TheBloke commited on
Commit
51d64bc
·
1 Parent(s): 7c31db3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +86 -2
README.md CHANGED
@@ -1,6 +1,12 @@
1
  ---
2
  inference: false
3
  license: other
 
 
 
 
 
 
4
  ---
5
 
6
  <!-- header start -->
@@ -29,6 +35,25 @@ It is the result of quantising to 4bit using [GPTQ-for-LLaMa](https://github.com
29
  * [4-bit, 5-bit and 8-bit GGML models for CPU(+GPU) inference](https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML)
30
  * [Unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/NousResearch/Nous-Hermes-13b)
31
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
32
  ## How to easily download and use this model in text-generation-webui
33
 
34
  ### Downloading the model
@@ -91,9 +116,68 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
91
  * Patreon: https://patreon.com/TheBlokeAI
92
  * Ko-Fi: https://ko-fi.com/TheBlokeAI
93
 
94
- **Patreon special mentions**: Aemon Algiz; Dmitiry Samsonov; Jonathan Leane; Illia Dulskyi; Khalefa Al-Ahmad; Nikolai Manek; senxiiz; Talal Aujan; vamX; Eugene Pentland; Lone Striker; Luke Pendergrass; Johann-Peter Hartmann.
95
 
96
  Thank you to all my generous patrons and donaters.
97
  <!-- footer end -->
98
 
99
- # Original model card: NousResearch's Nous-Hermes-13B
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  inference: false
3
  license: other
4
+ language:
5
+ - en
6
+ tags:
7
+ - llama
8
+ - self-instruct
9
+ - distillation
10
  ---
11
 
12
  <!-- header start -->
 
35
  * [4-bit, 5-bit and 8-bit GGML models for CPU(+GPU) inference](https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML)
36
  * [Unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/NousResearch/Nous-Hermes-13b)
37
 
38
+ ## Prompt Template
39
+
40
+ The model follows the Alpaca prompt format:
41
+ ```
42
+ ### Instruction:
43
+
44
+ ### Response:
45
+ ```
46
+
47
+ or
48
+
49
+ ```
50
+ ### Instruction:
51
+
52
+ ### Input:
53
+
54
+ ### Response:
55
+ ```
56
+
57
  ## How to easily download and use this model in text-generation-webui
58
 
59
  ### Downloading the model
 
116
  * Patreon: https://patreon.com/TheBlokeAI
117
  * Ko-Fi: https://ko-fi.com/TheBlokeAI
118
 
119
+ **Patreon special mentions**: Aemon Algiz; Dmitiry Samsonov; Mano Prime; David Flickinger; Jonathan Leane; Illia Dulskyi; Khalefa Al-Ahmad; Nikolai Manek; senxiiz; Talal Aujan; vamX; Eugene Pentland; Lone Striker; Luke Pendergrass; Johann-Peter Hartmann.
120
 
121
  Thank you to all my generous patrons and donaters.
122
  <!-- footer end -->
123
 
124
+ # Original model card: NousResearch's Nous-Hermes-13B
125
+
126
+ # Model Card: Nous-Hermes-13b
127
+
128
+ ## Model Description
129
+
130
+ Nous-Hermes-13b is a state-of-the-art language model fine-tuned on over 300,000 instructions. This model was fine-tuned by Nous Research, with Teknium and Karan4D leading the fine tuning process and dataset curation, Redmond AI sponsoring the compute, and several other contributors. The result is an enhanced Llama 13b model that rivals GPT-3.5-turbo in performance across a variety of tasks.
131
+
132
+ This model stands out for its long responses, low hallucination rate, and absence of OpenAI censorship mechanisms. The fine-tuning process was performed with a 2000 sequence length on an 8x a100 80GB DGX machine for over 50 hours.
133
+
134
+ ## Model Training
135
+
136
+ The model was trained almost entirely on synthetic GPT-4 outputs. This includes data from diverse sources such as GPTeacher, the general, roleplay v1&2, code instruct datasets, Nous Instruct & PDACTL (unpublished), CodeAlpaca, Evol_Instruct Uncensored, GPT4-LLM, and Unnatural Instructions.
137
+
138
+ Additional data inputs came from Camel-AI's Biology/Physics/Chemistry and Math Datasets, Airoboros' GPT-4 Dataset, and more from CodeAlpaca. The total volume of data encompassed over 300,000 instructions.
139
+
140
+ ## Collaborators
141
+ The model fine-tuning and the datasets were a collaboration of efforts and resources between Teknium, Karan4D, Nous Research, Huemin Art, and Redmond AI.
142
+
143
+ Huge shoutout and acknowledgement is deserved for all the dataset creators who generously share their datasets openly.
144
+
145
+ Special mention goes to @winglian, @erhartford, and @main_horse for assisting in some of the training issues.
146
+
147
+ Among the contributors of datasets, GPTeacher was made available by Teknium, Wizard LM by nlpxucan, and the Nous Research Instruct Dataset was provided by Karan4D and HueminArt.
148
+ The GPT4-LLM and Unnatural Instructions were provided by Microsoft, Airoboros dataset by jondurbin, Camel-AI datasets are from Camel-AI, and CodeAlpaca dataset by Sahil 2801.
149
+ If anyone was left out, please open a thread in the community tab.
150
+
151
+ ## Prompt Format
152
+
153
+ The model follows the Alpaca prompt format:
154
+ ```
155
+ ### Instruction:
156
+
157
+ ### Response:
158
+ ```
159
+
160
+ or
161
+
162
+ ```
163
+ ### Instruction:
164
+
165
+ ### Input:
166
+
167
+ ### Response:
168
+ ```
169
+
170
+ ## Resources for Applied Use Cases:
171
+ For an example of a back and forth chatbot using huggingface transformers and discord, check out: https://github.com/teknium1/alpaca-discord
172
+ For an example of a roleplaying discord bot, check out this: https://github.com/teknium1/alpaca-roleplay-discordbot
173
+
174
+ ## Future Plans
175
+ The model is currently being uploaded in FP16 format, and there are plans to convert the model to GGML and GPTQ 4bit quantizations. The team is also working on a full benchmark, similar to what was done for GPT4-x-Vicuna. We will try to get in discussions to get the model included in the GPT4All.
176
+
177
+ ## Benchmark Results
178
+ Benchmark results are coming soon.
179
+
180
+ ## Model Usage
181
+ The model is available for download on Hugging Face. It is suitable for a wide range of language tasks, from generating creative text to understanding and following complex instructions.
182
+
183
+ Compute provided by our project sponsor Redmond AI, thank you!!