[BUG]: stable diffusion example

[AlvL1225]

🐛 Describe the bug

when installing stable diffusion environments, everything works fine until doing below: **

git clone https://github.com/Lightning-AI/lightning && cd lightning && git reset --hard b04a7aa pip install -r requirements.txt && pip install .

**

the requirements.txt in lightning will change lots of things(like changing torchmetrics==0.6.0 to torchmetrics>=0.7.0, <0.10.1)

Environment

No response

[1SAA]

Those specific versions of packages are required by Lightning. There may raise compatibility problem if you install a mismatched version.

[AlvL1225]

Hi,

I just follow the steps:

conda env create -f environment.yaml conda activate ldm

pip install colossalai==0.1.10+torch1.11cu11.3 -f https://release.colossalai.org

and

git clone https://github.com/Lightning-AI/lightning && cd lightning && git reset --hard b04a7aa pip install -r requirements.txt && pip install .

but still get version mismatch

[AlvL1225]

Those specific versions of packages are required by Lightning. There may raise compatibility problem if you install a mismatched version.

pytorch-lightning==1.4.2 and lightning==1.9.0 both exist and their dependencies might conflict

[1SAA]

Hi @yli1994 ,

Thanks for your issue. I've already updated environment.yaml.

[AlvL1225]

Hi @yli1994 ,

Thanks for your issue. I've already updated environment.yaml.

Hi @1SAA

I tried default train_colossalai.yaml config and change batch size to 32(default 64 in yaml) but still get CUDA memory error(4 x 80G A100)

[1SAA]

Hi @yli1994,

could you show me more about your error and you environment. you can use the scripts in pytorch issues to get your environment. just open an issue in pytorch and you should know how to get your environment.

[binmakeswell]

We have updated a lot. This issue was closed due to inactivity. Thanks.