-17.0391", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/391", "detail": "**功能描述 / Feature Description**", "id": 111}
+{"title": "TypeError: 'ListDocsResponse' object is not subscriptable", "file": "2023-05-17.0393", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/393", "detail": "应该是用remain_docs.code和remain_docs.data吧?吗?", "id": 112}
+{"title": "[BUG] 加载chatglm模型报错:'NoneType' object has no attribute 'message_types_by_name'", "file": "2023-05-17.0398", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/398", "detail": "**问题描述 / Problem Description**", "id": 113}
+{"title": "[BUG] 执行 python webui.py 没有报错,但是ui界面提示 Something went wrong Expecting value: line 1 column 1 (char 0", "file": "2023-05-18.0399", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/399", "detail": "**环境配置**", "id": 114}
+{"title": "启动后调用api接口正常,过一会就不断的爆出 Since the angle classifier is not initialized", "file": "2023-05-18.0404", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/404", "detail": "**问题描述 / Problem Description**", "id": 115}
+{"title": "[BUG] write_check_file方法中,open函数未指定编码", "file": "2023-05-18.0408", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/408", "detail": "def write_check_file(filepath, docs):", "id": 116}
+{"title": "导入的PDF中存在图片,有大概率出现 “unsupported colorspace for 'png'”异常", "file": "2023-05-18.0409", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/409", "detail": "pix = fitz.Pixmap(doc, img[0])", "id": 117}
+{"title": "请问流程图是用什么软件画的", "file": "2023-05-18.0410", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/410", "detail": "draw.io", "id": 118}
+{"title": "mac 加载模型失败", "file": "2023-05-19.0417", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/417", "detail": "Explicitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.", "id": 119}
+{"title": "使用GPU本地运行知识库问答,提问第一个问题出现异常。", "file": "2023-05-20.0419", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/419", "detail": "配置文件model_config.py为:", "id": 120}
+{"title": "想加入讨论群", "file": "2023-05-20.0420", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/420", "detail": "OK", "id": 121}
+{"title": "有没有直接调用LLM的API,目前只有知识库的API?", "file": "2023-05-22.0426", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/426", "detail": "-------------------------------------------------------------------------------", "id": 122}
+{"title": "上传文件后出现 ERROR __init__() got an unexpected keyword argument 'autodetect_encoding'", "file": "2023-05-22.0428", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/428", "detail": "上传文件后出现这个问题:ERROR 2023-05-22 11:46:19,568-1d: __init__() got an unexpected keyword argument 'autodetect_encoding'", "id": 123}
+{"title": "想问下README中用到的流程图用什么软件画的", "file": "2023-05-22.0431", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/431", "detail": "**功能描述 / Feature Description**", "id": 124}
+{"title": "No matching distribution found for langchain==0.0.174", "file": "2023-05-23.0436", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/436", "detail": "ERROR: Could not find a version that satisfies the requirement langchain==0.0.174 ", "id": 125}
+{"title": "[FEATURE] bing是必须的么?", "file": "2023-05-23.0437", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/437", "detail": "从这个[脚步](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/configs/model_config.py#L129)里面发现需要申请bing api,如果不申请,纯用模型推理不可吗?", "id": 126}
+{"title": "同一台环境下部署了5.22号更新的langchain-chatglm v0.1.13和之前的版本,回复速度明显变慢", "file": "2023-05-23.0442", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/442", "detail": "新langchain-chatglm v0.1.13版本速度很慢", "id": 127}
+{"title": "Error reported during startup", "file": "2023-05-23.0443", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/443", "detail": "Traceback (most recent call last):", "id": 128}
+{"title": "ValueError: not enough values to unpack (expected 2, got 1)on of the issue", "file": "2023-05-24.0449", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/449", "detail": "File \".cache\\huggingface\\modules\\transformers_modules\\chatglm-6b-int4\\modeling_chatglm.py\", line 1280, in chat", "id": 129}
+{"title": "[BUG] API部署,流式输出的函数,少了个question", "file": "2023-05-24.0451", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/451", "detail": "**问题描述 / Problem Description**", "id": 130}
+{"title": "项目结构的简洁性保持", "file": "2023-05-24.0454", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/454", "detail": "**功能描述 / Feature Description**", "id": 131}
+{"title": "项目群扫码进不去了", "file": "2023-05-24.0455", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/455", "detail": "项目群扫码进不去了,是否可以加一下微信拉我进群,谢谢!微信号:daniel-0527", "id": 132}
+{"title": "请求拉我入群讨论,海硕一枚,专注于LLM等相关技术", "file": "2023-05-24.0461", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/461", "detail": "**功能描述 / Feature Description**", "id": 133}
+{"title": "[BUG] chatglm-6b模型报错OSError: Error no file named pytorch_model.bin found in directory /chatGLM/model/model-6b", "file": "2023-05-26.0474", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/474", "detail": "**1、简述:**", "id": 134}
+{"title": "现在本项目交流群二维码扫描不进去了,需要群主通过", "file": "2023-05-27.0478", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/478", "detail": "现在本项目交流群二维码扫描不进去了,需要群主通过", "id": 135}
+{"title": "RuntimeError: Only Tensors of floating point and complex dtype can require gradients", "file": "2023-05-28.0483", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/483", "detail": "刚更新了最新版本:", "id": 136}
+{"title": "RuntimeError: \"LayerNormKernelImpl\" not implemented for 'Half'", "file": "2023-05-28.0484", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/484", "detail": "已经解决了 params 只用两个参数 {'trust_remote_code': True, 'torch_dtype': torch.float16}", "id": 137}
+{"title": "[BUG] 文件未成功加载,请重新上传文件", "file": "2023-05-31.0504", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/504", "detail": "webui.py", "id": 138}
+{"title": "[BUG] bug 17 ,pdf和pdf为啥还不一样呢?为啥有的pdf能识别?有的pdf识别不了呢?", "file": "2023-05-31.0506", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/506", "detail": "bug 17 ,pdf和pdf为啥还不一样呢?为啥有的pdf能识别?有的pdf识别不了呢?", "id": 139}
+{"title": "[FEATURE] 简洁阐述功能 / Concise description of the feature", "file": "2023-05-31.0513", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/513", "detail": "**功能描述 / Feature Description**", "id": 140}
+{"title": "[BUG] webui.py 加载chatglm-6b-int4 失败", "file": "2023-06-02.0524", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/524", "detail": "**问题描述 / Problem Description**", "id": 141}
+{"title": "[BUG] webui.py 加载chatglm-6b模型异常", "file": "2023-06-02.0525", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/525", "detail": "**问题描述 / Problem Description**", "id": 142}
+{"title": "增加对chatgpt的embedding和api调用的支持", "file": "2023-06-02.0531", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/531", "detail": "能否支持openai的embedding api和对话的api?", "id": 143}
+{"title": "[FEATURE] 调整模型下载的位置", "file": "2023-06-02.0537", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/537", "detail": "模型默认下载到 $HOME/.cache/huggingface/,当 C 盘空间不足时无法完成模型的下载。configs/model_config.py 中也没有调整模型位置的参数。", "id": 144}
+{"title": "[BUG] langchain=0.0.174 出错", "file": "2023-06-04.0543", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/543", "detail": "**问题描述 / Problem Description**", "id": 145}
+{"title": "[BUG] 更新后加载本地模型路径不正确", "file": "2023-06-05.0545", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/545", "detail": "**问题描述 / Problem Description**", "id": 146}
+{"title": "SystemError: 8bit 模型需要 CUDA 支持,或者改用量化后模型!", "file": "2023-06-06.0550", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/550", "detail": "docker 部署后,启动docker,过会儿容器会自动退出,logs报错 SystemError: 8bit 模型需要 CUDA 支持,或者改用量化后模型! [NVIDIA Container Toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) 也已经安装了", "id": 147}
+{"title": "[BUG] 上传知识库超过1M报错", "file": "2023-06-06.0556", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/556", "detail": "**问题描述 / Problem Description**", "id": 148}
+{"title": "打开跨域访问后仍然报错,不能请求", "file": "2023-06-06.0560", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/560", "detail": "报错信息:", "id": 149}
+{"title": "dialogue_answering 里面的代码是不是没有用到?,没有看到调用", "file": "2023-06-07.0571", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/571", "detail": "dialogue_answering 是干啥的", "id": 150}
+{"title": "[BUG] 响应速度极慢,应从哪里入手优化?48C/128G/8卡", "file": "2023-06-07.0573", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/573", "detail": "运行环境:ubuntu20.04", "id": 151}
+{"title": "纯CPU环境下运行cli_demo时报错,提示找不到nvcuda.dll", "file": "2023-06-08.0576", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/576", "detail": "本地部署环境是纯CPU,之前的版本在纯CPU环境下能正常运行,但上传本地知识库经常出现encode问题。今天重新git项目后,运行时出现如下问题,请问该如何解决。", "id": 152}
+{"title": "如何加载本地的embedding模型(text2vec-large-chinese模型文件)", "file": "2023-06-08.0582", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/582", "detail": "因为需要离线部署,所以要把模型放到本地,我修改了chains/local_doc_qa.py中的HuggingFaceEmbeddings(),在其中加了一个cache_folder的参数,保证下载的文件在cache_folder中,model_name是text2vec-large-chinese。如cache_folder='/home/xx/model/text2vec-large-chinese', model_name='text2vec-large-chinese',这样仍然需要联网下载报错,请问大佬如何解决该问题?", "id": 153}
+{"title": "ChatGLM-6B 在另外服务器安装好了,请问如何修改model.cofnig.py 来使用它的接口呢??", "file": "2023-06-09.0588", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/588", "detail": "我本来想在这加一个api base url 但是运行web.py 发现 还是会去连huggingface 下载模型", "id": 154}
+{"title": "[BUG] raise partially initialized module 'charset_normalizer' has no attribute 'md__mypyc' when call interface `upload_file`", "file": "2023-06-10.0591", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/591", "detail": "**问题描述 / Problem Description**", "id": 155}
+{"title": "[BUG] raise OSError: [Errno 101] Network is unreachable when call interface upload_file and upload .pdf files", "file": "2023-06-10.0592", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/592", "detail": "**问题描述 / Problem Description**", "id": 156}
+{"title": "如果直接用vicuna作为基座大模型,需要修改的地方有哪些?", "file": "2023-06-12.0596", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/596", "detail": "vicuna模型有直接转换好的没有?也就是llama转换之后的vicuna。", "id": 157}
+{"title": "[BUG] 通过cli.py调用api时抛出AttributeError: 'NoneType' object has no attribute 'get'错误", "file": "2023-06-12.0598", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/598", "detail": "通过`python cli.py start api --ip localhost --port 8001` 命令调用api时,抛出:", "id": 158}
+{"title": "[BUG] 通过cli.py调用api时直接报错`langchain-ChatGLM: error: unrecognized arguments: start cli`", "file": "2023-06-12.0601", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/601", "detail": "通过python cli.py start cli启动cli_demo时,报错:", "id": 159}
+{"title": "[BUG] error: unrecognized arguments: --model-dir conf/models/", "file": "2023-06-12.0602", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/602", "detail": "关键字参数修改了吗?有没有文档啊?大佬", "id": 160}
+{"title": "[BUG] 上传文件全部失败", "file": "2023-06-12.0603", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/603", "detail": "ERROR: Exception in ASGI application", "id": 161}
+{"title": "[BUG] config 使用 chatyuan 无法启动", "file": "2023-06-12.0604", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/604", "detail": "\"chatyuan\": {", "id": 162}
+{"title": "使用fashchat api之后,后台报错APIError 如图所示", "file": "2023-06-12.0606", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/606", "detail": "我按照https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/fastchat.md", "id": 163}
+{"title": "[BUG] 启用上下文关联,每次embedding搜索到的内容都会比前一次多一段", "file": "2023-06-13.0613", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/613", "detail": "**问题描述 / Problem Description**", "id": 164}
+{"title": "local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法,其父类FAISS和VectorStore中也只有from_texts方法[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-14.0619", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/619", "detail": "local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法,其父类FAISS和VectorStore中也只有from_texts方法", "id": 165}
+{"title": "[BUG] TypeError: similarity_search_with_score_by_vector() got an unexpected keyword argument 'filter'", "file": "2023-06-14.0624", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/624", "detail": "**问题描述 / Problem Description**", "id": 166}
+{"title": "please delete this issue", "file": "2023-06-15.0633", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/633", "detail": "sorry, incorrect submission. Please remove this issue!", "id": 167}
+{"title": "[BUG] vue前端镜像构建失败", "file": "2023-06-15.0635", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/635", "detail": "**问题描述 / Problem Description**", "id": 168}
+{"title": "ChatGLM-6B模型能否回答英文问题?", "file": "2023-06-15.0640", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/640", "detail": "大佬,请问一下,如果本地知识文档是英文,ChatGLM-6B模型能否回答英文问题?不能的话,有没有替代的模型推荐,期待你的回复,谢谢", "id": 169}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-16.0644", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/644", "detail": "**问题描述 / Problem Description**", "id": 170}
+{"title": "KeyError: 3224", "file": "2023-06-16.0645", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/645", "detail": "```", "id": 171}
diff --git a/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.xlsx b/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.xlsx
new file mode 100644
index 0000000000000000000000000000000000000000..24119cc849e97ddbae577dfb1cc595fb1b512d37
Binary files /dev/null and b/knowledge_base/samples/content/test_files/langchain-ChatGLM_closed.xlsx differ
diff --git a/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.csv b/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.csv
new file mode 100644
index 0000000000000000000000000000000000000000..56ba7ca35d78a60443a7462f3ce99b2bc5d66cf6
--- /dev/null
+++ b/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.csv
@@ -0,0 +1,324 @@
+,title,file,url,detail,id
+0,效果如何优化,2023-04-04.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/14,如图所示,将该项目的README.md和该项目结合后,回答效果并不理想,请问可以从哪些方面进行优化,0
+1,怎么让模型严格根据检索的数据进行回答,减少胡说八道的回答呢,2023-04-04.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/15,举个例子:,1
+2,"When I try to run the `python knowledge_based_chatglm.py`, I got this error in macOS(M1 Max, OS 13.2)",2023-04-07.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/32,```python,2
+3,萌新求教大佬怎么改成AMD显卡或者CPU?,2023-04-10.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/48,把.cuda()去掉就行,3
+4,输出answer的时间很长,是否可以把文本向量化的部分提前做好存储起来?,2023-04-10.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/50,GPU:4090 24G显存,4
+5,报错Use `repo_type` argument if needed.,2023-04-11.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/57,Traceback (most recent call last):,5
+6,无法打开gradio的页面,2023-04-11.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/58,$ python webui.py,6
+7,支持word,那word里面的图片正常显示吗?,2023-04-12.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/60,如题,刚刚从隔壁转过来的,想先了解下,7
+8,detectron2 is not installed. Cannot use the hi_res partitioning strategy. Falling back to partitioning with the fast strategy.,2023-04-12.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/63,能够正常的跑起来,在加载content文件夹中的文件时,每加载一个文件都会提示:,8
+9,cpu上运行webui,step3 asking时报错,2023-04-12.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/66,web运行,文件加载都正常,asking时报错,9
+10,建议弄一个插件系统,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/67,如题弄成stable-diffusion-webui那种能装插件,再开一个存储库给使用者或插件开发,存储或下载插件。,10
+11,请教加载模型出错!?,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/75,AttributeError: module 'transformers_modules.chatglm-6b.configuration_chatglm' has no attribute 'ChatGLMConfig 怎么解决呀,11
+12,从本地知识检索内容的时候,是否可以设置相似度阈值,小于这个阈值的内容不返回,即使会小于设置的VECTOR_SEARCH_TOP_K参数呢?谢谢大佬,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/76,比如 问一些 你好/你是谁 等一些跟本地知识库无关的问题,12
+13,如何改成多卡推理?,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/77,+1,13
+14,能否弄个懒人包,可以一键体验?,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/78,能否弄个懒人包,可以一键体验?,14
+15,连续问问题会导致崩溃,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/79,看上去不是爆内存的问题,连续问问题后,会出现如下报错,15
+16,AttributeError: 'NoneType' object has no attribute 'as_retriever',2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/86,"环境:windows 11, anaconda/python 3.8",16
+17,FileNotFoundError: Could not find module 'nvcuda.dll' (or one of its dependencies). Try using the full path with constructor syntax.,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/87,请检查一下cuda或cudnn是否存在安装问题,17
