chris/MiniGPT-4
1
0
Fork 0
mirror of https://github.com/Vision-CAIR/MiniGPT-4.git synced 2025-04-05 10:30:45 +00:00
Code Issues Packages Projects Releases Wiki Activity

add guide to prepare vicuna

Browse Source
This commit is contained in:
Deyao Zhu 2023-04-18 16:01:24 +03:00
parent 8718057341
commit 307f0eeeba
2 changed files with 33 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
PrepareVicuna.md Normal file
View File

@ -0,0 +1,30 @@
## How to Prepare Vicuna Weight
Vicuna is an open-source LLAMA-based LLM that has a performance close to ChatGPT.
We currently use the v0 version of Vicuna-13B.
To prepare Vicuna’s weight, first download Vicuna’s **delta** weight from [https://huggingface.co/lmsys/vicuna-13b-delta-v0](https://huggingface.co/lmsys/vicuna-13b-delta-v0). In case you have git-lfs installed (https://git-lfs.com), this can be done by
```
git lfs install
git clone https://huggingface.co/lmsys/vicuna-13b-delta-v0
```
Note that this is not directly the working weight, but the difference between the working weight and the original weight of LLAMA-13B. (Due to LLAMA’s rules, we cannot distribute the weight of LLAMA.)
Then, you need to obtain the original LLAMA-13B weights in the HuggingFace format either following the instruction provided by HuggingFace [here](https://huggingface.co/docs/transformers/main/model_doc/llama) or from the Internet.
When these two weights are ready, we can use tools from Vicuna’s team to create the real working weight.
First, Install their library that is compatible with v0 Vicuna by
```
pip install git+https://github.com/huggingface/transformers@v0.1.10
```
Then, run the following command to create the final working weight
```
python -m fastchat.model.apply_delta --base /path/to/llama-13b-hf/ --target /path/to/save/working/vicuna/weight/ --delta /path/to/vicuna-13b-delta-v0/
```
Now you are good to go!

5
README.md
View File

@ -53,7 +53,8 @@ conda activate minigpt4
**2. Prepare the pretrained Vicuna weights**
The current version of MiniGPT-4 is built on the v0 versoin of Vicuna-13B.
Please refer to their instructions [here](https://huggingface.co/lmsys/vicuna-13b-delta-v0) to obtaining the weights.
Please refer to our instruction [here](PrepareVicuna.md)
to prepare the Vicuna weights.
The final weights would be in a single folder with the following structure:
```
@ -105,7 +106,7 @@ You can change the save path in the config file
torchrun --nproc-per-node NUM_GPU train.py --cfg-path train_configs/minigpt4_stage1_pretrain.yaml
```
**1. Second finetuning stage**
**2. Second finetuning stage**
In the second stage, we use a small high quality image-text pair dataset created by ourselves
and convert it to a conversation format to further align MiniGPT-4.