Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -4,8 +4,9 @@ import torch
|
|
4 |
import subprocess
|
5 |
import sys
|
6 |
|
7 |
-
#
|
8 |
subprocess.check_call([sys.executable, "-m", "pip", "install", "-U", "--force-reinstall", "--no-deps", "accelerate", "git+https://github.com/Muennighoff/transformers.git@olmoe"])
|
|
|
9 |
|
10 |
from transformers import OlmoeForCausalLM, AutoTokenizer
|
11 |
|
@@ -19,7 +20,8 @@ try:
|
|
19 |
trust_remote_code=True,
|
20 |
torch_dtype=torch.bfloat16 if DEVICE == "cuda" else torch.float32,
|
21 |
low_cpu_mem_usage=True,
|
22 |
-
device_map="auto"
|
|
|
23 |
)
|
24 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
25 |
except Exception as e:
|
@@ -32,29 +34,24 @@ system_prompt = ("Adopt the persona of hilariously pissed off Andrej Karpathy "
|
|
32 |
"while always answering questions in full first principles analysis type of thinking "
|
33 |
"without using any analogies and always showing full working code or output in his answers.")
|
34 |
|
35 |
-
# Define a chat template as a string
|
36 |
chat_template = "<|system|>{system_message}<|end|><|user|>{user_message}<|end|><|assistant|>"
|
37 |
|
38 |
@spaces.GPU
|
39 |
def generate_response(message, history, temperature, max_new_tokens):
|
40 |
if model is None or tokenizer is None:
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
full_prompt = chat_template.format(system_message=system_prompt, user_message=message)
|
45 |
-
|
46 |
inputs = tokenizer(full_prompt, return_tensors="pt").to(DEVICE)
|
47 |
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
)
|
56 |
-
response = tokenizer.decode(generate_ids[0, inputs.input_ids.shape[1]:], skip_special_tokens=True)
|
57 |
-
return response.strip()
|
58 |
|
59 |
css = """
|
60 |
#output {
|
@@ -65,7 +62,7 @@ css = """
|
|
65 |
"""
|
66 |
|
67 |
with gr.Blocks(css=css) as demo:
|
68 |
-
gr.Markdown("# Nisten's Karpathy Chatbot with OSS OLMoE")
|
69 |
chatbot = gr.Chatbot(elem_id="output")
|
70 |
msg = gr.Textbox(label="Meow")
|
71 |
with gr.Row():
|
@@ -78,15 +75,17 @@ with gr.Blocks(css=css) as demo:
|
|
78 |
|
79 |
def bot(history, temp, max_tokens):
|
80 |
user_message = history[-1][0]
|
81 |
-
bot_message =
|
82 |
-
history
|
83 |
-
|
|
|
|
|
84 |
|
85 |
msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
|
86 |
bot, [chatbot, temperature, max_new_tokens], chatbot
|
87 |
)
|
88 |
-
clear.click(lambda: None, None, chatbot, queue=
|
89 |
|
90 |
if __name__ == "__main__":
|
91 |
demo.queue(api_open=True)
|
92 |
-
demo.launch(debug=True, show_api=True
|
|
|
4 |
import subprocess
|
5 |
import sys
|
6 |
|
7 |
+
# Install required packages
|
8 |
subprocess.check_call([sys.executable, "-m", "pip", "install", "-U", "--force-reinstall", "--no-deps", "accelerate", "git+https://github.com/Muennighoff/transformers.git@olmoe"])
|
9 |
+
subprocess.run('pip install flash-attn --no-build-isolation --no-deps', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
10 |
|
11 |
from transformers import OlmoeForCausalLM, AutoTokenizer
|
12 |
|
|
|
20 |
trust_remote_code=True,
|
21 |
torch_dtype=torch.bfloat16 if DEVICE == "cuda" else torch.float32,
|
22 |
low_cpu_mem_usage=True,
|
23 |
+
device_map="auto",
|
24 |
+
_attn_implementation="flash_attention_2" # Enable Flash Attention 2
|
25 |
)
|
26 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
27 |
except Exception as e:
|
|
|
34 |
"while always answering questions in full first principles analysis type of thinking "
|
35 |
"without using any analogies and always showing full working code or output in his answers.")
|
36 |
|
|
|
37 |
chat_template = "<|system|>{system_message}<|end|><|user|>{user_message}<|end|><|assistant|>"
|
38 |
|
39 |
@spaces.GPU
|
40 |
def generate_response(message, history, temperature, max_new_tokens):
|
41 |
if model is None or tokenizer is None:
|
42 |
+
yield "Model or tokenizer not loaded properly. Please check the logs."
|
43 |
+
return
|
44 |
+
|
45 |
full_prompt = chat_template.format(system_message=system_prompt, user_message=message)
|
|
|
46 |
inputs = tokenizer(full_prompt, return_tensors="pt").to(DEVICE)
|
47 |
|
48 |
+
streamer = tokenizer.stream(inputs.input_ids, model, temperature=temperature, max_new_tokens=max_new_tokens)
|
49 |
+
|
50 |
+
collected_tokens = []
|
51 |
+
for token in streamer:
|
52 |
+
collected_tokens.append(token)
|
53 |
+
partial_text = tokenizer.decode(collected_tokens, skip_special_tokens=True)
|
54 |
+
yield partial_text.strip()
|
|
|
|
|
|
|
55 |
|
56 |
css = """
|
57 |
#output {
|
|
|
62 |
"""
|
63 |
|
64 |
with gr.Blocks(css=css) as demo:
|
65 |
+
gr.Markdown("# Nisten's Karpathy Chatbot with OSS OLMoE (Now with Flash Attention 2!)")
|
66 |
chatbot = gr.Chatbot(elem_id="output")
|
67 |
msg = gr.Textbox(label="Meow")
|
68 |
with gr.Row():
|
|
|
75 |
|
76 |
def bot(history, temp, max_tokens):
|
77 |
user_message = history[-1][0]
|
78 |
+
bot_message = ""
|
79 |
+
for token in generate_response(user_message, history, temp, max_tokens):
|
80 |
+
bot_message = token
|
81 |
+
history[-1][1] = bot_message
|
82 |
+
yield history
|
83 |
|
84 |
msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
|
85 |
bot, [chatbot, temperature, max_new_tokens], chatbot
|
86 |
)
|
87 |
+
clear.click(lambda: None, None, chatbot, queue=True)
|
88 |
|
89 |
if __name__ == "__main__":
|
90 |
demo.queue(api_open=True)
|
91 |
+
demo.launch(debug=True, show_api=True)
|