+18,加载txt文件失败?,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/89,![JppHrGOWFa](https://user-images.githubusercontent.com/109277248/232009383-bf7c46d1-a01e-4e0a-9de6-5b5ed3e36158.jpg),18
+19,NameError: name 'chatglm' is not defined,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/90,"This share link expires in 72 hours. For free permanent hosting and GPU upgrades (NEW!), check out Spaces: https://huggingface.co/spaces",19
+20,打不开地址?,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/91,报错数据如下:,20
+21,加载md文件出错,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/98,运行 webui.py后能访问页面,上传一个md文件后,日志中有错误。等待后能加载完成,提示可以提问了,但提问没反应,日志中有错误。 具体日志如下。,21
+22,建议增加获取在线知识的能力,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/101,建议增加获取在线知识的能力,22
+23,txt 未能成功加载,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/103,hinese. Creating a new one with MEAN pooling.,23
+24,pdf加载失败,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/105,e:\a.txt加载成功了,e:\a.pdf加载就失败,pdf文件里面前面几页是图片,后面都是文字,加载失败没有报更多错误,请问该怎么排查?,24
+25,一直停在文本加载处,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/108,一直停在文本加载处,25
+26," File ""/root/.cache/huggingface/modules/transformers_modules/chatglm-6b/modeling_chatglm.py"", line 440, in forward new_tensor_shape = mixed_raw_layer.size()[:-1] + ( TypeError: torch.Size() takes an iterable of 'int' (item 2 is 'float')",2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/113,按照最新的代码,发现,26
+27,后续会提供前后端分离的功能吗?,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/114,类似这种https://github.com/lm-sys/FastChat/tree/main/fastchat/serve,27
+28,安装依赖报错,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/115,(test) C:\Users\linh\Desktop\langchain-ChatGLM-master>pip install -r requirements.txt,28
+29,问特定问题会出现爆显存,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/117,正常提问没问题。,29
+30,Expecting value: line 1 column 1 (char 0),2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/118,运行后 第一步加载配置一直报错:,30
+31,embedding https://huggingface.co/GanymedeNil/text2vec-large-chinese/tree/main是免费的,效果比对openai的如何?,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/119,-------------------------------------------------------------------------------,31
+32,这是什么错误,在Colab上运行的。,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/120,libcuda.so.1: cannot open shared object file: No such file or directory,32
+33,只想用自己的lora微调后的模型进行对话,不想加载任何本地文档,该如何调整?,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/121,能出一个单独的教程吗,33
+34,"租的gpu,Running on local URL: http://0.0.0.0:7860 To create a public link, set `share=True` in `launch()`. 浏览器上访问不了???",2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/122,(chatglm20230401) root@autodl-container-e82d11963c-10ece0d7:~/autodl-tmp/chatglm/langchain-ChatGLM-20230418# python3.9 webui.py,34
+35,本地部署中的报错请教,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/124,"您好,在本地运行langchain-ChatGLM过程中,环境及依赖的包都已经满足条件,但是运行webui.py,报错如下(运行cli_demo.py报错类似),请问是哪里出了错呢?盼望您的回复,谢谢!",35
+36,报错。The dtype of attention mask (torch.int64) is not bool,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/131,The dtype of attention mask (torch.int64) is not bool,36
+37,[求助] pip install -r requirements.txt 的时候出现以下报错。。。有大佬帮忙看看怎么搞么,下的release里面的包,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/134,$ pip install -r requirements.txt,37
+38,如何提升根据问题搜索到对应知识的准确率,2023-04-19.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/136,外链知识库最大的问题在于问题是短文本,知识是中长文本。如何根据问题精准的搜索到对应的知识是个最大的问题。这类本地化项目不像百度,由无数的网页,基本上每个问题都可以找到对应的页面。,38
+39,是否可以增加向量召回的阈值设定,有些召回内容相关性太低,导致模型胡言乱语,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/140,如题,39
+40,输入长度问题,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/141,感谢作者支持ptuning微调模型。,40
+41,已有部署好的chatGLM-6b,如何通过接口接入?,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/144,已有部署好的chatGLM-6b,如何通过接口接入,而不是重新加载一个模型;,41
+42,执行web_demo.py后,显示Killed,就退出了,是不是配置不足呢?,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/146,![图片](https://user-images.githubusercontent.com/26102866/233256425-c7aab999-11d7-4de9-867b-23ef18d519e4.png),42
+43,执行python cli_demo1.py,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/147,Traceback (most recent call last):,43
+44,报错:ImportError: cannot import name 'GENERATION_CONFIG_NAME' from 'transformers.utils',2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/149,(mychatGLM) PS D:\Users\admin3\zrh\langchain-ChatGLM> python cli_demo.py,44
+45,上传文件并加载知识库时,会不停地出现临时文件,2023-04-21.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/153,环境:ubuntu 18.04,45
+46,向知识库中添加文件后点击”上传文件并加载知识库“后Segmentation fault报错。,2023-04-23.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/161,运行服务后的提示如下:,46
+47,langchain-serve 集成,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/162,Hey 我是来自 [langchain-serve](https://github.com/jina-ai/langchain-serve) 的dev!,47
+48,大佬们,wsl的ubuntu怎么配置用cuda加速,装了运行后发现是cpu在跑,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/164,大佬们,wsl的ubuntu怎么配置用cuda加速,装了运行后发现是cpu在跑,48
+49,在github codespaces docker运行出错,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/165,docker run -d --restart=always --name chatglm -p 7860:7860 -v /www/wwwroot/code/langchain-ChatGLM:/chatGLM chatglm,49
+50,有计划接入Moss模型嘛,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/166,后续会开展测试,目前主要在优化langchain部分效果,如果有兴趣也欢迎提PR,50
+51,怎么实现 API 部署?,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/168,利用 fastapi 实现 API 部署方式,具体怎么实现,有方法说明吗?,51
+52, 'NoneType' object has no attribute 'message_types_by_name'报错,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/169,_HISTOGRAMPROTO = DESCRIPTOR.message_types_by_name['HistogramProto'],52
+53,能否指定自己训练的text2vector模型?,2023-04-25.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/172,请问大佬:,53
+54,关于项目支持的模型以及quantization_bit潜在的影响的问题,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/176,作者您好~,54
+55,运行python3.9 api.py WARNING: You must pass the application as an import string to enable 'reload' or 'workers'.,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/179,api.py文件最下面改成这样试试:,55
+56,ValidationError: 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted (type=value_error.extra),2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/180,ValidationError: 1 validation error for HuggingFaceEmbeddings,56
+57,如果没有检索到相关性比较高的,回答“我不知道”,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/181,如果通过设计system_template,让模型在搜索到的文档都不太相关的情况下回答“我不知道”,57
+58,请问如果不能联网,6B之类的文件从本地上传需要放到哪里,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/182,感谢大佬的项目,很有启发~,58
+59,知识库问答--输入新的知识库名称是中文的话,会报error,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/184,知识库问答--输入新的知识库名称是中文的话,会报error,选择要加载的知识库那里也不显示之前添加的知识库,59
+60,现在能通过问题匹配的相似度值,来直接返回文档中的文段,而不经过模型吗?因为有些答案在文档中,模型自己回答,不能回答文档中的答案,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/186,现在能通过问题匹配的相似度值,来直接返回文档中的文段,而不经过模型吗?因为有些答案在文档中,模型自己回答,不能回答文档中的答案。也就是说,提供向量检索回答+模型回答相结合的策略。如果相似度值高于一定数值,直接返回文档中的文本,没有高于就返回模型的回答或者不知道,60
+61,"TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation",2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/188,"Mac 运行 python3 ./webui.py 报 TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation",61
+62,Not Enough Memory,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/190,"运行命令行程序python cli_demo.py, 已经成功加载pdf文件, 报“DefaultCPUAllocator: not enough memory: you tried to allocate 458288380900 bytes”错误,请问哪里可以配置default memory",62
+63,参与开发问题,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/191,1.是否需要进专门的开发群,63
+64,对话框中代码片段格式需改进,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/192,最好能改进下输出代码片段的格式,目前输出的格式还不友好。,64
+65,请问未来有可能支持belle吗,2023-04-28.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/195,如题,谢谢大佬,65
+66,TypeError: cannot unpack non-iterable NoneType object,2023-04-28.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/200,"When i tried to change the knowledge vector store through `init_knowledge_vector_store`, the error `TypeError: cannot unpack non-iterable NoneType object` came out.",66
+67,生成结果,2023-04-28.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/202,你好,想问一下langchain+chatglm-6B,找到相似匹配的prompt,是直接返回prompt对应的答案信息,还是chatglm-6B在此基础上自己优化答案?,67
+68,在win、ubuntu下都出现这个错误:attributeerror: 't5forconditionalgeneration' object has no attribute 'stream_chat',2023-04-29.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/207,在win、ubuntu。下载完模型后,没办法修改代码以执行本地模型,每次都要重新输入路径; LLM 模型、Embedding 模型支持也都在官网下的,在其他项目(wenda)下可以使用,68
+69,[FEATURE] knowledge_based_chatglm.py: renamed or missing?,2023-04-30.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/210,"Not found. Was it renamed? Or, is it missing? How can I get it?",69
+70,sudo apt-get install -y nvidia-container-toolkit-base执行报错,2023-05-01.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/211,**问题描述 / Problem Description**,70
+71,效果不佳几乎答不上来,2023-05-01.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/212,提供了50条问答的docx文件,71
+72,有没有可能新增一个基于chatglm api调用的方式构建langchain,2023-05-02.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/218,我有两台8G GPU/40G内存的服务器,一个台做成了chatglm的api ;想基于另外一台服务器部署langchain;网上好像没有类似的代码。,72
+73,电脑是intel的集成显卡; 运行时告知我找不到nvcuda.dll,模型无法运行,2023-05-02.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/219,您好,我的电脑是intel的集成显卡,不过CPU是i5-11400 @ 2.60GHz ,内存64G;,73
+74,根据langchain官方的文档和使用模式,是否可以改Faiss为Elasticsearch?会需要做哪些额外调整?求解,2023-05-03.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/221,本人新手小白,由于业务模式的原因(有一些自己的场景和优化),希望利用Elasticsearch做这个体系内部的检索机制,不知道是否可以替换,同时,还会涉及到哪些地方的改动?或者说可能会有哪些其他影响,希望作者和大佬们不吝赐教!,74
+75,请问未来有可能支持t5吗,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/224,请问可能支持基於t5的模型吗?,75
+76,[BUG] 内存溢出 / torch.cuda.OutOfMemoryError:,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/229,**问题描述 / Problem Description**,76
+77,报错 No module named 'chatglm_llm',2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/230,明明已经安装了包,却在python里吊不出来,77
+78,能出一个api部署的描述文档吗,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/233,**功能描述 / Feature Description**,78
+79,使用docs/API.md 出错,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/234,使用API.md文档2种方法,出错,79
+80,加载pdf文档报错?,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/238,ew one with MEAN pooling.,80
+81,上传的本地知识文件后再次上传不能显示,只显示成功了一个,别的上传成功后再次刷新就没了,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/239,您好,项目有很大启发,感谢~,81
+82,创建了新的虚拟环境,安装了相关包,并且自动下载了相关的模型,但是仍旧出现:OSError: Unable to load weights from pytorch checkpoint file for,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/240,![78ac8e663fdc312d0e9d78da95925c4](https://user-images.githubusercontent.com/34124260/236378728-9ea4424f-0f7f-4013-9d33-820b723de321.png),82
+83,[BUG] 数据加载不进来,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/243,使用的.txt格式,utf-8编码,报以下错误,83
+84,不能读取pdf,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/244,请问是webui还是cli_demo,84
+85,本地txt文件有500M,加载的时候很慢,如何提高速度?,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/251,![yayRzxSYHP](https://user-images.githubusercontent.com/109277248/236592902-f5ab338d-c1e9-43dc-ae16-9df2cd3c1378.jpg),85
+86,[BUG] gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/253,gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库,86
+87,[FEATURE] 可以支持 OpenAI 的模型嘛?比如 GPT-3、GPT-3.5、GPT-4;embedding 增加 text-embedding-ada-002,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/254,**功能描述 / Feature Description**,87
+88,[FEATURE] 能否增加对于milvus向量数据库的支持 / Concise description of the feature,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/256,**功能描述 / Feature Description**,88
+89,CPU和GPU上跑,除了速度有区别,准确率效果回答上有区别吗?,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/259,理论上没有区别,89
+90,m1,请问在生成回答时怎么看是否使用了mps or cpu?,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/260,m1,请问在生成回答时怎么看是否使用了mps or cpu?,90
+91,知识库一刷新就没了,2023-05-07.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/263,知识库上传后刷新就没了,91
+92,本地部署报没有模型,2023-05-07.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/267,建议在下载llm和embedding模型至本地后在configs/model_config中写入模型本地存储路径后再运行,92
+93,[BUG] python3: can't open file 'webui.py': [Errno 2] No such file or directory,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/269,**问题描述 / Problem Description**,93
+94,模块缺失提示,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/271,因为已有自己使用的docker环境,直接启动webui.py,提示,94
+95,"运行api.py后,执行curl -X POST ""http://127.0.0.1:7861"" 报错?",2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/272,"执行curl -X POST ""http://127.0.0.1:7861"" \ -H 'Content-Type: application/json' \ -d '{""prompt"": ""你好"", ""history"": []}',报错怎么解决",95
+96,[BUG] colab安装requirements提示protobuf版本问题?,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/273,pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.,96
+97,请问项目里面向量相似度使用了什么方法计算呀?,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/275,基本按照langchain里的FAISS.similarity_search_with_score_by_vector实现,97
+98,[BUG] 安装detectron2后,pdf无法加载,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/276,**问题描述 / Problem Description**,98
+99,[BUG] 使用ChatYuan-V2模型无法流式输出,会报错,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/277,一方面好像是ChatYuan本身不支持stream_chat,有人在clueai那边提了issue他们说还没开发,所以估计这个attribute调不起来;但是另一方面看报错好像是T5模型本身就不是decoder-only模型,所以不能流式输出吧(个人理解),99
+100,[BUG] 无法加载text2vec模型,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/278,**问题描述 / Problem Description**,100
+101,请问能否增加网络搜索功能,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/281,请问能否增加网络搜索功能,101
+102,[FEATURE] 结构化数据sql、excel、csv啥时会支持呐。,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/283,**功能描述 / Feature Description**,102
+103,TypeError: ChatGLM._call() got an unexpected keyword argument 'stop',2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/284,No sentence-transformers model found with name D:\DevProject\langchain-ChatGLM\GanymedeNil\text2vec-large-chinese. Creating a new one with MEAN pooling.,103
+104,关于api.py的一些bug和设计逻辑问题?,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/285,首先冒昧的问一下,这个api.py,开发者大佬们是在自己电脑上测试后确实没问题吗?,104
+105,有没有租用的算力平台上,运行api.py后,浏览器http://localhost:7861/报错,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/287,是不是租用的gpu平台上都会出现这个问题???,105
+106,请问一下项目中有用到文档段落切割方法吗?,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/288,text_load中的文档切割方法用上了吗?在代码中看好像没有用到?,106
+107,"报错 raise ValueError(f""Knowledge base {knowledge_base_id} not found"") ValueError: Knowledge base ./vector_store not found",2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/289,"File ""/root/autodl-tmp/chatglm/langchain-ChatGLM-master/api.py"", line 183, in chat",107
+108,能接入vicuna模型吗,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/290,目前本地已经有了vicuna模型能直接接入吗?,108
+109,[BUG] 提问公式相关问题大概率爆显存,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/291,**问题描述 / Problem Description**,109
+110,安装pycocotools失败,找了好多方法都不能解决。,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/292,**问题描述 / Problem Description**,110
+111,使用requirements安装,PyTorch安装的是CPU版本,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/294,如题目,使用requirements安装,PyTorch安装的是CPU版本,运行程序的时候,也是使用CPU在工作。,111
+112,能不能给一个毛坯服务器的部署教程,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/298,“开发部署”你当成服务器的部署教程用就行了。,112
+113, Error(s) in loading state_dict for ChatGLMForConditionalGeneration:,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/299,运行中出现的问题,7860的端口页面显示不出来,求助。,113
+114,ChatYuan-large-v2模型加载失败,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/300,**实际结果 / Actual Result**,114
+115,新增摘要功能,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/303,你好,后续会考虑新增对长文本信息进行推理和语音理解功能吗?比如生成摘要,115
+116,[BUG] pip install -r requirements.txt 出错,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/304,pip install langchain -i https://pypi.org/simple,116
+117,[BUG] 上传知识库文件报错,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/305,![19621e29eaa547d01213bee53d81e6a](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/7f6ceb46-e494-4b0e-939c-23b585a6d9d8),117
+118,[BUG] AssertionError: Component with id 41 not a valid input component.,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/306,**问题描述 / Problem Description**,118
+119,[BUG] CUDA out of memory with container deployment,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/310,**问题描述 / Problem Description**,119
+120,[FEATURE] 增加微调训练功能,2023-05-11.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/311,**功能描述 / Feature Description**,120
+121,如何使用多卡部署,多个gpu,2023-05-11.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/315,"机器上有多个gpu,如何全使用了",121
+122,请问这个知识库问答,和chatglm的关系是什么,2023-05-11.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/319,这个知识库问答,哪部分关联到了chatglm,是不是没有这个chatglm,知识库问答也可单单拎出来,122
