Converting stable diffusion 2.0's ckpt file to diffusers' format

[larme]

Describe the bug

Cannot convert stable diffusion 2.0's ckpt file into diffusers format using convert_original_stable_diffusion_to_diffusers.py

Reproduction

Assuming you have used the stable diffusion 2.0 pipeline before, you can just run the following commands to reproduce the error:

python convert_original_stable_diffusion_to_diffusers.py \
       --checkpoint_path ~/.cache/huggingface/hub/models--stabilityai--stable-diffusion-2/snapshots/1e128c8891e52218b74cde8f26dbfc701cb99d79/768-v-ema.ckpt \
       --original_config_file ~/codes/stablediffusion/configs/stable-diffusion/v2-inference.yaml \
       --dump_path ~/tmp/sd2 \
       --prediction_type v_prediction \
       --image_size 768

where ~/codes/stablediffusion is a copy of https://github.com/stability-AI/stablediffusion/

Logs

the entire log is here:

https://gist.github.com/larme/f348b35780da3d2f763d66abc2251433

System Info

• diffusers version: 0.18.0.dev0
• Platform: Linux-6.1.0-9-amd64-x86_64-with-glibc2.35
• Python version: 3.10.6
• PyTorch version (GPU?): 2.0.1+cu117 (True)
• Huggingface_hub version: 0.16.4
• Transformers version: 4.30.2
• Accelerate version: not installed
• xFormers version: not installed
• Using GPU in script?: no
• Using distributed or parallel set-up in script?: no

Who can help?

@patrickvonplaten @say

[cmdr2]

I can reproduce this as well, seems to be from this change: bc9a8ce#diff-0613bbbcc958a37615e7638e09307201784c90c2661611f745fe4f67f68f51efL863-L869

[patrickvonplaten]

Thanks for the issue, looking into it!

[patrickvonplaten]

Hey @cmdr2,

Very sorry about this, you were 100% correct. I've corrected everything and just tested in with #4041.

Please let me know if it isn't fixed.

[cmdr2]

Thanks @patrickvonplaten for the quick fix!

Tested that it works well for me now.

No problem, and sorry I didn't feel confident enough about my changes to propose a patch.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.