lasted

Remove log

return total_second_length * 60 * (1.5 if use_teacache else 2.5) * (1 + ((steps - 25) / 100))

More time

self_attn_mask_1 = attention_mask.view(batch_size, 1, 1, seq_len).expand(-1, -1, seq_len, -1)

self_attn_mask_1 = attention_mask.view(batch_size, 1, 1, seq_len).expand(1, 1, seq_len, 1)

Original code

Closer

No worker_video_experimental

Move code

Less diff

All latents

Complete error message

Upload 4 files

Upload 3 files

Add advices

Make code closer

Prod version

New code again

Optimization

Write the code the same way

Complete code

Merge file creation

Fix

Increase seed

Make code closer

Introduce batch

torch.cat(

Remove is_last_frame

Merge image_encoder_last_hidden_state

Continue merge

Upload 4 files

transformers==4.43.0

Upload 3 files

Code from video

Fix name

Duplicate code

No logs

No yield

Add logs

Use worker instead

Upload 4 files

Upload 3 files

av==12.3.0

numpy==1.26.4

numpy==2.3.0

Restore

Only reverse indices

New example

Upload Example4.webp