+123,[BUG] 运行的时候报错ImportError: libcudnn.so.8: cannot open shared object file: No such file or directory,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/324,**问题描述 / Problem Description**raceback (most recent call last):,123
+124,webui启动成功,但有报错,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/325,**问题描述 / Problem Description**,124
+125,切换MOSS的时候报错,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/327,danshi但是发布的源码中,,125
+126,vicuna模型是否能接入?,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/328,您好!关于MOSS模型和vicuna模型,都是AutoModelForCausalLM来加载模型的,但是稍作更改(模型路径这些)会报这个错误。这个错误的造成是什么,126
+127,你好,请问一下在阿里云CPU服务器上跑可以吗?可以的话比较理想的cpu配置是什么?,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/330,你好,请问一下在阿里云CPU服务器上跑可以吗?可以的话比较理想的cpu配置是什么?,127
+128,你好,请问8核32g的CPU可以跑多轮对话吗?,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/331,什么样的cpu配置比较好呢?我目前想部署CPU下的多轮对话?,128
+129,[BUG] 聊天内容输入超过10000个字符系统出现错误,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/332,聊天内容输入超过10000个字符系统出现错误,如下图所示:,129
+130,能增加API的多用户访问接口部署吗?,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/333,默认部署程序仅支持单用户访问,多用户则需要排队访问。测试过相关的几个Github多用户工程,但是其中一些仍然不满足要求。本节将系统介绍如何实现多用户同时访问ChatGLM的部署接口,包括http、websocket(流式输出,stream)和web页面等方式,主要目录如下所示。,130
+131,多卡部署,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/334,用单机多卡或多机多卡,fastapi部署模型,怎样提高并发,131
+132,WEBUI能否指定知识库目录?,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/335,**功能描述 / Feature Description**,132
+133,[BUG] Cannot read properties of undefined (reading 'error'),2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/336,**问题描述 / Problem Description**,133
+134,[BUG] 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted.,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/337,模型加载到 100% 后出现问题:,134
+135,上传知识库需要重启能不能修复一下,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/338,挺严重的这个问题,135
+136,[BUG] 4块v100卡爆显存,在LLM会话模式也一样,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/339,**问题描述 / Problem Description**,136
+137,针对上传的文件配置不同的TextSpliter,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/341,1. 目前的ChineseTextSpliter切分对英文尤其是代码文件不友好,而且限制固定长度;导致对话结果不如人意,137
+138,[FEATURE] 未来可增加Bloom系列模型吗?根据甲骨易的测试,这系列中文评测效果不错,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/346,**功能描述 / Feature Description**,138
+139,[BUG] v0.1.12打包镜像后启动webui.py失败 / Concise description of the issue,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/347,**问题描述 / Problem Description**,139
+140,切换MOSS模型时报错,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/349,昨天问了下,说是transformers版本不对,需要4.30.0,发现没有这个版本,今天更新到4.29.1,依旧报错,错误如下,140
+141,[BUG] pdf文档加载失败,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/350,**问题描述 / Problem Description**,141
+142,建议可以在后期增强一波注释,这样也有助于更多人跟进提PR,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/351,知道作者和团队在疯狂更新审查代码,只是建议后续稳定后可以把核心代码进行一些注释的补充,从而能帮助更多人了解各个模块作者的思路从而提出更好的优化。,142
+143,[FEATURE] MOSS 量化版支援,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/353,**功能描述 / Feature Description**,143
+144,[BUG] moss模型无法加载,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/356,**问题描述 / Problem Description**,144
+145,[BUG] load_doc_qa.py 中的 load_file 函数有bug,2023-05-14.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/358,原函数为:,145
+146,[FEATURE] API模式,知识库加载优化,2023-05-14.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/359,如题,当前版本,每次调用本地知识库接口,都将加载一次知识库,是否有更好的方式?,146
+147,运行Python api.py脚本后端部署后,怎么使用curl命令调用?,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/361,也就是说,我现在想做个对话机器人,想和公司的前后端联调?怎么与前后端相互调用呢?可私信,有偿解答!!!,147
+148,上传知识库需要重启能不能修复一下,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/363,上传知识库需要重启能不能修复一下,148
+149,[BUG] pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/364,我的python是3.8.5的,149
+150,pip install gradio 报错,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/367,大佬帮我一下,150
+151,[BUG] pip install gradio 一直卡不动,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/369,![aba82742dd9d4d242181662eb5027a7](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/cd9600d9-f6e7-46b7-b1be-30ed8b99f76b),151
+152,[BUG] 简洁阐述问题 / Concise description of the issue,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/370,初次加载本地知识库成功,但提问后,就无法重写加载本地知识库,152
+153,[FEATURE] 简洁阐述功能 / Concise description of the feature,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/371,**功能描述 / Feature Description**,153
+154,在windows上,模型文件默认会安装到哪,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/372,-------------------------------------------------------------------------------,154
+155,[FEATURE] 兼顾对话管理,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/374,如何在知识库检索的情况下,兼顾对话管理?,155
+156,llm device: cpu embedding device: cpu,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/376,**问题描述 / Problem Description**,156
+157,[FEATURE] 简洁阐述功能 /文本文件的知识点之间使用什么分隔符可以分割?,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/377,**功能描述 / Feature Description**,157
+158,[BUG] 上传文件失败:PermissionError: [WinError 32] 另一个程序正在使用此文件,进程无法访问。,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/379,**问题描述 / Problem Description**,158
+159,[BUG] 执行python api.py 报错,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/383,错误信息,159
+160,model_kwargs extra fields not permitted (type=value_error.extra),2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/384,"大家好,请问这个有遇到的么,?",160
+161,[BUG] 简洁阐述问题 / Concise description of the issue,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/385,执行的时候出现了ls1 = [ls[0]],161
+162,[FEATURE] 性能优化,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/388,**功能描述 / Feature Description**,162
+163,"[BUG] Moss模型问答,RuntimeError: probability tensor contains either inf, nan or element < 0",2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/390,**问题描述 / Problem Description**,163
+164,有没有人知道v100GPU的32G显存,会报错吗?支持V100GPU吗?,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/392,**问题描述 / Problem Description**,164
+165,针对于编码问题比如'gbk' codec can't encode character '\xab' in position 14: illegal multibyte sequence粗浅的解决方法,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/397,**功能描述 / Feature Description**,165
+166,Could not import sentence_transformers python package. Please install it with `pip install sentence_transformers`.,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/400,**问题描述 / Problem Description**,166
+167,支持模型问答与检索问答,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/401,不同的query,根据意图不一致,回答也应该不一样。,167
+168,文本分割的时候,能不能按照txt文件的每行进行分割,也就是按照换行符号\n进行分割???,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/403,下面的代码应该怎么修改?,168
+169,local_doc_qa/local_doc_chat 接口响应是串行,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/405,**问题描述 / Problem Description**,169
+170,"为什么找到出处了,但是还是无法回答该问题?",2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/406,![图片](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/3349611/1fc81d61-2409-4330-9065-fdda1a27c86a),170
+171,"请问下:知识库测试中的:添加单条内容,如果换成文本导入是是怎样的格式?我发现添加单条内容测试效果很好.",2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/412,"我发现在知识库测试中`添加单条内容`,并且勾选`禁止内容分句入库`,即使 `不开启上下文关联`的测试效果都非常好.",171
+172,[BUG] 无法配置知识库,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/413,**问题描述 / Problem Description**,172
+173,[BUG] 部署在阿里PAI平台的EAS上访问页面是白屏,2023-05-19.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/414,**问题描述 / Problem Description**,173
+174,API部署后调用/local_doc_qa/local_doc_chat 返回Knowledge base samples not found,2023-05-19.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/416,入参,174
+175,[FEATURE] 上传word另存为的txt文件报 'ascii' codec can't decode byte 0xb9 in position 6: ordinal not in range(128),2023-05-20.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/421,上传word另存为的txt文件报,175
+176,创建保存的知识库刷新后没有出来,这个知识库是永久保存的吗?可以连外部的 向量知识库吗?,2023-05-21.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/422,创建保存的知识库刷新后没有出来,这个知识库是永久保存的吗?可以连外部的 向量知识库吗?,176
+177,[BUG] 用colab运行,无法加载模型,报错:'NoneType' object has no attribute 'message_types_by_name',2023-05-21.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/423,**问题描述 / Problem Description**,177
+178,请问是否需要用到向量数据库?以及什么时候需要用到向量数据库?,2023-05-21.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/424,目前用的是 text2vec , 请问是否需要用到向量数据库?以及什么时候需要用到向量数据库?,178
+179,huggingface模型引用问题,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/427,它最近似乎变成了一个Error?,179
+180,你好,加载本地txt文件出现这个killed错误,TXT文件有100M左右大小。原因是?谢谢。,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/429,"",180
+181,想请问一下,关于对本地知识的管理是如何管理?例如:通过http API接口添加数据 或者 删除某条数据,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/430,例如:通过http API接口添加、删除、修改 某条数据。,181
+182,[FEATURE] 双栏pdf识别问题,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/432,试了一下模型,感觉对单栏pdf识别的准确性较高,但是由于使用的基本是ocr的技术,对一些双栏pdf论文识别出来有很多问题,请问有什么办法改善吗?,182
+183,部署启动小问题,小弟初学求大佬解答,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/433,1.python loader/image_loader.py时,提示ModuleNotFoundError: No module named 'configs',但是跑python webui.py还是还能跑,183
+184,能否支持检测到目录下文档有增加而去增量加载文档,不影响前台对话,其实就是支持读写分离。如果能支持查询哪些文档向量化了,删除过时文档等就更好了,谢谢。,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/434,**功能描述 / Feature Description**,184
+185,[BUG] 简洁阐述问题 / windows 下cuda错误,请用https://github.com/Keith-Hon/bitsandbytes-windows.git,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/435,pip install git+https://github.com/Keith-Hon/bitsandbytes-windows.git,185
+186,"[BUG] from commit 33bbb47, Required library version not found: libbitsandbytes_cuda121_nocublaslt.so. Maybe you need to compile it from source?",2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/438,**问题描述 / Problem Description**,186
+187,[BUG] 简洁阐述问题 / Concise description of the issue上传60m的txt文件报错,显示超时,请问这个能上传的文件大小有限制吗,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/439,"ERROR 2023-05-23 11:13:09,627-1d: Timeout reached while detecting encoding for ./docs/GLM模型格式数据.txt",187
+188,[BUG] TypeError: issubclass() arg 1 must be a class,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/440,**问题描述**,188
+189,"执行python3 webui.py后,一直提示”模型未成功加载,请到页面左上角""模型配置""选项卡中重新选择后点击""加载模型""按钮“",2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/441,**问题描述 / Problem Description**,189
+190,是否能提供网页文档得导入支持,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/444,现在很多都是在线文档作为协作得工具,所以通过URL导入在线文档需求更大,190
+191,[BUG] history 索引问题,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/445,在比较对话框的history和模型chat function 中的history时, 发现并不匹配,在传入 llm._call 时,history用的索引是不是有点问题,导致上一轮对话的内容并不输入给模型。,191
+192,[BUG] moss_llm没有实现,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/447,有些方法没支持,如history_len,192
+193,请问langchain-ChatGLM如何删除一条本地知识库的数据?,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/448,例如:用户刚刚提交了一条错误的数据到本地知识库中了,现在如何在本地知识库从找到,并且对此删除。,193
+194,[BUG] 简洁阐述问题 / UnboundLocalError: local variable 'resp' referenced before assignment,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/450,"在最新一版的代码中, 运行api.py 出现了以上错误(UnboundLocalError: local variable 'resp' referenced before assignment), 通过debug的方式观察到local_doc_qa.llm.generatorAnswer(prompt=question, history=history,streaming=True)可能不返回任何值。",194
+195,请问有没有 PROMPT_TEMPLATE 能让模型不回答敏感问题,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/452,## PROMPT_TEMPLATE问题,195
+196,[BUG] 测试环境 Python 版本有误,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/456,**问题描述 / Problem Description**,196
+197,[BUG] webui 部署后样式不正确,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/458,**问题描述 / Problem Description**,197
+198,配置默认LLM模型的问题,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/459,**问题描述 / Problem Description**,198
+199,[FEATURE]是时候更新一下autoDL的镜像了,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/460,如题,跑了下autoDL的镜像,发现是4.27号的,git pull新版本的代码功能+老的依赖环境,各种奇奇怪怪的问题。,199
+200,[BUG] tag:0.1.13 以cpu模式下,想使用本地模型无法跑起来,各种路径参数问题,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/462,-------------------------------------------------------------------------------,200
+201,[BUG] 有没有同学遇到过这个错!!!加载本地txt文件出现这个killed错误,TXT文件有100M左右大小。,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/463,运行cli_demo.py。是本地的txt文件太大了吗?100M左右。,201
+202,API版本能否提供WEBSOCKET的流式接口,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/464,webui 版本中,采用了WS的流式输出,整体感知反应很快,202
+203,[BUG] 安装bug记录,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/465,按照[install文档](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/INSTALL.md)安装的,,203
+204,VUE的pnmp i执行失败的修复-用npm i命令即可,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/466,感谢作者!非常棒的应用,用的很开心。,204
+205,请教个问题,有没有人知道cuda11.4是否支持???,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/467,请教个问题,有没有人知道cuda11.4是否支持???,205
+206,请问有实现多轮问答中基于问题的搜索上下文关联么,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/468,在基于知识库的多轮问答中,第一个问题讲述了一个主题,后续的问题描述没有包含这个主题的关键词,但又存在上下文的关联。如果用后续问题去搜索知识库有可能会搜索出无关的信息,从而导致大模型无法正确回答问题。请问这个项目要考虑这种情况吗?,206
+207,[BUG] 内存不足的问题,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/470,我用了本地的chatglm-6b-int4模型,然后显示了内存不足(win10+32G内存+1080ti11G),一般需要多少内存才足够?这个bug应该如何解决?,207
+208,[BUG] 纯内网环境安装pycocotools失败,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/472,**问题描述 / Problem Description**,208
+209,[BUG] webui.py 重新加载模型会导致 KeyError,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/473,**问题描述 / Problem Description**,209
+210,chatyuan无法使用,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/475,**问题描述 / Problem Description**,210
+211,[BUG] 文本分割模型AliTextSplitter存在bug,会把“.”作为分割符,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/476,"阿里达摩院的语义分割模型存在bug,默认会把"".”作为分割符进行分割而不管上下文语义。是否还有其他分割符则未知。建议的修改方案:把“.”统一替换为其他字符,分割后再替换回来。或者添加其他分割模型。",211
+212,[BUG] RuntimeError: Error in faiss::FileIOReader::FileIOReader(const char*) a,2023-05-27.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/479,**问题描述 / Problem Description**,212
+213,[FEATURE] 安装,为什么conda create要额外指定路径 用-p ,而不是默认的/envs下面,2023-05-28.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/481,##**功能描述 / Feature Description**,213
+214,[小白求助] 通过Anaconda执行webui.py后,无法打开web链接,2023-05-28.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/485,在执行webui.py命令后,http://0.0.0.0:7860复制到浏览器后无法打开,显示“无法访问此网站”。,214
+215,[BUG] 使用 p-tuningv2后的模型,重新加载报错,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/486,把p-tunningv2训练完后的相关文件放到了p-tunningv2文件夹下,勾选使用p-tuningv2点重新加载模型,控制台输错错误信息:,215
+216,[小白求助] 服务器上执行webui.py后,在本地无法打开web链接,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/487,此项目执行在xxx.xx.xxx.xxx服务器上,我在webui.py上的代码为 (demo,216
+217,[FEATURE] 能不能支持VisualGLM-6B,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/488,**功能描述 / Feature Description**,217
+218,你好,问一下各位,后端api部署的时候,支持多用户同时问答吗???,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/489,支持多用户的话,最多支持多少用户问答?根据硬件而定吧?,218
+219,V100GPU显存占满,而利用率却为0,这是为什么?,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/491,"",219
+220,[求助] 如果在公司内部搭建产品知识库,使用INT-4模型,200人规模需要配置多少显存的服务器?,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/492,如题,计划给公司搭一个在线知识库。,220
+221,你好,请教个问题,目前问答回复需要20秒左右,如何提高速度?V10032G服务器。,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/493,**问题描述 / Problem Description**,221
+222,[FEATURE] 如何实现只匹配下文,而不要上文的结果,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/494,在构建自己的知识库时,主要采用问答对的形式,那么也就是我需要的回答是在我的问题下面的内容,但是目前设置了chunk_size的值以后匹配的是上下文的内容,但我实际并不需要上文的。为了实现更完整的展示下面的答案,我只能调大chunk_size的值,但实际上上文的一半内容都是我不需要的。也就是扔了一半没用的东西给prompt,在faiss.py中我也没找到这块的一些描述,请问该如何进行修改呢?,222
+223,你好,问一下,我调用api.py部署,为什么用ip加端口可以使用postman调用,而改为域名使用postman无法调用?,2023-05-30.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/497,![5ufBSWxLyF](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/70e2fbac-5699-48d0-b0d1-3dc84fd042c2),223
+224,调用api.py中的stream_chat,返回source_documents中出现中文乱码。,2023-05-30.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/498,-------------------------------------------------------------------------------,224
+225,[BUG] 捉个虫,api.py中的stream_chat解析json问题,2023-05-30.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/501,**问题描述 / Problem Description**,225
+226,windows本地部署遇到了omp错误,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/502,**问题描述 / Problem Description**,226
+227,"[BUG] bug14 ,""POST /local_doc_qa/upload_file HTTP/1.1"" 422 Unprocessable Entity",2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/503,上传的文件报错,返回错误,api.py,227
+228,你好,请教个问题,api.py部署的时候,如何改为多线程调用?谢谢,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/505,目前的api.py脚本不支持多线程,228
+229,你好,请教一下。api.py部署的时候,能不能提供给后端流失返回结果。,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/507,curl -X 'POST' \,229
+230,流式输出,流式接口,使用server-sent events技术。,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/508,想这样一样,https://blog.csdn.net/weixin_43228814/article/details/130063010,230
