-
Notifications
You must be signed in to change notification settings - Fork 11
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Problems with the Weight of resolution 256 stage2 #24
Comments
I did not make this mistake. |
1 similar comment
I did not make this mistake. |
Besides the error log, apparently the two ckpts of stage 2 have different files size under this link While the size of s2_256 is same with the stage 1 file And I have check the inner content, s2_256.pt contains numbers of transformer blocks, which is the stage 1 structure in the paper. s2_256 state_dict['module']
|
Could you try to see if the weight of 512 has this problem? |
The weight of 512 is fine. All other stages could be complete using the given ckpts except stage 2 of 256. |
Maybe it was the wrong version when I uploaded it, and now it seems there is no way to fix it, because I have left my job. You can retrain according to the training script. I'm sorry. |
The released stage 2 weight of resolution 256 seems to be incomplete, the error log is shown below.
File "/home/user/data/PT/PCDMs/stage2_batchtest_inpaint_model.py", line 126, in inference pose_proj.load_state_dict(pose_proj_dict)
RuntimeError: Error(s) in loading state_dict for ControlNetConditioningEmbedding: Missing key(s) in state_dict: "conv_in.weight", "conv_in.bias", "blocks.0.weight", "blocks.0.bias", "blocks.1.weight", "blocks.1.bias", "blocks.2.weight", "blocks.2.bias", "blocks.3.weight", "blocks.3.bias", "blocks.4.weight", "blocks.4.bias", "blocks.5.weight", "blocks.5.bias", "conv_out.weight", "conv_out.bias".
I also find the stage 2 .pt file is quite different between resolution 256 and 512.
Maybe its actually the stage 1 weight of 256?
Could you please help check this file or I wrongly process this step? thank you very much.
The text was updated successfully, but these errors were encountered: