mirror of
https://github.com/oobabooga/text-generation-webui.git
synced 2024-11-27 01:59:14 +01:00
24 lines
923 B
Markdown
24 lines
923 B
Markdown
An alternative way of reducing the GPU memory usage of models is to use the `DeepSpeed ZeRO-3` optimization.
|
|
|
|
With this, I have been able to load a 6b model (GPT-J 6B) with less than 6GB of VRAM. The speed of text generation is very decent and much better than what would be accomplished with `--auto-devices --gpu-memory 6`.
|
|
|
|
As far as I know, DeepSpeed is only available for Linux at the moment.
|
|
|
|
### How to use it
|
|
|
|
1. Install DeepSpeed:
|
|
|
|
```
|
|
conda install -c conda-forge mpi4py mpich
|
|
pip install -U deepspeed
|
|
```
|
|
|
|
2. Start the web UI replacing `python` with `deepspeed --num_gpus=1` and adding the `--deepspeed` flag. Example:
|
|
|
|
```
|
|
deepspeed --num_gpus=1 server.py --deepspeed --chat --model gpt-j-6B
|
|
```
|
|
|
|
### Learn more
|
|
|
|
For more information, check out [this comment](https://github.com/oobabooga/text-generation-webui/issues/40#issuecomment-1412038622) by 81300, who came up with the DeepSpeed support in this web UI. |