+231,计划增加流式输出功能吗?ChatGLM模型通过api方式调用响应时间慢怎么破,Fastapi流式接口来解惑,能快速提升响应速度,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/509,**问题描述 / Problem Description**,231
+232,[BUG] 知识库上传时发生ERROR (could not open xxx for reading: No such file or directory),2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/510,**问题描述 / Problem Description**,232
+233,api.py脚本打算增加SSE流式输出吗?,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/511,curl调用的时候可以检测第一个字,从而提升回复的体验,233
+234,[BUG] 使用tornado实现webSocket,可以多个客户端同时连接,并且实现流式回复,但是多个客户端同时使用,答案就很乱,是模型不支持多线程吗,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/512,import asyncio,234
+235,支持 chinese_alpaca_plus_lora 吗 基于llama的,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/514,支持 chinese_alpaca_plus_lora 吗 基于llama的,https://github.com/ymcui/Chinese-LLaMA-Alpaca这个项目的,235
+236,[BUG] 现在能读图片的pdf了,但是文字的pdf反而读不了了,什么情况???,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/515,**问题描述 / Problem Description**,236
+237,在推理的过程中卡住不动,进程无法正常结束,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/516,**问题描述 / Problem Description**,237
+238,curl调用的时候,从第二轮开始,curl如何传参可以实现多轮对话?,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/517,第一轮调用:,238
+239,建议添加api.py部署后的日志管理功能?,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/518,-------------------------------------------------------------------------------,239
+240,有大佬知道,怎么多线程部署api.py脚本吗?,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/519,api.py部署后,使用下面的请求,时间较慢,好像是单线程,如何改为多线程部署api.py:,240
+241,[BUG] 上传文件到知识库 任何格式与内容都永远失败,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/520,上传知识库的时候,传txt无法解析,就算是穿content/sample里的样例txt也无法解析,上传md、pdf等都无法加载,会持续性等待,等到了超过30分钟也不行。,241
+242,关于prompt_template的问题,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/521,请问这段prompt_template是什么意思,要怎么使用?可以给一个具体模板参考下吗?,242
+243,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/522,**问题描述 / Problem Description**,243
+244,"中文分词句号处理(关于表达金额之间的""."")",2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/523,建议处理12.6亿元的这样的分词,最好别分成12 和6亿这样的,需要放到一起,244
+245,ImportError: cannot import name 'inference' from 'paddle',2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/526,在网上找了一圈,有说升级paddle的,我做了还是没有用,有说安装paddlepaddle的,我找了豆瓣的镜像源,但安装报错cannot detect archive format,245
+246,[BUG] webscoket 接口串行问题(/local_doc_qa/stream-chat/{knowledge_base_id}),2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/527,**问题描述 / Problem Description**,246
+247,[FEATURE] 刷新页面更新知识库列表,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/528,**功能描述以及改进方案**,247
+248,[BUG] 使用ptuning微调模型后,问答效果并不好,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/530,### 未调用ptuning,248
+249,[BUG] 多轮对话效果不佳,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/532,在进行多轮对话的时候,无论设置的history_len是多少,效果都不好。事实上我将其设置成了最大值10,但在对话中,仍然无法实现多轮对话:,249
+250,"RuntimeError: MPS backend out of memory (MPS allocated: 18.00 GB, other allocations: 4.87 MB, max allowed: 18.13 GB)",2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/533,**问题描述**,250
+251, 请大家重视这个issue!真正使用肯定是多用户并发问答,希望增加此功能!!!,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/534,这得看你有多少显卡,251
+252,在启动项目的时候如何使用到多张gpu啊?,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/535,**在启动项目的时候如何使用到多张gpu啊?**,252
+253, 使用流式输出的时候,curl调用的格式是什么?,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/536,"app.websocket(""/local_doc_qa/stream-chat/{knowledge_base_id}"")(stream_chat)中的knowledge_base_id应该填什么???",253
+254,使用本地 vicuna-7b模型启动错误,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/538,环境: ubuntu 22.04 cuda 12.1 没有安装nccl,使用rtx2080与m60显卡并行计算,254
+255,为什么会不调用GPU直接调用CPU呢,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/539,我的阿里云配置是16G显存,用默认代码跑webui.py时提示,255
+256,上传多个文件时会互相覆盖,2023-06-03.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/541,1、在同一个知识库中上传多个文件时会互相覆盖,无法结合多个文档的知识,有大佬知道怎么解决吗?,256
+257,[BUG] ‘gcc’不是内部或外部命令/LLM对话只能持续一轮,2023-06-03.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/542,No compiled kernel found.,257
+258,以API模式启动项目却没有知识库的接口列表?,2023-06-04.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/544,请问如何获取知识库的接口列表?如果没有需要自行编写的话,可不可以提供相关的获取方式,感谢,258
+259,程序以API模式启动的时候,如何才能让接口以stream模式被调用呢?,2023-06-05.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/546,作者您好,我在以API模式进行程序启动后,我发现接口响应时间很长,怎么样才能让接口以stream模式被调用呢?我想实现像webui模式的回答那样,259
+260,关于原文中表格转为文本后数据相关度问题。,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/547,原文中表格数据转换为文本,以 (X-Y:值;...) 的格式每一行组织成一句话,但这样做后发现相关度较低,效果很差,有何好的方案吗?,260
+261,启动后LLM和知识库问答模式均只有最后一轮记录,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/548,拉取最新代码,问答时,每次页面只显示最后一次问答记录,需要修改什么参数才可以保留历史记录?,261
+262,提供system message配置,以便于让回答不要超出知识库范围,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/549,**功能描述 / Feature Description**,262
+263,[BUG] 使用p-tunningv2报错,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/551,按照readme的指示把p-tunningv2训练完后的文件放到了p-tunningv2文件夹下,勾选使用p-tuningv2点重新加载模型,控制台提示错误信息:,263
+264,[BUG] 智障,这么多问题,也好意思放出来,浪费时间,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/553,。。。,264
+265,[FEATURE] 我看代码文件中有一个ali_text_splitter.py,为什么不用他这个文本分割器了?,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/554,我看代码文件中有一个ali_text_splitter.py,为什么不用他这个文本分割器了?,265
+266,加载文档函数报错,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/557,"def load_file(filepath, sentence_size=SENTENCE_SIZE):",266
+267,参考指引安装docker后,运行cli_demo.py,提示killed,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/558,root@b3d1bd08095c:/chatGLM# python3 cli_demo.py,267
+268,注意:如果安装错误,注意这两个包的版本 wandb==0.11.0 protobuf==3.18.3,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/559,Error1: 如果启动异常报错 `protobuf` 需要更新到 `protobuf==3.18.3 `,268
+269,知识库对长文的知识相关度匹配不太理想有何优化方向,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/563,我们可能录入一个文章有 1W 字,里面涉及这个文章主题的很多角度问题,我们针对他提问,他相关度匹配的内容和实际我们需要的答案相差很大怎么办。,269
+270,使用stream-chat函数进行流式输出的时候,能使用curl调用吗?,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/565,为什么下面这样调用会报错???,270
+271,有大佬实践过 并行 或者 多线程 的部署方案吗?,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/566,+1,271
+272,多线程部署遇到问题?,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/567,"",272
+273,[BUG] 用fastchat加载vicuna-13b模型进行知识库的问答有token的限制错误,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/569,当我开启fastchat的vicuna-13b的api服务,然后config那里配置好(api本地测试过可以返回结果),然后知识库加载好之后(知识库大概有1000多个文档,用chatGLM可以正常推理),进行问答时出现token超过限制,就问了一句hello;,273
+274,现在的添加知识库,文件多了总是报错,也不知道自己加载了哪些文件,报错后也不知道是全部失败还是一部分成功;希望能有个加载指定文件夹作为知识库的功能,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/574,**功能描述 / Feature Description**,274
+275,[BUG] moss模型本地加载报错,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/577,moss模型本地加载报错:,275
+276,加载本地moss模型报错Can't instantiate abstract class MOSSLLM with abstract methods _history_len,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/578,(vicuna) ps@ps[13:56:20]:/data/chat/langchain-ChatGLM2/langchain-ChatGLM-0.1.13$ python webui.py --model-dir local_models --model moss --no-remote-model,276
+277,[FEATURE] 能增加在前端页面控制prompt_template吗?或是能支持前端页面选择使用哪个prompt?,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/579,目前只能在config里修改一个prompt,想在多个不同场景切换比较麻烦,277
+278,[BUG] streamlit ui的bug,在增加知识库时会报错,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/580,**问题描述 / Problem Description**,278
+279,[FEATURE] webui/webui_st可以支持history吗?目前仅能一次对话,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/581,试了下webui和webui_st都不支持历史对话啊,只能对话一次,不能默认开启所有history吗?,279
+280,启动python cli_demo.py --model chatglm-6b-int4-qe报错,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/585,下载好模型,和相关依赖环境,之间运行`python cli_demo.py --model chatglm-6b-int4-qe`报错了:,280
+281,重新构建知识库报错,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/586,**问题描述 / Problem Description**,281
+282,[FEATURE] 能否屏蔽paddle,我不需要OCR,效果差依赖环境还很复杂,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/587,希望能不依赖paddle,282
+283,question :文档向量化这个可以自己手动实现么?,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/589,现有公司级数据500G+,需要使用这个功能,请问如何手动实现这个向量化,然后并加载,283
+284,view前端能进行流式的返回吗??,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/590,view前端能进行流式的返回吗??,284
+285,"[BUG] Load parallel cpu kernel failed, using default cpu kernel code",2023-06-11.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/594,**问题描述 / Problem Description**,285
+286,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-11.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/595,**问题描述 / Problem Description**,286
+287,我在上传本地知识库时提示KeyError: 'name'错误,本地知识库都是.txt文件,文件数量大约是2000+。,2023-06-12.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/597,"",287
+288,model_config.py中有vicuna-13b-hf模型的配置信息,但是好像还是不可用?,2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/600,@dongyihua543,288
+289,"ImportError: Using SOCKS proxy, but the 'socksio' package is not installed. Make sure to install httpx using `pip install httpx[socks]`.",2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/605,应该代理问题,但是尝试了好多方法都解决不了,,289
+290,[BUG] similarity_search_with_score_by_vector在找不到匹配的情况下出错,2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/607,在设置匹配阈值 VECTOR_SEARCH_SCORE_THRESHOLD 的情况下,vectorstore会返回空,此时上述处理函数会出错,290
+291,[FEATURE] 请问如何搭建英文知识库呢,2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/609,**功能描述 / Feature Description**,291
+292,谁有vicuna权重?llama转换之后的,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/611,**问题描述 / Problem Description**,292
+293,[FEATURE] API能实现上传文件夹的功能么?,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/612,用户懒得全选所有的文件,就想上传个文件夹,请问下API能实现这个功能么?,293
+294,请问在多卡部署后,上传单个文件作为知识库,用的是单卡在生成向量还是多卡?,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/614,目前我检测我本地多卡部署的,好像生成知识库向量的时候用的还是单卡,294
+295,[BUG] python webui.py提示非法指令,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/615,(/data/conda-langchain [root@chatglm langchain-ChatGLM]# python webui.py,295
+296,知识库文件跨行切分问题,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/616,我的知识库文件txt文件,是一行一条知识,用\n分行。,296
+297,[FEATURE] bing搜索问答有流式的API么?,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/617,web端是有这个bing搜索回答,但api接口没有发现,大佬能给个提示么?,297
+298,希望出一个macos m2的安装教程,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/620,mac m2安装,模型加载成功了,知识库文件也上传成功了,但是一问答就会报错,报错内容如下,298
+299,为【出处】提供高亮显示,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/621,具体出处里面,对相关的内容高亮显示,不包含前后文。,299
+300,[BUG] CPU运行cli_demo.py,不回答,hang住,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/622,没有GPU;32G内存的ubuntu机器。,300
+301,关于删除知识库里面的文档后,LLM知识库对话的时候还是会返回该被删除文档的内容,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/623,如题,在vue前端成功执行删除知识库里面文档A.txt后,未能也在faiss索引中也删除该文档,LLM还是会返回这个A.txt的内容,并且以A.txt为出处,未能达到删除的效果,301
+302,"[BUG] 调用知识库进行问答,显存会一直叠加",2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/625,"14G的显存,调用的chatglm-6b-int8模型,进行知识库问答时,最多问答四次就会爆显存了,观察了一下显存使用情况,每一次使用就会增加一次显存,请问这样是正常的吗?是否有什么配置需要开启可以解决这个问题?例如进行一次知识库问答清空上次问题的显存?",302
+303,[BUG] web页面 重新构建数据库 失败,导致 原来的上传的数据库都没了,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/626,web页面 重新构建数据库 失败,导致 原来的上传的数据库都没了,303
+304,在CPU上运行webui.py报错Tensor on device cpu is not on the expected device meta!,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/627,在CPU上运行python webui.py能启动,但最后有:RuntimeError: Tensor on device cpu is not on the expected device meta!,304
+305,"OSError: [WinError 1114] 动态链接库(DLL)初始化例程失败。 Error loading ""E:\xxx\envs\langchain\lib\site-packages\torch\lib\caffe2_nvrtc.dll"" or one of its dependencies.哪位大佬知道如何解决吗?",2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/629,**问题描述 / Problem Description**,305
+306,[BUG] WEBUI删除知识库文档,会导致知识库问答失败,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/632,如题,从知识库已有文件中选择要删除的文件,点击删除后,在问答框输入内容回车报错,306
+307,更新后的版本中,删除知识库中的文件,再提问出现error错误,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/634,针对更新版本,识别到一个问题,过程如下:,307
+308,我配置好了环境,想要实现本地知识库的问答?可是它返回给我的,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/637,没有总结,只有相关度的回复,但是我看演示里面表现的,回复是可以实现总结的,我去查询代码,308
+309,[BUG] NPM run dev can not successfully start the VUE frontend,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/638,**问题描述 / Problem Description**,309
+310,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/639,**问题描述 / Problem Description**,310
+311,提一个模型加载的bug,我在截图中修复了,你们有空可以看一下。,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/642,![model_load_bug](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/59411575/4432adc4-ccdd-45d9-aafc-5f2d1963403b),311
+312,[求助]关于设置embedding model路径的问题,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/643,如题,我之前成功跑起来过一次,但因环境丢失重新配置 再运行webui就总是报错,312
+313,Lora微调后的模型可以直接使用吗,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/646,看model_config.py里是有USE_LORA这个参数的,但是在cli_demo.py和webui.py这两个里面都没有用到,实际测试下来模型没有微调的效果,想问问现在这个功能实现了吗,313
+314,write_check_file在tmp_files目录下生成的load_file.txt是否需要一直保留,占用空间很大,在建完索引后能否删除,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/647,**功能描述 / Feature Description**,314
+315,[BUG] /local_doc_qa/list_files?knowledge_base_id=test删除知识库bug,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/649,1.新建test知识库并上传文件(在vue前端完成并检查后端发现确实生成了test文件夹以及下面的content和vec_store,315
+316,[BUG] vue webui无法加载知识库,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/650,拉取了最新的代码,分别运行了后端api和前端web,点击知识库,始终只能显示simple,无法加载知识库,316
+317,不能本地加载moss模型吗?,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/652,手动下载模型设置local_model_path路径依旧提示缺少文件,该如何正确配置?,317
+318,macos m2 pro docker 安装失败,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/654,macos m2 pro docker 安装失败,318
+319, [BUG] mac m1 pro 运行提示 zsh: segmentation fault,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/655,运行: python webui.py,319
+320,安装 requirements 报错,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/656,(langchainchatglm) D:\github\langchain-ChatGLM>pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple/,320
+321,[BUG] AssertionError,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/658,**问题描述 / Problem Description**,321
+322,[FEATURE] 支持AMD win10 本地部署吗?,2023-06-18.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/660,**功能描述 / Feature Description**,322
diff --git a/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.jsonl b/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a34816e89f4e085b305df0bcf99971a5dcbd490f
--- /dev/null
+++ b/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.jsonl
@@ -0,0 +1,323 @@
+{"title": "效果如何优化", "file": "2023-04-04.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/14", "detail": "如图所示,将该项目的README.md和该项目结合后,回答效果并不理想,请问可以从哪些方面进行优化", "id": 0}
+{"title": "怎么让模型严格根据检索的数据进行回答,减少胡说八道的回答呢", "file": "2023-04-04.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/15", "detail": "举个例子:", "id": 1}
+{"title": "When I try to run the `python knowledge_based_chatglm.py`, I got this error in macOS(M1 Max, OS 13.2)", "file": "2023-04-07.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/32", "detail": "```python", "id": 2}
+{"title": "萌新求教大佬怎么改成AMD显卡或者CPU?", "file": "2023-04-10.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/48", "detail": "把.cuda()去掉就行", "id": 3}
+{"title": "输出answer的时间很长,是否可以把文本向量化的部分提前做好存储起来?", "file": "2023-04-10.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/50", "detail": "GPU:4090 24G显存", "id": 4}
+{"title": "报错Use `repo_type` argument if needed.", "file": "2023-04-11.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/57", "detail": "Traceback (most recent call last):", "id": 5}
+{"title": "无法打开gradio的页面", "file": "2023-04-11.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/58", "detail": "$ python webui.py", "id": 6}
+{"title": "支持word,那word里面的图片正常显示吗?", "file": "2023-04-12.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/60", "detail": "如题,刚刚从隔壁转过来的,想先了解下", "id": 7}
+{"title": "detectron2 is not installed. Cannot use the hi_res partitioning strategy. Falling back to partitioning with the fast strategy.", "file": "2023-04-12.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/63", "detail": "能够正常的跑起来,在加载content文件夹中的文件时,每加载一个文件都会提示:", "id": 8}
+{"title": "cpu上运行webui,step3 asking时报错", "file": "2023-04-12.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/66", "detail": "web运行,文件加载都正常,asking时报错", "id": 9}
+{"title": "建议弄一个插件系统", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/67", "detail": "如题弄成stable-diffusion-webui那种能装插件,再开一个存储库给使用者或插件开发,存储或下载插件。", "id": 10}
+{"title": "请教加载模型出错!?", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/75", "detail": "AttributeError: module 'transformers_modules.chatglm-6b.configuration_chatglm' has no attribute 'ChatGLMConfig 怎么解决呀", "id": 11}
+{"title": "从本地知识检索内容的时候,是否可以设置相似度阈值,小于这个阈值的内容不返回,即使会小于设置的VECTOR_SEARCH_TOP_K参数呢?谢谢大佬", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/76", "detail": "比如 问一些 你好/你是谁 等一些跟本地知识库无关的问题", "id": 12}
