SD V2 - shape mismatch: value tensor of shape [1024] cannot be broadcast to indexing result of shape [0, 768]

[shrutidhange]

Hi Team,
I am able to run the sd v2 related code using command

python main.py --name actor --base ./configs/perfusion_custom_sd_v2.yaml --basedir ./ckpt -t True --gpus 0,

the training runs fine but when I try to create the immages I get the following errors related to shape mismatch: value tensor of shape [1024] cannot be broadcast to indexing result of shape [0, 768]

I am using below command to generate images

python scripts/perfusion_txt2img.py --ddim_eta 0.0 --steps 50 --scale 6.0 --beta 0.7 --tau 0.15 --n_samples 4 --n_iter 1 --personalized_ckpt ./ckpt/2023-11-27T09-04-38_actor/models/last.ckpt --prompt "photo of actor"

Loading model from ./ckpt/v1-5-pruned-emaonly.ckpt and ./ckpt/2023-11-27T09-39-45_teddy/models/last.ckpt
Global Step: 840000
data: 0%| | 0/1 [00:00<?, ?it/s]
Sampling: 0%| | 0/1 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/scripts/perfusion_txt2img.py", line 357, in
main()
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/scripts/perfusion_txt2img.py", line 317, in main
encoding = model.cond_stage_model.encode(prompts, embedding_manager=model.embedding_manager)
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/ldm/modules/encoders/modules.py", line 285, in encode
return self(text, **kwargs)
File "/home/ubuntu/miniconda3/envs/aiml2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/ldm/modules/encoders/modules.py", line 280, in forward
z = self.transformer(input_ids=tokens, **kwargs)
File "/home/ubuntu/miniconda3/envs/aiml2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/ldm/modules/encoders/modules.py", line 259, in transformer_forward
return self.text_model(
File "/home/ubuntu/miniconda3/envs/aiml2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/ldm/modules/encoders/modules.py", line 219, in text_encoder_forward
hidden_states = self.embeddings(input_ids=input_ids, position_ids=position_ids,
File "/home/ubuntu/miniconda3/envs/aiml2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/ldm/modules/encoders/modules.py", line 143, in embedding_forward
inputs_embeds = embedding_manager(input_ids, inputs_embeds)
File "/home/ubuntu/miniconda3/envs/aiml2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/ubuntu/code/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization/perfusion/embedding_manager.py", line 117, in forward
embedded_text[tokenized_text == placeholder_token] = placeholder_embedding
RuntimeError: shape mismatch: value tensor of shape [1024] cannot be broadcast to indexing result of shape [0, 768]