alpa-projects / alpa

Training and serving large-scale neural networks with auto parallelization.
https://alpa.ai
Apache License 2.0
3.07k stars 358 forks source link

Dreambooth fine-tune support #851

Open huanzhizun opened 1 year ago

huanzhizun commented 1 year ago

Can alpa support the fine-tuning of Dreambooth for faster training?

zhisbug commented 1 year ago

@huanzhizun Could you give more details about the Dreambooth implementation?

huanzhizun commented 1 year ago

https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth.py Above link is the dreambooth implementation. When I try to apply Alpa to it, I find that there are some problems along with JAX.