+{"title": "如何改成多卡推理?", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/77", "detail": "+1", "id": 13}
+{"title": "能否弄个懒人包,可以一键体验?", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/78", "detail": "能否弄个懒人包,可以一键体验?", "id": 14}
+{"title": "连续问问题会导致崩溃", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/79", "detail": "看上去不是爆内存的问题,连续问问题后,会出现如下报错", "id": 15}
+{"title": "AttributeError: 'NoneType' object has no attribute 'as_retriever'", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/86", "detail": "环境:windows 11, anaconda/python 3.8", "id": 16}
+{"title": "FileNotFoundError: Could not find module 'nvcuda.dll' (or one of its dependencies). Try using the full path with constructor syntax.", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/87", "detail": "请检查一下cuda或cudnn是否存在安装问题", "id": 17}
+{"title": "加载txt文件失败?", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/89", "detail": "![JppHrGOWFa](https://user-images.githubusercontent.com/109277248/232009383-bf7c46d1-a01e-4e0a-9de6-5b5ed3e36158.jpg)", "id": 18}
+{"title": "NameError: name 'chatglm' is not defined", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/90", "detail": "This share link expires in 72 hours. For free permanent hosting and GPU upgrades (NEW!), check out Spaces: https://huggingface.co/spaces", "id": 19}
+{"title": "打不开地址?", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/91", "detail": "报错数据如下:", "id": 20}
+{"title": "加载md文件出错", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/98", "detail": "运行 webui.py后能访问页面,上传一个md文件后,日志中有错误。等待后能加载完成,提示可以提问了,但提问没反应,日志中有错误。 具体日志如下。", "id": 21}
+{"title": "建议增加获取在线知识的能力", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/101", "detail": "建议增加获取在线知识的能力", "id": 22}
+{"title": "txt 未能成功加载", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/103", "detail": "hinese. Creating a new one with MEAN pooling.", "id": 23}
+{"title": "pdf加载失败", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/105", "detail": "e:\\a.txt加载成功了,e:\\a.pdf加载就失败,pdf文件里面前面几页是图片,后面都是文字,加载失败没有报更多错误,请问该怎么排查?", "id": 24}
+{"title": "一直停在文本加载处", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/108", "detail": "一直停在文本加载处", "id": 25}
+{"title": " File \"/root/.cache/huggingface/modules/transformers_modules/chatglm-6b/modeling_chatglm.py\", line 440, in forward new_tensor_shape = mixed_raw_layer.size()[:-1] + ( TypeError: torch.Size() takes an iterable of 'int' (item 2 is 'float')", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/113", "detail": "按照最新的代码,发现", "id": 26}
+{"title": "后续会提供前后端分离的功能吗?", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/114", "detail": "类似这种https://github.com/lm-sys/FastChat/tree/main/fastchat/serve", "id": 27}
+{"title": "安装依赖报错", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/115", "detail": "(test) C:\\Users\\linh\\Desktop\\langchain-ChatGLM-master>pip install -r requirements.txt", "id": 28}
+{"title": "问特定问题会出现爆显存", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/117", "detail": "正常提问没问题。", "id": 29}
+{"title": "Expecting value: line 1 column 1 (char 0)", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/118", "detail": "运行后 第一步加载配置一直报错:", "id": 30}
+{"title": "embedding https://huggingface.co/GanymedeNil/text2vec-large-chinese/tree/main是免费的,效果比对openai的如何?", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/119", "detail": "-------------------------------------------------------------------------------", "id": 31}
+{"title": "这是什么错误,在Colab上运行的。", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/120", "detail": "libcuda.so.1: cannot open shared object file: No such file or directory", "id": 32}
+{"title": "只想用自己的lora微调后的模型进行对话,不想加载任何本地文档,该如何调整?", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/121", "detail": "能出一个单独的教程吗", "id": 33}
+{"title": "租的gpu,Running on local URL: http://0.0.0.0:7860 To create a public link, set `share=True` in `launch()`. 浏览器上访问不了???", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/122", "detail": "(chatglm20230401) root@autodl-container-e82d11963c-10ece0d7:~/autodl-tmp/chatglm/langchain-ChatGLM-20230418# python3.9 webui.py", "id": 34}
+{"title": "本地部署中的报错请教", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/124", "detail": "您好,在本地运行langchain-ChatGLM过程中,环境及依赖的包都已经满足条件,但是运行webui.py,报错如下(运行cli_demo.py报错类似),请问是哪里出了错呢?盼望您的回复,谢谢!", "id": 35}
+{"title": "报错。The dtype of attention mask (torch.int64) is not bool", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/131", "detail": "The dtype of attention mask (torch.int64) is not bool", "id": 36}
+{"title": "[求助] pip install -r requirements.txt 的时候出现以下报错。。。有大佬帮忙看看怎么搞么,下的release里面的包", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/134", "detail": "$ pip install -r requirements.txt", "id": 37}
+{"title": "如何提升根据问题搜索到对应知识的准确率", "file": "2023-04-19.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/136", "detail": "外链知识库最大的问题在于问题是短文本,知识是中长文本。如何根据问题精准的搜索到对应的知识是个最大的问题。这类本地化项目不像百度,由无数的网页,基本上每个问题都可以找到对应的页面。", "id": 38}
+{"title": "是否可以增加向量召回的阈值设定,有些召回内容相关性太低,导致模型胡言乱语", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/140", "detail": "如题", "id": 39}
+{"title": "输入长度问题", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/141", "detail": "感谢作者支持ptuning微调模型。", "id": 40}
+{"title": "已有部署好的chatGLM-6b,如何通过接口接入?", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/144", "detail": "已有部署好的chatGLM-6b,如何通过接口接入,而不是重新加载一个模型;", "id": 41}
+{"title": "执行web_demo.py后,显示Killed,就退出了,是不是配置不足呢?", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/146", "detail": "![图片](https://user-images.githubusercontent.com/26102866/233256425-c7aab999-11d7-4de9-867b-23ef18d519e4.png)", "id": 42}
+{"title": "执行python cli_demo1.py", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/147", "detail": "Traceback (most recent call last):", "id": 43}
+{"title": "报错:ImportError: cannot import name 'GENERATION_CONFIG_NAME' from 'transformers.utils'", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/149", "detail": "(mychatGLM) PS D:\\Users\\admin3\\zrh\\langchain-ChatGLM> python cli_demo.py", "id": 44}
+{"title": "上传文件并加载知识库时,会不停地出现临时文件", "file": "2023-04-21.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/153", "detail": "环境:ubuntu 18.04", "id": 45}
+{"title": "向知识库中添加文件后点击”上传文件并加载知识库“后Segmentation fault报错。", "file": "2023-04-23.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/161", "detail": "运行服务后的提示如下:", "id": 46}
+{"title": "langchain-serve 集成", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/162", "detail": "Hey 我是来自 [langchain-serve](https://github.com/jina-ai/langchain-serve) 的dev!", "id": 47}
+{"title": "大佬们,wsl的ubuntu怎么配置用cuda加速,装了运行后发现是cpu在跑", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/164", "detail": "大佬们,wsl的ubuntu怎么配置用cuda加速,装了运行后发现是cpu在跑", "id": 48}
+{"title": "在github codespaces docker运行出错", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/165", "detail": "docker run -d --restart=always --name chatglm -p 7860:7860 -v /www/wwwroot/code/langchain-ChatGLM:/chatGLM chatglm", "id": 49}
+{"title": "有计划接入Moss模型嘛", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/166", "detail": "后续会开展测试,目前主要在优化langchain部分效果,如果有兴趣也欢迎提PR", "id": 50}
+{"title": "怎么实现 API 部署?", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/168", "detail": "利用 fastapi 实现 API 部署方式,具体怎么实现,有方法说明吗?", "id": 51}
+{"title": " 'NoneType' object has no attribute 'message_types_by_name'报错", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/169", "detail": "_HISTOGRAMPROTO = DESCRIPTOR.message_types_by_name['HistogramProto']", "id": 52}
+{"title": "能否指定自己训练的text2vector模型?", "file": "2023-04-25.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/172", "detail": "请问大佬:", "id": 53}
+{"title": "关于项目支持的模型以及quantization_bit潜在的影响的问题", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/176", "detail": "作者您好~", "id": 54}
+{"title": "运行python3.9 api.py WARNING: You must pass the application as an import string to enable 'reload' or 'workers'.", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/179", "detail": "api.py文件最下面改成这样试试:", "id": 55}
+{"title": "ValidationError: 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted (type=value_error.extra)", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/180", "detail": "ValidationError: 1 validation error for HuggingFaceEmbeddings", "id": 56}
+{"title": "如果没有检索到相关性比较高的,回答“我不知道”", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/181", "detail": "如果通过设计system_template,让模型在搜索到的文档都不太相关的情况下回答“我不知道”", "id": 57}
+{"title": "请问如果不能联网,6B之类的文件从本地上传需要放到哪里", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/182", "detail": "感谢大佬的项目,很有启发~", "id": 58}
+{"title": "知识库问答--输入新的知识库名称是中文的话,会报error", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/184", "detail": "知识库问答--输入新的知识库名称是中文的话,会报error,选择要加载的知识库那里也不显示之前添加的知识库", "id": 59}
+{"title": "现在能通过问题匹配的相似度值,来直接返回文档中的文段,而不经过模型吗?因为有些答案在文档中,模型自己回答,不能回答文档中的答案", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/186", "detail": "现在能通过问题匹配的相似度值,来直接返回文档中的文段,而不经过模型吗?因为有些答案在文档中,模型自己回答,不能回答文档中的答案。也就是说,提供向量检索回答+模型回答相结合的策略。如果相似度值高于一定数值,直接返回文档中的文本,没有高于就返回模型的回答或者不知道", "id": 60}
+{"title": "TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/188", "detail": "Mac 运行 python3 ./webui.py 报 TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation", "id": 61}
+{"title": "Not Enough Memory", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/190", "detail": "运行命令行程序python cli_demo.py, 已经成功加载pdf文件, 报“DefaultCPUAllocator: not enough memory: you tried to allocate 458288380900 bytes”错误,请问哪里可以配置default memory", "id": 62}
+{"title": "参与开发问题", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/191", "detail": "1.是否需要进专门的开发群", "id": 63}
+{"title": "对话框中代码片段格式需改进", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/192", "detail": "最好能改进下输出代码片段的格式,目前输出的格式还不友好。", "id": 64}
+{"title": "请问未来有可能支持belle吗", "file": "2023-04-28.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/195", "detail": "如题,谢谢大佬", "id": 65}
+{"title": "TypeError: cannot unpack non-iterable NoneType object", "file": "2023-04-28.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/200", "detail": "When i tried to change the knowledge vector store through `init_knowledge_vector_store`, the error `TypeError: cannot unpack non-iterable NoneType object` came out.", "id": 66}
+{"title": "生成结果", "file": "2023-04-28.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/202", "detail": "你好,想问一下langchain+chatglm-6B,找到相似匹配的prompt,是直接返回prompt对应的答案信息,还是chatglm-6B在此基础上自己优化答案?", "id": 67}
+{"title": "在win、ubuntu下都出现这个错误:attributeerror: 't5forconditionalgeneration' object has no attribute 'stream_chat'", "file": "2023-04-29.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/207", "detail": "在win、ubuntu。下载完模型后,没办法修改代码以执行本地模型,每次都要重新输入路径; LLM 模型、Embedding 模型支持也都在官网下的,在其他项目(wenda)下可以使用", "id": 68}
+{"title": "[FEATURE] knowledge_based_chatglm.py: renamed or missing?", "file": "2023-04-30.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/210", "detail": "Not found. Was it renamed? Or, is it missing? How can I get it?", "id": 69}
+{"title": "sudo apt-get install -y nvidia-container-toolkit-base执行报错", "file": "2023-05-01.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/211", "detail": "**问题描述 / Problem Description**", "id": 70}
+{"title": "效果不佳几乎答不上来", "file": "2023-05-01.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/212", "detail": "提供了50条问答的docx文件", "id": 71}
+{"title": "有没有可能新增一个基于chatglm api调用的方式构建langchain", "file": "2023-05-02.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/218", "detail": "我有两台8G GPU/40G内存的服务器,一个台做成了chatglm的api ;想基于另外一台服务器部署langchain;网上好像没有类似的代码。", "id": 72}
+{"title": "电脑是intel的集成显卡; 运行时告知我找不到nvcuda.dll,模型无法运行", "file": "2023-05-02.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/219", "detail": "您好,我的电脑是intel的集成显卡,不过CPU是i5-11400 @ 2.60GHz ,内存64G;", "id": 73}
+{"title": "根据langchain官方的文档和使用模式,是否可以改Faiss为Elasticsearch?会需要做哪些额外调整?求解", "file": "2023-05-03.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/221", "detail": "本人新手小白,由于业务模式的原因(有一些自己的场景和优化),希望利用Elasticsearch做这个体系内部的检索机制,不知道是否可以替换,同时,还会涉及到哪些地方的改动?或者说可能会有哪些其他影响,希望作者和大佬们不吝赐教!", "id": 74}
+{"title": "请问未来有可能支持t5吗", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/224", "detail": "请问可能支持基於t5的模型吗?", "id": 75}
+{"title": "[BUG] 内存溢出 / torch.cuda.OutOfMemoryError:", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/229", "detail": "**问题描述 / Problem Description**", "id": 76}
+{"title": "报错 No module named 'chatglm_llm'", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/230", "detail": "明明已经安装了包,却在python里吊不出来", "id": 77}
+{"title": "能出一个api部署的描述文档吗", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/233", "detail": "**功能描述 / Feature Description**", "id": 78}
+{"title": "使用docs/API.md 出错", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/234", "detail": "使用API.md文档2种方法,出错", "id": 79}
+{"title": "加载pdf文档报错?", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/238", "detail": "ew one with MEAN pooling.", "id": 80}
+{"title": "上传的本地知识文件后再次上传不能显示,只显示成功了一个,别的上传成功后再次刷新就没了", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/239", "detail": "您好,项目有很大启发,感谢~", "id": 81}
+{"title": "创建了新的虚拟环境,安装了相关包,并且自动下载了相关的模型,但是仍旧出现:OSError: Unable to load weights from pytorch checkpoint file for", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/240", "detail": "![78ac8e663fdc312d0e9d78da95925c4](https://user-images.githubusercontent.com/34124260/236378728-9ea4424f-0f7f-4013-9d33-820b723de321.png)", "id": 82}
+{"title": "[BUG] 数据加载不进来", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/243", "detail": "使用的.txt格式,utf-8编码,报以下错误", "id": 83}
+{"title": "不能读取pdf", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/244", "detail": "请问是webui还是cli_demo", "id": 84}
+{"title": "本地txt文件有500M,加载的时候很慢,如何提高速度?", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/251", "detail": "![yayRzxSYHP](https://user-images.githubusercontent.com/109277248/236592902-f5ab338d-c1e9-43dc-ae16-9df2cd3c1378.jpg)", "id": 85}
+{"title": "[BUG] gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/253", "detail": "gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库", "id": 86}
+{"title": "[FEATURE] 可以支持 OpenAI 的模型嘛?比如 GPT-3、GPT-3.5、GPT-4;embedding 增加 text-embedding-ada-002", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/254", "detail": "**功能描述 / Feature Description**", "id": 87}
+{"title": "[FEATURE] 能否增加对于milvus向量数据库的支持 / Concise description of the feature", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/256", "detail": "**功能描述 / Feature Description**", "id": 88}
+{"title": "CPU和GPU上跑,除了速度有区别,准确率效果回答上有区别吗?", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/259", "detail": "理论上没有区别", "id": 89}
+{"title": "m1,请问在生成回答时怎么看是否使用了mps or cpu?", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/260", "detail": "m1,请问在生成回答时怎么看是否使用了mps or cpu?", "id": 90}
+{"title": "知识库一刷新就没了", "file": "2023-05-07.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/263", "detail": "知识库上传后刷新就没了", "id": 91}
+{"title": "本地部署报没有模型", "file": "2023-05-07.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/267", "detail": "建议在下载llm和embedding模型至本地后在configs/model_config中写入模型本地存储路径后再运行", "id": 92}
+{"title": "[BUG] python3: can't open file 'webui.py': [Errno 2] No such file or directory", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/269", "detail": "**问题描述 / Problem Description**", "id": 93}
+{"title": "模块缺失提示", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/271", "detail": "因为已有自己使用的docker环境,直接启动webui.py,提示", "id": 94}
+{"title": "运行api.py后,执行curl -X POST \"http://127.0.0.1:7861\" 报错?", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/272", "detail": "执行curl -X POST \"http://127.0.0.1:7861\" \\ -H 'Content-Type: application/json' \\ -d '{\"prompt\": \"你好\", \"history\": []}',报错怎么解决", "id": 95}
+{"title": "[BUG] colab安装requirements提示protobuf版本问题?", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/273", "detail": "pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.", "id": 96}
+{"title": "请问项目里面向量相似度使用了什么方法计算呀?", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/275", "detail": "基本按照langchain里的FAISS.similarity_search_with_score_by_vector实现", "id": 97}
+{"title": "[BUG] 安装detectron2后,pdf无法加载", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/276", "detail": "**问题描述 / Problem Description**", "id": 98}
+{"title": "[BUG] 使用ChatYuan-V2模型无法流式输出,会报错", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/277", "detail": "一方面好像是ChatYuan本身不支持stream_chat,有人在clueai那边提了issue他们说还没开发,所以估计这个attribute调不起来;但是另一方面看报错好像是T5模型本身就不是decoder-only模型,所以不能流式输出吧(个人理解)", "id": 99}
