Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,87 @@
|
|
1 |
---
|
2 |
license: other
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: other
|
3 |
+
inference: false
|
4 |
---
|
5 |
+
# Vicuna 7B GPTQ 4-bit 128g
|
6 |
+
|
7 |
+
This repository contains the [Vicuna 7B model](https://huggingface.co/lmsys/vicuna-7b-delta-v0) quantised using [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa).
|
8 |
+
|
9 |
+
The original Vicuna 7B repository contains deltas rather than weights. Rather than merging the deltas myself, I used the model files from https://huggingface.co/helloollel/vicuna-7b.
|
10 |
+
|
11 |
+
## Provided files
|
12 |
+
|
13 |
+
Two model files are provided. You don't need both, choose the one you prefer.
|
14 |
+
|
15 |
+
Details of the files provided:
|
16 |
+
* `vicuna-7B-GPTQ-4bit-128g.pt`
|
17 |
+
* pt format file, created with the latest [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa) code.
|
18 |
+
* Command to create:
|
19 |
+
* `python3 llama.py vicuna-7B c4 --wbits 4 --true-sequential --act-order --groupsize 128 --save vicuna-7B-GPTQ-4bit-128g.pt`
|
20 |
+
* `vicuna-7B-GPTQ-4bit-128g.safetensors`
|
21 |
+
* newer `safetensors` format, with improved file security, created with the latest [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa) code.
|
22 |
+
* Command to create:
|
23 |
+
* `python3 llama.py vicuna-7B c4 --wbits 4 --true-sequential --act-order --groupsize 128 --save_safetensors vicuna-7B-GPTQ-4bit-128g.safetensors`
|
24 |
+
|
25 |
+
## How to run these GPTQ models in `text-generation-webui`
|
26 |
+
|
27 |
+
These model files were created with the latest GPTQ code, and require that the latest GPTQ-for-LLaMa is used inside the UI.
|
28 |
+
|
29 |
+
Here are the commands I used to clone the Triton branch of GPTQ-for-LLaMa, clone text-generation-webui, and install GPTQ into the UI:
|
30 |
+
```
|
31 |
+
git clone https://github.com/qwopqwop200/GPTQ-for-LLaMa
|
32 |
+
git clone https://github.com/oobabooga/text-generation-webui
|
33 |
+
mkdir -p text-generation-webui/repositories
|
34 |
+
ln -s GPTQ-for-LLaMa text-generation-webui/repositories/GPTQ-for-LLaMa
|
35 |
+
```
|
36 |
+
|
37 |
+
Then install this model into `text-generation-webui/models` and launch the UI as follows:
|
38 |
+
```
|
39 |
+
cd text-generation-webui
|
40 |
+
python server.py --model vicuna-7B-GPTQ-4bit-128g --wbits 4 --groupsize 128 # add any other command line args you want
|
41 |
+
```
|
42 |
+
|
43 |
+
The above commands assume you have installed all dependencies for GPTQ-for-LLaMa and text-generation-webui. Please see their respective repositories for further information.
|
44 |
+
|
45 |
+
If you are on Windows, or cannot use the Triton branch of GPTQ for any other reason, you can instead use the CUDA branch:
|
46 |
+
```
|
47 |
+
git clone https://github.com/qwopqwop200/GPTQ-for-LLaMa -b cuda
|
48 |
+
cd GPTQ-for-LLaMa
|
49 |
+
python setup_cuda.py install
|
50 |
+
```
|
51 |
+
Then link that into `text-generation-webui/repositories` as described above.
|
52 |
+
|
53 |
+
# Vicuna Model Card
|
54 |
+
|
55 |
+
## Model details
|
56 |
+
|
57 |
+
**Model type:**
|
58 |
+
Vicuna is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.
|
59 |
+
It is an auto-regressive language model, based on the transformer architecture.
|
60 |
+
|
61 |
+
**Model date:**
|
62 |
+
Vicuna was trained between March 2023 and April 2023.
|
63 |
+
|
64 |
+
**Organizations developing the model:**
|
65 |
+
The Vicuna team with members from UC Berkeley, CMU, Stanford, and UC San Diego.
|
66 |
+
|
67 |
+
**Paper or resources for more information:**
|
68 |
+
https://vicuna.lmsys.org/
|
69 |
+
|
70 |
+
**License:**
|
71 |
+
Apache License 2.0
|
72 |
+
|
73 |
+
**Where to send questions or comments about the model:**
|
74 |
+
https://github.com/lm-sys/FastChat/issues
|
75 |
+
|
76 |
+
## Intended use
|
77 |
+
**Primary intended uses:**
|
78 |
+
The primary use of Vicuna is research on large language models and chatbots.
|
79 |
+
|
80 |
+
**Primary intended users:**
|
81 |
+
The primary intended users of the model are researchers and hobbyists in natural language processing, machine learning, and artificial intelligence.
|
82 |
+
|
83 |
+
## Training dataset
|
84 |
+
70K conversations collected from ShareGPT.com.
|
85 |
+
|
86 |
+
## Evaluation dataset
|
87 |
+
A preliminary evaluation of the model quality is conducted by creating a set of 80 diverse questions and utilizing GPT-4 to judge the model outputs. See https://vicuna.lmsys.org/ for more details.
|