+{"title": "[BUG] 无法加载text2vec模型", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/278", "detail": "**问题描述 / Problem Description**", "id": 100}
+{"title": "请问能否增加网络搜索功能", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/281", "detail": "请问能否增加网络搜索功能", "id": 101}
+{"title": "[FEATURE] 结构化数据sql、excel、csv啥时会支持呐。", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/283", "detail": "**功能描述 / Feature Description**", "id": 102}
+{"title": "TypeError: ChatGLM._call() got an unexpected keyword argument 'stop'", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/284", "detail": "No sentence-transformers model found with name D:\\DevProject\\langchain-ChatGLM\\GanymedeNil\\text2vec-large-chinese. Creating a new one with MEAN pooling.", "id": 103}
+{"title": "关于api.py的一些bug和设计逻辑问题?", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/285", "detail": "首先冒昧的问一下,这个api.py,开发者大佬们是在自己电脑上测试后确实没问题吗?", "id": 104}
+{"title": "有没有租用的算力平台上,运行api.py后,浏览器http://localhost:7861/报错", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/287", "detail": "是不是租用的gpu平台上都会出现这个问题???", "id": 105}
+{"title": "请问一下项目中有用到文档段落切割方法吗?", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/288", "detail": "text_load中的文档切割方法用上了吗?在代码中看好像没有用到?", "id": 106}
+{"title": "报错 raise ValueError(f\"Knowledge base {knowledge_base_id} not found\") ValueError: Knowledge base ./vector_store not found", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/289", "detail": "File \"/root/autodl-tmp/chatglm/langchain-ChatGLM-master/api.py\", line 183, in chat", "id": 107}
+{"title": "能接入vicuna模型吗", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/290", "detail": "目前本地已经有了vicuna模型能直接接入吗?", "id": 108}
+{"title": "[BUG] 提问公式相关问题大概率爆显存", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/291", "detail": "**问题描述 / Problem Description**", "id": 109}
+{"title": "安装pycocotools失败,找了好多方法都不能解决。", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/292", "detail": "**问题描述 / Problem Description**", "id": 110}
+{"title": "使用requirements安装,PyTorch安装的是CPU版本", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/294", "detail": "如题目,使用requirements安装,PyTorch安装的是CPU版本,运行程序的时候,也是使用CPU在工作。", "id": 111}
+{"title": "能不能给一个毛坯服务器的部署教程", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/298", "detail": "“开发部署”你当成服务器的部署教程用就行了。", "id": 112}
+{"title": " Error(s) in loading state_dict for ChatGLMForConditionalGeneration:", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/299", "detail": "运行中出现的问题,7860的端口页面显示不出来,求助。", "id": 113}
+{"title": "ChatYuan-large-v2模型加载失败", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/300", "detail": "**实际结果 / Actual Result**", "id": 114}
+{"title": "新增摘要功能", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/303", "detail": "你好,后续会考虑新增对长文本信息进行推理和语音理解功能吗?比如生成摘要", "id": 115}
+{"title": "[BUG] pip install -r requirements.txt 出错", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/304", "detail": "pip install langchain -i https://pypi.org/simple", "id": 116}
+{"title": "[BUG] 上传知识库文件报错", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/305", "detail": "![19621e29eaa547d01213bee53d81e6a](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/7f6ceb46-e494-4b0e-939c-23b585a6d9d8)", "id": 117}
+{"title": "[BUG] AssertionError: Component with id 41 not a valid input component.", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/306", "detail": "**问题描述 / Problem Description**", "id": 118}
+{"title": "[BUG] CUDA out of memory with container deployment", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/310", "detail": "**问题描述 / Problem Description**", "id": 119}
+{"title": "[FEATURE] 增加微调训练功能", "file": "2023-05-11.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/311", "detail": "**功能描述 / Feature Description**", "id": 120}
+{"title": "如何使用多卡部署,多个gpu", "file": "2023-05-11.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/315", "detail": "机器上有多个gpu,如何全使用了", "id": 121}
+{"title": "请问这个知识库问答,和chatglm的关系是什么", "file": "2023-05-11.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/319", "detail": "这个知识库问答,哪部分关联到了chatglm,是不是没有这个chatglm,知识库问答也可单单拎出来", "id": 122}
+{"title": "[BUG] 运行的时候报错ImportError: libcudnn.so.8: cannot open shared object file: No such file or directory", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/324", "detail": "**问题描述 / Problem Description**raceback (most recent call last):", "id": 123}
+{"title": "webui启动成功,但有报错", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/325", "detail": "**问题描述 / Problem Description**", "id": 124}
+{"title": "切换MOSS的时候报错", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/327", "detail": "danshi但是发布的源码中,", "id": 125}
+{"title": "vicuna模型是否能接入?", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/328", "detail": "您好!关于MOSS模型和vicuna模型,都是AutoModelForCausalLM来加载模型的,但是稍作更改(模型路径这些)会报这个错误。这个错误的造成是什么", "id": 126}
+{"title": "你好,请问一下在阿里云CPU服务器上跑可以吗?可以的话比较理想的cpu配置是什么?", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/330", "detail": "你好,请问一下在阿里云CPU服务器上跑可以吗?可以的话比较理想的cpu配置是什么?", "id": 127}
+{"title": "你好,请问8核32g的CPU可以跑多轮对话吗?", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/331", "detail": "什么样的cpu配置比较好呢?我目前想部署CPU下的多轮对话?", "id": 128}
+{"title": "[BUG] 聊天内容输入超过10000个字符系统出现错误", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/332", "detail": "聊天内容输入超过10000个字符系统出现错误,如下图所示:", "id": 129}
+{"title": "能增加API的多用户访问接口部署吗?", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/333", "detail": "默认部署程序仅支持单用户访问,多用户则需要排队访问。测试过相关的几个Github多用户工程,但是其中一些仍然不满足要求。本节将系统介绍如何实现多用户同时访问ChatGLM的部署接口,包括http、websocket(流式输出,stream)和web页面等方式,主要目录如下所示。", "id": 130}
+{"title": "多卡部署", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/334", "detail": "用单机多卡或多机多卡,fastapi部署模型,怎样提高并发", "id": 131}
+{"title": "WEBUI能否指定知识库目录?", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/335", "detail": "**功能描述 / Feature Description**", "id": 132}
+{"title": "[BUG] Cannot read properties of undefined (reading 'error')", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/336", "detail": "**问题描述 / Problem Description**", "id": 133}
+{"title": "[BUG] 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted.", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/337", "detail": "模型加载到 100% 后出现问题:", "id": 134}
+{"title": "上传知识库需要重启能不能修复一下", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/338", "detail": "挺严重的这个问题", "id": 135}
+{"title": "[BUG] 4块v100卡爆显存,在LLM会话模式也一样", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/339", "detail": "**问题描述 / Problem Description**", "id": 136}
+{"title": "针对上传的文件配置不同的TextSpliter", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/341", "detail": "1. 目前的ChineseTextSpliter切分对英文尤其是代码文件不友好,而且限制固定长度;导致对话结果不如人意", "id": 137}
+{"title": "[FEATURE] 未来可增加Bloom系列模型吗?根据甲骨易的测试,这系列中文评测效果不错", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/346", "detail": "**功能描述 / Feature Description**", "id": 138}
+{"title": "[BUG] v0.1.12打包镜像后启动webui.py失败 / Concise description of the issue", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/347", "detail": "**问题描述 / Problem Description**", "id": 139}
+{"title": "切换MOSS模型时报错", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/349", "detail": "昨天问了下,说是transformers版本不对,需要4.30.0,发现没有这个版本,今天更新到4.29.1,依旧报错,错误如下", "id": 140}
+{"title": "[BUG] pdf文档加载失败", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/350", "detail": "**问题描述 / Problem Description**", "id": 141}
+{"title": "建议可以在后期增强一波注释,这样也有助于更多人跟进提PR", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/351", "detail": "知道作者和团队在疯狂更新审查代码,只是建议后续稳定后可以把核心代码进行一些注释的补充,从而能帮助更多人了解各个模块作者的思路从而提出更好的优化。", "id": 142}
+{"title": "[FEATURE] MOSS 量化版支援", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/353", "detail": "**功能描述 / Feature Description**", "id": 143}
+{"title": "[BUG] moss模型无法加载", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/356", "detail": "**问题描述 / Problem Description**", "id": 144}
+{"title": "[BUG] load_doc_qa.py 中的 load_file 函数有bug", "file": "2023-05-14.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/358", "detail": "原函数为:", "id": 145}
+{"title": "[FEATURE] API模式,知识库加载优化", "file": "2023-05-14.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/359", "detail": "如题,当前版本,每次调用本地知识库接口,都将加载一次知识库,是否有更好的方式?", "id": 146}
+{"title": "运行Python api.py脚本后端部署后,怎么使用curl命令调用?", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/361", "detail": "也就是说,我现在想做个对话机器人,想和公司的前后端联调?怎么与前后端相互调用呢?可私信,有偿解答!!!", "id": 147}
+{"title": "上传知识库需要重启能不能修复一下", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/363", "detail": "上传知识库需要重启能不能修复一下", "id": 148}
+{"title": "[BUG] pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/364", "detail": "我的python是3.8.5的", "id": 149}
+{"title": "pip install gradio 报错", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/367", "detail": "大佬帮我一下", "id": 150}
+{"title": "[BUG] pip install gradio 一直卡不动", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/369", "detail": "![aba82742dd9d4d242181662eb5027a7](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/cd9600d9-f6e7-46b7-b1be-30ed8b99f76b)", "id": 151}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/370", "detail": "初次加载本地知识库成功,但提问后,就无法重写加载本地知识库", "id": 152}
+{"title": "[FEATURE] 简洁阐述功能 / Concise description of the feature", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/371", "detail": "**功能描述 / Feature Description**", "id": 153}
+{"title": "在windows上,模型文件默认会安装到哪", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/372", "detail": "-------------------------------------------------------------------------------", "id": 154}
+{"title": "[FEATURE] 兼顾对话管理", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/374", "detail": "如何在知识库检索的情况下,兼顾对话管理?", "id": 155}
+{"title": "llm device: cpu embedding device: cpu", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/376", "detail": "**问题描述 / Problem Description**", "id": 156}
+{"title": "[FEATURE] 简洁阐述功能 /文本文件的知识点之间使用什么分隔符可以分割?", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/377", "detail": "**功能描述 / Feature Description**", "id": 157}
+{"title": "[BUG] 上传文件失败:PermissionError: [WinError 32] 另一个程序正在使用此文件,进程无法访问。", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/379", "detail": "**问题描述 / Problem Description**", "id": 158}
+{"title": "[BUG] 执行python api.py 报错", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/383", "detail": "错误信息", "id": 159}
+{"title": "model_kwargs extra fields not permitted (type=value_error.extra)", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/384", "detail": "大家好,请问这个有遇到的么,?", "id": 160}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/385", "detail": "执行的时候出现了ls1 = [ls[0]]", "id": 161}
+{"title": "[FEATURE] 性能优化", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/388", "detail": "**功能描述 / Feature Description**", "id": 162}
+{"title": "[BUG] Moss模型问答,RuntimeError: probability tensor contains either inf, nan or element < 0", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/390", "detail": "**问题描述 / Problem Description**", "id": 163}
+{"title": "有没有人知道v100GPU的32G显存,会报错吗?支持V100GPU吗?", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/392", "detail": "**问题描述 / Problem Description**", "id": 164}
+{"title": "针对于编码问题比如'gbk' codec can't encode character '\\xab' in position 14: illegal multibyte sequence粗浅的解决方法", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/397", "detail": "**功能描述 / Feature Description**", "id": 165}
+{"title": "Could not import sentence_transformers python package. Please install it with `pip install sentence_transformers`.", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/400", "detail": "**问题描述 / Problem Description**", "id": 166}
+{"title": "支持模型问答与检索问答", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/401", "detail": "不同的query,根据意图不一致,回答也应该不一样。", "id": 167}
+{"title": "文本分割的时候,能不能按照txt文件的每行进行分割,也就是按照换行符号\\n进行分割???", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/403", "detail": "下面的代码应该怎么修改?", "id": 168}
+{"title": "local_doc_qa/local_doc_chat 接口响应是串行", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/405", "detail": "**问题描述 / Problem Description**", "id": 169}
+{"title": "为什么找到出处了,但是还是无法回答该问题?", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/406", "detail": "![图片](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/3349611/1fc81d61-2409-4330-9065-fdda1a27c86a)", "id": 170}
+{"title": "请问下:知识库测试中的:添加单条内容,如果换成文本导入是是怎样的格式?我发现添加单条内容测试效果很好.", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/412", "detail": "我发现在知识库测试中`添加单条内容`,并且勾选`禁止内容分句入库`,即使 `不开启上下文关联`的测试效果都非常好.", "id": 171}
+{"title": "[BUG] 无法配置知识库", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/413", "detail": "**问题描述 / Problem Description**", "id": 172}
+{"title": "[BUG] 部署在阿里PAI平台的EAS上访问页面是白屏", "file": "2023-05-19.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/414", "detail": "**问题描述 / Problem Description**", "id": 173}
+{"title": "API部署后调用/local_doc_qa/local_doc_chat 返回Knowledge base samples not found", "file": "2023-05-19.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/416", "detail": "入参", "id": 174}
+{"title": "[FEATURE] 上传word另存为的txt文件报 'ascii' codec can't decode byte 0xb9 in position 6: ordinal not in range(128)", "file": "2023-05-20.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/421", "detail": "上传word另存为的txt文件报", "id": 175}
+{"title": "创建保存的知识库刷新后没有出来,这个知识库是永久保存的吗?可以连外部的 向量知识库吗?", "file": "2023-05-21.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/422", "detail": "创建保存的知识库刷新后没有出来,这个知识库是永久保存的吗?可以连外部的 向量知识库吗?", "id": 176}
+{"title": "[BUG] 用colab运行,无法加载模型,报错:'NoneType' object has no attribute 'message_types_by_name'", "file": "2023-05-21.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/423", "detail": "**问题描述 / Problem Description**", "id": 177}
+{"title": "请问是否需要用到向量数据库?以及什么时候需要用到向量数据库?", "file": "2023-05-21.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/424", "detail": "目前用的是 text2vec , 请问是否需要用到向量数据库?以及什么时候需要用到向量数据库?", "id": 178}
+{"title": "huggingface模型引用问题", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/427", "detail": "它最近似乎变成了一个Error?", "id": 179}
+{"title": "你好,加载本地txt文件出现这个killed错误,TXT文件有100M左右大小。原因是?谢谢。", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/429", "detail": "", "id": 180}
+{"title": "想请问一下,关于对本地知识的管理是如何管理?例如:通过http API接口添加数据 或者 删除某条数据", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/430", "detail": "例如:通过http API接口添加、删除、修改 某条数据。", "id": 181}
+{"title": "[FEATURE] 双栏pdf识别问题", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/432", "detail": "试了一下模型,感觉对单栏pdf识别的准确性较高,但是由于使用的基本是ocr的技术,对一些双栏pdf论文识别出来有很多问题,请问有什么办法改善吗?", "id": 182}
+{"title": "部署启动小问题,小弟初学求大佬解答", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/433", "detail": "1.python loader/image_loader.py时,提示ModuleNotFoundError: No module named 'configs',但是跑python webui.py还是还能跑", "id": 183}
+{"title": "能否支持检测到目录下文档有增加而去增量加载文档,不影响前台对话,其实就是支持读写分离。如果能支持查询哪些文档向量化了,删除过时文档等就更好了,谢谢。", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/434", "detail": "**功能描述 / Feature Description**", "id": 184}
+{"title": "[BUG] 简洁阐述问题 / windows 下cuda错误,请用https://github.com/Keith-Hon/bitsandbytes-windows.git", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/435", "detail": "pip install git+https://github.com/Keith-Hon/bitsandbytes-windows.git", "id": 185}
+{"title": "[BUG] from commit 33bbb47, Required library version not found: libbitsandbytes_cuda121_nocublaslt.so. Maybe you need to compile it from source?", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/438", "detail": "**问题描述 / Problem Description**", "id": 186}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue上传60m的txt文件报错,显示超时,请问这个能上传的文件大小有限制吗", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/439", "detail": "ERROR 2023-05-23 11:13:09,627-1d: Timeout reached while detecting encoding for ./docs/GLM模型格式数据.txt", "id": 187}
+{"title": "[BUG] TypeError: issubclass() arg 1 must be a class", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/440", "detail": "**问题描述**", "id": 188}
+{"title": "执行python3 webui.py后,一直提示”模型未成功加载,请到页面左上角\"模型配置\"选项卡中重新选择后点击\"加载模型\"按钮“", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/441", "detail": "**问题描述 / Problem Description**", "id": 189}
+{"title": "是否能提供网页文档得导入支持", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/444", "detail": "现在很多都是在线文档作为协作得工具,所以通过URL导入在线文档需求更大", "id": 190}
+{"title": "[BUG] history 索引问题", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/445", "detail": "在比较对话框的history和模型chat function 中的history时, 发现并不匹配,在传入 llm._call 时,history用的索引是不是有点问题,导致上一轮对话的内容并不输入给模型。", "id": 191}
+{"title": "[BUG] moss_llm没有实现", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/447", "detail": "有些方法没支持,如history_len", "id": 192}
+{"title": "请问langchain-ChatGLM如何删除一条本地知识库的数据?", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/448", "detail": "例如:用户刚刚提交了一条错误的数据到本地知识库中了,现在如何在本地知识库从找到,并且对此删除。", "id": 193}
+{"title": "[BUG] 简洁阐述问题 / UnboundLocalError: local variable 'resp' referenced before assignment", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/450", "detail": "在最新一版的代码中, 运行api.py 出现了以上错误(UnboundLocalError: local variable 'resp' referenced before assignment), 通过debug的方式观察到local_doc_qa.llm.generatorAnswer(prompt=question, history=history,streaming=True)可能不返回任何值。", "id": 194}
+{"title": "请问有没有 PROMPT_TEMPLATE 能让模型不回答敏感问题", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/452", "detail": "## PROMPT_TEMPLATE问题", "id": 195}
+{"title": "[BUG] 测试环境 Python 版本有误", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/456", "detail": "**问题描述 / Problem Description**", "id": 196}
+{"title": "[BUG] webui 部署后样式不正确", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/458", "detail": "**问题描述 / Problem Description**", "id": 197}
+{"title": "配置默认LLM模型的问题", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/459", "detail": "**问题描述 / Problem Description**", "id": 198}
+{"title": "[FEATURE]是时候更新一下autoDL的镜像了", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/460", "detail": "如题,跑了下autoDL的镜像,发现是4.27号的,git pull新版本的代码功能+老的依赖环境,各种奇奇怪怪的问题。", "id": 199}
+{"title": "[BUG] tag:0.1.13 以cpu模式下,想使用本地模型无法跑起来,各种路径参数问题", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/462", "detail": "-------------------------------------------------------------------------------", "id": 200}
+{"title": "[BUG] 有没有同学遇到过这个错!!!加载本地txt文件出现这个killed错误,TXT文件有100M左右大小。", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/463", "detail": "运行cli_demo.py。是本地的txt文件太大了吗?100M左右。", "id": 201}
+{"title": "API版本能否提供WEBSOCKET的流式接口", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/464", "detail": "webui 版本中,采用了WS的流式输出,整体感知反应很快", "id": 202}
+{"title": "[BUG] 安装bug记录", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/465", "detail": "按照[install文档](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/INSTALL.md)安装的,", "id": 203}
+{"title": "VUE的pnmp i执行失败的修复-用npm i命令即可", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/466", "detail": "感谢作者!非常棒的应用,用的很开心。", "id": 204}
+{"title": "请教个问题,有没有人知道cuda11.4是否支持???", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/467", "detail": "请教个问题,有没有人知道cuda11.4是否支持???", "id": 205}
+{"title": "请问有实现多轮问答中基于问题的搜索上下文关联么", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/468", "detail": "在基于知识库的多轮问答中,第一个问题讲述了一个主题,后续的问题描述没有包含这个主题的关键词,但又存在上下文的关联。如果用后续问题去搜索知识库有可能会搜索出无关的信息,从而导致大模型无法正确回答问题。请问这个项目要考虑这种情况吗?", "id": 206}
+{"title": "[BUG] 内存不足的问题", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/470", "detail": "我用了本地的chatglm-6b-int4模型,然后显示了内存不足(win10+32G内存+1080ti11G),一般需要多少内存才足够?这个bug应该如何解决?", "id": 207}
+{"title": "[BUG] 纯内网环境安装pycocotools失败", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/472", "detail": "**问题描述 / Problem Description**", "id": 208}
+{"title": "[BUG] webui.py 重新加载模型会导致 KeyError", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/473", "detail": "**问题描述 / Problem Description**", "id": 209}
+{"title": "chatyuan无法使用", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/475", "detail": "**问题描述 / Problem Description**", "id": 210}
+{"title": "[BUG] 文本分割模型AliTextSplitter存在bug,会把“.”作为分割符", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/476", "detail": "阿里达摩院的语义分割模型存在bug,默认会把\".”作为分割符进行分割而不管上下文语义。是否还有其他分割符则未知。建议的修改方案:把“.”统一替换为其他字符,分割后再替换回来。或者添加其他分割模型。", "id": 211}
+{"title": "[BUG] RuntimeError: Error in faiss::FileIOReader::FileIOReader(const char*) a", "file": "2023-05-27.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/479", "detail": "**问题描述 / Problem Description**", "id": 212}
+{"title": "[FEATURE] 安装,为什么conda create要额外指定路径 用-p ,而不是默认的/envs下面", "file": "2023-05-28.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/481", "detail": "##**功能描述 / Feature Description**", "id": 213}
+{"title": "[小白求助] 通过Anaconda执行webui.py后,无法打开web链接", "file": "2023-05-28.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/485", "detail": "在执行webui.py命令后,http://0.0.0.0:7860复制到浏览器后无法打开,显示“无法访问此网站”。", "id": 214}
+{"title": "[BUG] 使用 p-tuningv2后的模型,重新加载报错", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/486", "detail": "把p-tunningv2训练完后的相关文件放到了p-tunningv2文件夹下,勾选使用p-tuningv2点重新加载模型,控制台输错错误信息:", "id": 215}
+{"title": "[小白求助] 服务器上执行webui.py后,在本地无法打开web链接", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/487", "detail": "此项目执行在xxx.xx.xxx.xxx服务器上,我在webui.py上的代码为 (demo", "id": 216}
+{"title": "[FEATURE] 能不能支持VisualGLM-6B", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/488", "detail": "**功能描述 / Feature Description**", "id": 217}
+{"title": "你好,问一下各位,后端api部署的时候,支持多用户同时问答吗???", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/489", "detail": "支持多用户的话,最多支持多少用户问答?根据硬件而定吧?", "id": 218}
+{"title": "V100GPU显存占满,而利用率却为0,这是为什么?", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/491", "detail": "", "id": 219}
+{"title": "[求助] 如果在公司内部搭建产品知识库,使用INT-4模型,200人规模需要配置多少显存的服务器?", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/492", "detail": "如题,计划给公司搭一个在线知识库。", "id": 220}
+{"title": "你好,请教个问题,目前问答回复需要20秒左右,如何提高速度?V10032G服务器。", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/493", "detail": "**问题描述 / Problem Description**", "id": 221}
+{"title": "[FEATURE] 如何实现只匹配下文,而不要上文的结果", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/494", "detail": "在构建自己的知识库时,主要采用问答对的形式,那么也就是我需要的回答是在我的问题下面的内容,但是目前设置了chunk_size的值以后匹配的是上下文的内容,但我实际并不需要上文的。为了实现更完整的展示下面的答案,我只能调大chunk_size的值,但实际上上文的一半内容都是我不需要的。也就是扔了一半没用的东西给prompt,在faiss.py中我也没找到这块的一些描述,请问该如何进行修改呢?", "id": 222}
+{"title": "你好,问一下,我调用api.py部署,为什么用ip加端口可以使用postman调用,而改为域名使用postman无法调用?", "file": "2023-05-30.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/497", "detail": "![5ufBSWxLyF](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/70e2fbac-5699-48d0-b0d1-3dc84fd042c2)", "id": 223}
+{"title": "调用api.py中的stream_chat,返回source_documents中出现中文乱码。", "file": "2023-05-30.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/498", "detail": "-------------------------------------------------------------------------------", "id": 224}
+{"title": "[BUG] 捉个虫,api.py中的stream_chat解析json问题", "file": "2023-05-30.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/501", "detail": "**问题描述 / Problem Description**", "id": 225}
+{"title": "windows本地部署遇到了omp错误", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/502", "detail": "**问题描述 / Problem Description**", "id": 226}
+{"title": "[BUG] bug14 ,\"POST /local_doc_qa/upload_file HTTP/1.1\" 422 Unprocessable Entity", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/503", "detail": "上传的文件报错,返回错误,api.py", "id": 227}
+{"title": "你好,请教个问题,api.py部署的时候,如何改为多线程调用?谢谢", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/505", "detail": "目前的api.py脚本不支持多线程", "id": 228}
+{"title": "你好,请教一下。api.py部署的时候,能不能提供给后端流失返回结果。", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/507", "detail": "curl -X 'POST' \\", "id": 229}
+{"title": "流式输出,流式接口,使用server-sent events技术。", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/508", "detail": "想这样一样,https://blog.csdn.net/weixin_43228814/article/details/130063010", "id": 230}
+{"title": "计划增加流式输出功能吗?ChatGLM模型通过api方式调用响应时间慢怎么破,Fastapi流式接口来解惑,能快速提升响应速度", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/509", "detail": "**问题描述 / Problem Description**", "id": 231}
+{"title": "[BUG] 知识库上传时发生ERROR (could not open xxx for reading: No such file or directory)", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/510", "detail": "**问题描述 / Problem Description**", "id": 232}
+{"title": "api.py脚本打算增加SSE流式输出吗?", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/511", "detail": "curl调用的时候可以检测第一个字,从而提升回复的体验", "id": 233}
+{"title": "[BUG] 使用tornado实现webSocket,可以多个客户端同时连接,并且实现流式回复,但是多个客户端同时使用,答案就很乱,是模型不支持多线程吗", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/512", "detail": "import asyncio", "id": 234}
+{"title": "支持 chinese_alpaca_plus_lora 吗 基于llama的", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/514", "detail": "支持 chinese_alpaca_plus_lora 吗 基于llama的,https://github.com/ymcui/Chinese-LLaMA-Alpaca这个项目的", "id": 235}
+{"title": "[BUG] 现在能读图片的pdf了,但是文字的pdf反而读不了了,什么情况???", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/515", "detail": "**问题描述 / Problem Description**", "id": 236}
+{"title": "在推理的过程中卡住不动,进程无法正常结束", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/516", "detail": "**问题描述 / Problem Description**", "id": 237}
+{"title": "curl调用的时候,从第二轮开始,curl如何传参可以实现多轮对话?", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/517", "detail": "第一轮调用:", "id": 238}
+{"title": "建议添加api.py部署后的日志管理功能?", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/518", "detail": "-------------------------------------------------------------------------------", "id": 239}
+{"title": "有大佬知道,怎么多线程部署api.py脚本吗?", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/519", "detail": "api.py部署后,使用下面的请求,时间较慢,好像是单线程,如何改为多线程部署api.py:", "id": 240}
+{"title": "[BUG] 上传文件到知识库 任何格式与内容都永远失败", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/520", "detail": "上传知识库的时候,传txt无法解析,就算是穿content/sample里的样例txt也无法解析,上传md、pdf等都无法加载,会持续性等待,等到了超过30分钟也不行。", "id": 241}
+{"title": "关于prompt_template的问题", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/521", "detail": "请问这段prompt_template是什么意思,要怎么使用?可以给一个具体模板参考下吗?", "id": 242}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/522", "detail": "**问题描述 / Problem Description**", "id": 243}
+{"title": "中文分词句号处理(关于表达金额之间的\".\")", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/523", "detail": "建议处理12.6亿元的这样的分词,最好别分成12 和6亿这样的,需要放到一起", "id": 244}
+{"title": "ImportError: cannot import name 'inference' from 'paddle'", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/526", "detail": "在网上找了一圈,有说升级paddle的,我做了还是没有用,有说安装paddlepaddle的,我找了豆瓣的镜像源,但安装报错cannot detect archive format", "id": 245}
+{"title": "[BUG] webscoket 接口串行问题(/local_doc_qa/stream-chat/{knowledge_base_id})", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/527", "detail": "**问题描述 / Problem Description**", "id": 246}
+{"title": "[FEATURE] 刷新页面更新知识库列表", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/528", "detail": "**功能描述以及改进方案**", "id": 247}
+{"title": "[BUG] 使用ptuning微调模型后,问答效果并不好", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/530", "detail": "### 未调用ptuning", "id": 248}
+{"title": "[BUG] 多轮对话效果不佳", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/532", "detail": "在进行多轮对话的时候,无论设置的history_len是多少,效果都不好。事实上我将其设置成了最大值10,但在对话中,仍然无法实现多轮对话:", "id": 249}
+{"title": "RuntimeError: MPS backend out of memory (MPS allocated: 18.00 GB, other allocations: 4.87 MB, max allowed: 18.13 GB)", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/533", "detail": "**问题描述**", "id": 250}
+{"title": " 请大家重视这个issue!真正使用肯定是多用户并发问答,希望增加此功能!!!", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/534", "detail": "这得看你有多少显卡", "id": 251}
+{"title": "在启动项目的时候如何使用到多张gpu啊?", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/535", "detail": "**在启动项目的时候如何使用到多张gpu啊?**", "id": 252}
+{"title": " 使用流式输出的时候,curl调用的格式是什么?", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/536", "detail": "app.websocket(\"/local_doc_qa/stream-chat/{knowledge_base_id}\")(stream_chat)中的knowledge_base_id应该填什么???", "id": 253}
+{"title": "使用本地 vicuna-7b模型启动错误", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/538", "detail": "环境: ubuntu 22.04 cuda 12.1 没有安装nccl,使用rtx2080与m60显卡并行计算", "id": 254}
+{"title": "为什么会不调用GPU直接调用CPU呢", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/539", "detail": "我的阿里云配置是16G显存,用默认代码跑webui.py时提示", "id": 255}
+{"title": "上传多个文件时会互相覆盖", "file": "2023-06-03.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/541", "detail": "1、在同一个知识库中上传多个文件时会互相覆盖,无法结合多个文档的知识,有大佬知道怎么解决吗?", "id": 256}
+{"title": "[BUG] ‘gcc’不是内部或外部命令/LLM对话只能持续一轮", "file": "2023-06-03.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/542", "detail": "No compiled kernel found.", "id": 257}
+{"title": "以API模式启动项目却没有知识库的接口列表?", "file": "2023-06-04.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/544", "detail": "请问如何获取知识库的接口列表?如果没有需要自行编写的话,可不可以提供相关的获取方式,感谢", "id": 258}
+{"title": "程序以API模式启动的时候,如何才能让接口以stream模式被调用呢?", "file": "2023-06-05.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/546", "detail": "作者您好,我在以API模式进行程序启动后,我发现接口响应时间很长,怎么样才能让接口以stream模式被调用呢?我想实现像webui模式的回答那样", "id": 259}
+{"title": "关于原文中表格转为文本后数据相关度问题。", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/547", "detail": "原文中表格数据转换为文本,以 (X-Y:值;...) 的格式每一行组织成一句话,但这样做后发现相关度较低,效果很差,有何好的方案吗?", "id": 260}
+{"title": "启动后LLM和知识库问答模式均只有最后一轮记录", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/548", "detail": "拉取最新代码,问答时,每次页面只显示最后一次问答记录,需要修改什么参数才可以保留历史记录?", "id": 261}
+{"title": "提供system message配置,以便于让回答不要超出知识库范围", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/549", "detail": "**功能描述 / Feature Description**", "id": 262}
+{"title": "[BUG] 使用p-tunningv2报错", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/551", "detail": "按照readme的指示把p-tunningv2训练完后的文件放到了p-tunningv2文件夹下,勾选使用p-tuningv2点重新加载模型,控制台提示错误信息:", "id": 263}
+{"title": "[BUG] 智障,这么多问题,也好意思放出来,浪费时间", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/553", "detail": "。。。", "id": 264}
+{"title": "[FEATURE] 我看代码文件中有一个ali_text_splitter.py,为什么不用他这个文本分割器了?", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/554", "detail": "我看代码文件中有一个ali_text_splitter.py,为什么不用他这个文本分割器了?", "id": 265}
+{"title": "加载文档函数报错", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/557", "detail": "def load_file(filepath, sentence_size=SENTENCE_SIZE):", "id": 266}
+{"title": "参考指引安装docker后,运行cli_demo.py,提示killed", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/558", "detail": "root@b3d1bd08095c:/chatGLM# python3 cli_demo.py", "id": 267}
+{"title": "注意:如果安装错误,注意这两个包的版本 wandb==0.11.0 protobuf==3.18.3", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/559", "detail": "Error1: 如果启动异常报错 `protobuf` 需要更新到 `protobuf==3.18.3 `", "id": 268}
+{"title": "知识库对长文的知识相关度匹配不太理想有何优化方向", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/563", "detail": "我们可能录入一个文章有 1W 字,里面涉及这个文章主题的很多角度问题,我们针对他提问,他相关度匹配的内容和实际我们需要的答案相差很大怎么办。", "id": 269}
+{"title": "使用stream-chat函数进行流式输出的时候,能使用curl调用吗?", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/565", "detail": "为什么下面这样调用会报错???", "id": 270}
+{"title": "有大佬实践过 并行 或者 多线程 的部署方案吗?", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/566", "detail": "+1", "id": 271}
+{"title": "多线程部署遇到问题?", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/567", "detail": "", "id": 272}
+{"title": "[BUG] 用fastchat加载vicuna-13b模型进行知识库的问答有token的限制错误", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/569", "detail": "当我开启fastchat的vicuna-13b的api服务,然后config那里配置好(api本地测试过可以返回结果),然后知识库加载好之后(知识库大概有1000多个文档,用chatGLM可以正常推理),进行问答时出现token超过限制,就问了一句hello;", "id": 273}
+{"title": "现在的添加知识库,文件多了总是报错,也不知道自己加载了哪些文件,报错后也不知道是全部失败还是一部分成功;希望能有个加载指定文件夹作为知识库的功能", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/574", "detail": "**功能描述 / Feature Description**", "id": 274}
+{"title": "[BUG] moss模型本地加载报错", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/577", "detail": "moss模型本地加载报错:", "id": 275}
+{"title": "加载本地moss模型报错Can't instantiate abstract class MOSSLLM with abstract methods _history_len", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/578", "detail": "(vicuna) ps@ps[13:56:20]:/data/chat/langchain-ChatGLM2/langchain-ChatGLM-0.1.13$ python webui.py --model-dir local_models --model moss --no-remote-model", "id": 276}
+{"title": "[FEATURE] 能增加在前端页面控制prompt_template吗?或是能支持前端页面选择使用哪个prompt?", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/579", "detail": "目前只能在config里修改一个prompt,想在多个不同场景切换比较麻烦", "id": 277}
+{"title": "[BUG] streamlit ui的bug,在增加知识库时会报错", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/580", "detail": "**问题描述 / Problem Description**", "id": 278}
+{"title": "[FEATURE] webui/webui_st可以支持history吗?目前仅能一次对话", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/581", "detail": "试了下webui和webui_st都不支持历史对话啊,只能对话一次,不能默认开启所有history吗?", "id": 279}
+{"title": "启动python cli_demo.py --model chatglm-6b-int4-qe报错", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/585", "detail": "下载好模型,和相关依赖环境,之间运行`python cli_demo.py --model chatglm-6b-int4-qe`报错了:", "id": 280}
+{"title": "重新构建知识库报错", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/586", "detail": "**问题描述 / Problem Description**", "id": 281}
+{"title": "[FEATURE] 能否屏蔽paddle,我不需要OCR,效果差依赖环境还很复杂", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/587", "detail": "希望能不依赖paddle", "id": 282}
+{"title": "question :文档向量化这个可以自己手动实现么?", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/589", "detail": "现有公司级数据500G+,需要使用这个功能,请问如何手动实现这个向量化,然后并加载", "id": 283}
+{"title": "view前端能进行流式的返回吗??", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/590", "detail": "view前端能进行流式的返回吗??", "id": 284}
+{"title": "[BUG] Load parallel cpu kernel failed, using default cpu kernel code", "file": "2023-06-11.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/594", "detail": "**问题描述 / Problem Description**", "id": 285}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-11.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/595", "detail": "**问题描述 / Problem Description**", "id": 286}
+{"title": "我在上传本地知识库时提示KeyError: 'name'错误,本地知识库都是.txt文件,文件数量大约是2000+。", "file": "2023-06-12.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/597", "detail": "", "id": 287}
+{"title": "model_config.py中有vicuna-13b-hf模型的配置信息,但是好像还是不可用?", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/600", "detail": "@dongyihua543", "id": 288}
+{"title": "ImportError: Using SOCKS proxy, but the 'socksio' package is not installed. Make sure to install httpx using `pip install httpx[socks]`.", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/605", "detail": "应该代理问题,但是尝试了好多方法都解决不了,", "id": 289}
+{"title": "[BUG] similarity_search_with_score_by_vector在找不到匹配的情况下出错", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/607", "detail": "在设置匹配阈值 VECTOR_SEARCH_SCORE_THRESHOLD 的情况下,vectorstore会返回空,此时上述处理函数会出错", "id": 290}
+{"title": "[FEATURE] 请问如何搭建英文知识库呢", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/609", "detail": "**功能描述 / Feature Description**", "id": 291}
+{"title": "谁有vicuna权重?llama转换之后的", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/611", "detail": "**问题描述 / Problem Description**", "id": 292}
+{"title": "[FEATURE] API能实现上传文件夹的功能么?", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/612", "detail": "用户懒得全选所有的文件,就想上传个文件夹,请问下API能实现这个功能么?", "id": 293}
+{"title": "请问在多卡部署后,上传单个文件作为知识库,用的是单卡在生成向量还是多卡?", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/614", "detail": "目前我检测我本地多卡部署的,好像生成知识库向量的时候用的还是单卡", "id": 294}
+{"title": "[BUG] python webui.py提示非法指令", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/615", "detail": "(/data/conda-langchain [root@chatglm langchain-ChatGLM]# python webui.py", "id": 295}
+{"title": "知识库文件跨行切分问题", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/616", "detail": "我的知识库文件txt文件,是一行一条知识,用\\n分行。", "id": 296}
+{"title": "[FEATURE] bing搜索问答有流式的API么?", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/617", "detail": "web端是有这个bing搜索回答,但api接口没有发现,大佬能给个提示么?", "id": 297}
+{"title": "希望出一个macos m2的安装教程", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/620", "detail": "mac m2安装,模型加载成功了,知识库文件也上传成功了,但是一问答就会报错,报错内容如下", "id": 298}
+{"title": "为【出处】提供高亮显示", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/621", "detail": "具体出处里面,对相关的内容高亮显示,不包含前后文。", "id": 299}
+{"title": "[BUG] CPU运行cli_demo.py,不回答,hang住", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/622", "detail": "没有GPU;32G内存的ubuntu机器。", "id": 300}
+{"title": "关于删除知识库里面的文档后,LLM知识库对话的时候还是会返回该被删除文档的内容", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/623", "detail": "如题,在vue前端成功执行删除知识库里面文档A.txt后,未能也在faiss索引中也删除该文档,LLM还是会返回这个A.txt的内容,并且以A.txt为出处,未能达到删除的效果", "id": 301}
+{"title": "[BUG] 调用知识库进行问答,显存会一直叠加", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/625", "detail": "14G的显存,调用的chatglm-6b-int8模型,进行知识库问答时,最多问答四次就会爆显存了,观察了一下显存使用情况,每一次使用就会增加一次显存,请问这样是正常的吗?是否有什么配置需要开启可以解决这个问题?例如进行一次知识库问答清空上次问题的显存?", "id": 302}
+{"title": "[BUG] web页面 重新构建数据库 失败,导致 原来的上传的数据库都没了", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/626", "detail": "web页面 重新构建数据库 失败,导致 原来的上传的数据库都没了", "id": 303}
+{"title": "在CPU上运行webui.py报错Tensor on device cpu is not on the expected device meta!", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/627", "detail": "在CPU上运行python webui.py能启动,但最后有:RuntimeError: Tensor on device cpu is not on the expected device meta!", "id": 304}
+{"title": "OSError: [WinError 1114] 动态链接库(DLL)初始化例程失败。 Error loading \"E:\\xxx\\envs\\langchain\\lib\\site-packages\\torch\\lib\\caffe2_nvrtc.dll\" or one of its dependencies.哪位大佬知道如何解决吗?", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/629", "detail": "**问题描述 / Problem Description**", "id": 305}
+{"title": "[BUG] WEBUI删除知识库文档,会导致知识库问答失败", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/632", "detail": "如题,从知识库已有文件中选择要删除的文件,点击删除后,在问答框输入内容回车报错", "id": 306}
+{"title": "更新后的版本中,删除知识库中的文件,再提问出现error错误", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/634", "detail": "针对更新版本,识别到一个问题,过程如下:", "id": 307}
+{"title": "我配置好了环境,想要实现本地知识库的问答?可是它返回给我的", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/637", "detail": "没有总结,只有相关度的回复,但是我看演示里面表现的,回复是可以实现总结的,我去查询代码", "id": 308}
+{"title": "[BUG] NPM run dev can not successfully start the VUE frontend", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/638", "detail": "**问题描述 / Problem Description**", "id": 309}
+{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/639", "detail": "**问题描述 / Problem Description**", "id": 310}
+{"title": "提一个模型加载的bug,我在截图中修复了,你们有空可以看一下。", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/642", "detail": "![model_load_bug](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/59411575/4432adc4-ccdd-45d9-aafc-5f2d1963403b)", "id": 311}
+{"title": "[求助]关于设置embedding model路径的问题", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/643", "detail": "如题,我之前成功跑起来过一次,但因环境丢失重新配置 再运行webui就总是报错", "id": 312}
+{"title": "Lora微调后的模型可以直接使用吗", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/646", "detail": "看model_config.py里是有USE_LORA这个参数的,但是在cli_demo.py和webui.py这两个里面都没有用到,实际测试下来模型没有微调的效果,想问问现在这个功能实现了吗", "id": 313}
+{"title": "write_check_file在tmp_files目录下生成的load_file.txt是否需要一直保留,占用空间很大,在建完索引后能否删除", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/647", "detail": "**功能描述 / Feature Description**", "id": 314}
+{"title": "[BUG] /local_doc_qa/list_files?knowledge_base_id=test删除知识库bug", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/649", "detail": "1.新建test知识库并上传文件(在vue前端完成并检查后端发现确实生成了test文件夹以及下面的content和vec_store", "id": 315}
+{"title": "[BUG] vue webui无法加载知识库", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/650", "detail": "拉取了最新的代码,分别运行了后端api和前端web,点击知识库,始终只能显示simple,无法加载知识库", "id": 316}
+{"title": "不能本地加载moss模型吗?", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/652", "detail": "手动下载模型设置local_model_path路径依旧提示缺少文件,该如何正确配置?", "id": 317}
+{"title": "macos m2 pro docker 安装失败", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/654", "detail": "macos m2 pro docker 安装失败", "id": 318}
+{"title": " [BUG] mac m1 pro 运行提示 zsh: segmentation fault", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/655", "detail": "运行: python webui.py", "id": 319}
+{"title": "安装 requirements 报错", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/656", "detail": "(langchainchatglm) D:\\github\\langchain-ChatGLM>pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple/", "id": 320}
+{"title": "[BUG] AssertionError", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/658", "detail": "**问题描述 / Problem Description**", "id": 321}
+{"title": "[FEATURE] 支持AMD win10 本地部署吗?", "file": "2023-06-18.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/660", "detail": "**功能描述 / Feature Description**", "id": 322}
diff --git a/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.xlsx b/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.xlsx
new file mode 100644
index 0000000000000000000000000000000000000000..3503e2d2b0f35967323f3dc93ce290046e7d68dc
Binary files /dev/null and b/knowledge_base/samples/content/test_files/langchain-ChatGLM_open.xlsx differ
diff --git a/knowledge_base/samples/content/test_files/langchain.pdf b/knowledge_base/samples/content/test_files/langchain.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..eb439f31915a553f9b7f589b69026730b64ca500
--- /dev/null
+++ b/knowledge_base/samples/content/test_files/langchain.pdf
@@ -0,0 +1,6446 @@
+%PDF-1.7
+%����
+1 0 obj
+<<
+/Type /Catalog
+/Pages 2 0 R
+/Lang (tr)
+/StructTreeRoot 3 0 R
+/MarkInfo <<
+/Marked true
+>>
+/Metadata 4 0 R
+/ViewerPreferences 5 0 R
+>>
+endobj
+6 0 obj
+<<
+/Author (TuRGuT)
+/Creator
+/CreationDate (D:20230722114502+03'00')
+/ModDate (D:20230722114502+03'00')
+/Producer
+/rgid (PB:372669736_AS:11431281179897226@1691423911543)
+>>
+endobj
+2 0 obj
+<<
+/Type /Pages
+/Count 8
+/Kids [7 0 R 8 0 R 9 0 R 10 0 R 11 0 R 12 0 R 13 0 R 14 0 R]
+>>
+endobj
+3 0 obj
+<<
+/Type /StructTreeRoot
+/RoleMap 15 0 R
+/ParentTree 16 0 R
+/K [17 0 R]
+/ParentTreeNextKey 7
+>>
+endobj
+4 0 obj
+<<
+/Length 3086
+/Type /Metadata
+/Subtype /XML
+>>
+stream
+
+
+
+Microsoft® Word Microsoft 365 için
+
+TuRGuT
+
+Microsoft® Word Microsoft 365 için2023-07-22T11:45:02+03:002023-07-22T11:45:02+03:00
+
+uuid:3151B3E9-B7DC-42C1-93F2-C1A06B88013Cuuid:3151B3E9-B7DC-42C1-93F2-C1A06B88013C
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+endstream
+endobj
+5 0 obj
+<<
+/DisplayDocTitle true
+>>
+endobj
+7 0 obj
+<<
+/Type /Page
+/Resources <<
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+/ExtGState <<
+/G3 18 0 R
+/G11 19 0 R
+/G12 20 0 R
+>>
+/XObject <<
+/X8 21 0 R
+/X10 22 0 R
+/X13 23 0 R
+>>
+/Font <<
+/F4 24 0 R
+/F5 25 0 R
+/F6 26 0 R
+/F7 27 0 R
+>>
+>>
+/MediaBox [0 0 594.95996 840.95996]
+/Annots [28 0 R 29 0 R 30 0 R 31 0 R 32 0 R 33 0 R 34 0 R 35 0 R 36 0 R 37 0 R
+38 0 R 39 0 R 40 0 R]
+/Contents 41 0 R
+/StructParents 0
+/Parent 2 0 R
+>>
+endobj
+8 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F2 43 0 R
+/F3 44 0 R
+/F4 45 0 R
+/F5 46 0 R
+/F6 47 0 R
+/F7 48 0 R
+/F8 49 0 R
+/F9 50 0 R
+/F10 51 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/XObject <<
+/Image22 54 0 R
+/Image24 55 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/Annots [56 0 R 57 0 R]
+/MediaBox [0 0 595.32 842.04]
+/Contents 58 0 R
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 0
+>>
+endobj
+9 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F5 46 0 R
+/F4 45 0 R
+/F10 51 0 R
+/F2 43 0 R
+/F11 59 0 R
+/F6 47 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/MediaBox [0 0 595.32 842.04]
+/Contents 60 0 R
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 1
+>>
+endobj
+10 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F5 46 0 R
+/F2 43 0 R
+/F11 59 0 R
+/F6 47 0 R
+/F4 45 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/XObject <<
+/Image47 61 0 R
+/Image48 62 0 R
+/Image49 63 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/MediaBox [0 0 595.32 842.04]
+/Contents 64 0 R
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 2
+>>
+endobj
+11 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F5 46 0 R
+/F2 43 0 R
+/F11 59 0 R
+/F12 65 0 R
+/F10 51 0 R
+/F6 47 0 R
+/F13 66 0 R
+/F9 50 0 R
+/F4 45 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/XObject <<
+/Image62 67 0 R
+/Image63 68 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/MediaBox [0 0 595.32 842.04]
+/Contents 69 0 R
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 3
+>>
+endobj
+12 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F5 46 0 R
+/F2 43 0 R
+/F11 59 0 R
+/F12 65 0 R
+/F10 51 0 R
+/F4 45 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/XObject <<
+/Image66 70 0 R
+/Image67 71 0 R
+/Image68 72 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/MediaBox [0 0 595.32 842.04]
+/Contents 73 0 R
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 4
+>>
+endobj
+13 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F5 46 0 R
+/F10 51 0 R
+/F4 45 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/MediaBox [0 0 595.32 842.04]
+/Contents 74 0 R
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 5
+>>
+endobj
+14 0 obj
+<<
+/Type /Page
+/Parent 2 0 R
+/Resources <<
+/Font <<
+/F1 42 0 R
+/F5 46 0 R
+/F10 51 0 R
+/F4 75 0 R
+>>
+/ExtGState <<
+/GS7 52 0 R
+/GS8 53 0 R
+>>
+/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]
+>>
+/MediaBox [0 0 595.32 842.04]
+/Contents [76 0 R 77 0 R]
+/Group <<
+/Type /Group
+/S /Transparency
+/CS /DeviceRGB
+>>
+/Tabs /S
+/StructParents 6
+/Annots [78 0 R]
+>>
+endobj
+15 0 obj
+<<
+/Footnote /Note
+/Endnote /Note
+/Textbox /Sect
+/Header /Sect
+/Footer /Sect
+/InlineShape /Sect
+/Annotation /Sect
+/Artifact /Sect
+/Workbook /Document
+/Worksheet /Part
+/Macrosheet /Part
+/Chartsheet /Part
+/Dialogsheet /Part
+/Slide /Part
+/Chart /Sect
+/Diagram /Figure
+/Title /H1
+>>
+endobj
+16 0 obj
+<<
+/Nums [0 [79 0 R 80 0 R 81 0 R 82 0 R 83 0 R 84 0 R 85 0 R 86 0 R 87 0 R 88 0 R
+89 0 R 90 0 R 91 0 R 92 0 R 93 0 R]
+ 1 [94 0 R 95 0 R 96 0 R 97 0 R 98 0 R 99 0 R 100 0 R 100 0 R 101 0 R 102 0 R
+103 0 R 104 0 R 105 0 R 106 0 R 107 0 R 108 0 R 109 0 R 110 0 R 111 0 R 112 0 R
+113 0 R 114 0 R 115 0 R 116 0 R 117 0 R 118 0 R 119 0 R 120 0 R]
+ 2 [121 0 R 122 0 R 123 0 R 124 0 R 125 0 R 126 0 R 127 0 R 128 0 R 129 0 R 130 0 R
+131 0 R 131 0 R 131 0 R 132 0 R 133 0 R 134 0 R 135 0 R 136 0 R 137 0 R 138 0 R
+139 0 R 140 0 R 141 0 R 142 0 R 143 0 R 144 0 R 145 0 R 146 0 R 147 0 R 148 0 R
+149 0 R]
+ 3 [150 0 R 151 0 R 152 0 R 153 0 R 154 0 R 155 0 R 156 0 R 157 0 R 158 0 R 159 0 R
+160 0 R 161 0 R 162 0 R 163 0 R 164 0 R 165 0 R 166 0 R 167 0 R 168 0 R 169 0 R
+170 0 R 171 0 R 172 0 R 173 0 R 174 0 R 175 0 R 176 0 R]
+ 4 [177 0 R 178 0 R 179 0 R 180 0 R 181 0 R 182 0 R 183 0 R 184 0 R 185 0 R 186 0 R
+187 0 R 188 0 R 189 0 R 190 0 R 191 0 R 192 0 R 193 0 R 194 0 R 195 0 R 196 0 R
+197 0 R 197 0 R 197 0 R 197 0 R 198 0 R 199 0 R 200 0 R]
+5 [201 0 R 201 0 R 202 0 R 203 0 R 204 0 R 205 0 R 206 0 R 207 0 R 208 0 R 209 0 R
+210 0 R 211 0 R 212 0 R 213 0 R 214 0 R 215 0 R 216 0 R 217 0 R 218 0 R 219 0 R
+220 0 R 221 0 R 222 0 R]
+ 6 [223 0 R 224 0 R 225 0 R 226 0 R]
+]
+>>
+endobj
+17 0 obj
+<<
+/P 3 0 R
+/S /Document
+/Type /StructElem
+/K [79 0 R 80 0 R 81 0 R 82 0 R 83 0 R 84 0 R 85 0 R 86 0 R 87 0 R 88 0 R
+89 0 R 227 0 R 91 0 R 92 0 R 93 0 R 94 0 R 95 0 R 96 0 R 97 0 R 98 0 R
+228 0 R 100 0 R 229 0 R 102 0 R 104 0 R 105 0 R 230 0 R 118 0 R 120 0 R 121 0 R
+123 0 R 124 0 R 125 0 R 126 0 R 147 0 R 127 0 R 128 0 R 129 0 R 148 0 R 130 0 R
+131 0 R 132 0 R 133 0 R 149 0 R 134 0 R 135 0 R 136 0 R 231 0 R 143 0 R 145 0 R
+146 0 R 150 0 R 152 0 R 153 0 R 154 0 R 155 0 R 175 0 R 156 0 R 157 0 R 158 0 R
+232 0 R 164 0 R 165 0 R 166 0 R 167 0 R 168 0 R 169 0 R 170 0 R 171 0 R 172 0 R
+173 0 R 174 0 R 177 0 R 178 0 R 179 0 R 180 0 R 181 0 R 182 0 R 183 0 R 184 0 R
+185 0 R 186 0 R 187 0 R 188 0 R 189 0 R 191 0 R 233 0 R 194 0 R 195 0 R 234 0 R
+197 0 R 201 0 R 235 0 R 203 0 R 204 0 R 236 0 R 206 0 R 207 0 R 208 0 R 209 0 R
+210 0 R 237 0 R 238 0 R 224 0 R 225 0 R 226 0 R]
+>>
+endobj
+18 0 obj
+<<
+/ca 1
+/BM /Normal
+>>
+endobj
+19 0 obj
+<<
+/ca .2
+/BM /Normal
+>>
+endobj
+20 0 obj
+<<
+/CA 1
+/ca 1
+/LC 0
+/LJ 0
+/LW 1
+/ML 4
+/SA true
+/BM /Normal
+>>
+endobj
+21 0 obj
+<<
+/Length 4601
+/Type /XObject
+/Subtype /Image
+/Width 320
+/Height 320
+/ColorSpace /DeviceRGB
+/SMask 239 0 R
+/BitsPerComponent 8
+/Filter /FlateDecode
+>>
+stream
+x��yp��}�h i�f&i2�2S�4d�IҤM�I�v��1M�6�N2iC�M�dhJ��Ʒ�|��`cc|b�ԧ$�>�u�CƦHZ�m�p��:˫Z��~�Ϯ�<���}w�w?�~��_�R Ѕ7����������f���������R�:f���z����\i�t�*\�n��]߸�Kw�7�Q63�p�:r�d�����k�-�U����t���!�z��1�l�k�k��͍h�mlm=�N���.l��6�k��j���ce�p��p�� ?
m�k���N��
+�[2{��o �O}�����m�glۣ�M�% 8�X�����^h?\mm
��&*���Dj��o]fGJy}�֥����W.��
+�^#�O1X��|�b[}[��� ����u�+oc[˹�v����)��V^v�����h��sFJyk��t��K�
�-�� ��)&mG��[��Z�
JP
+�+Sl�V����˗���Gޗ"���%{O���ȇ�,Ej籬s�/�rF �}S��t���6�Z����;[��
+�c(6�5)f;��j�mki�ұE}��M?Kx��[k��}f�J�'�
��1hV�.6��6���"�X�:���7Q��D��9��\���cDTik��3��-�#�Q��7�o�[�G�!�Ў[G�%�$py��J;��n�}��j�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�����W� ʥ�)��%+�M�_,Y�o��b�JS�Kֹd���KK����X����Œ����/���7E�d��)��%+�M�_,Y�o��b�JS�KV����X��_���o�_,E��U@�B?{/�Y��Y�~�U@�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt�r�-�#�Q���~��!�U�Џ<:D����G�(W�B?���[�G�\u�ȣC��n�yt��d��6���/O۳[>3�Su�ȣKĹ�>��"�?�X7l��i�w�p�|n�N߳������o>/��~���~��%������;F��o�%�Q�OM����Ç�s��T�Џ<�D�U3���П�Y��������������u澽��m�:���f/�-�#�.�����Ok+��3gE��G/�9!�3�~h��,%���~��%����߿�T8�?~��'2s�G�<�~�<�@�[�ȣK�x�YTn[zY�Vn`mş?���Wx�k�3���9�Osu�Iy�q�o1#�.��}d�Ɯc�=x�WG� F����P���bF]"�_sBUU�X֖������YRs