Repository Issue Activity (beta)

#832 Ace-step - auto caption: .wav: 'NoneType' object has no attribute 'device'

#839 # Bug report — FLUX.2-klein-9B training fails in ai-toolkit (WSL2 / CUDA)

#1009 [Feature Request] Support LoRA training for MiniMax-H3 Ref2VA (Reference-to-Video-Audio)

#1008 Image grid behaving strangly for videos.

#806 Models are not downloading.

#826 Automagic v2 not working at all

#403 element 0 of tensors does not require grad and does not have a grad_fn

#1007 New Version is 14x slower then the one with the torchao error

#1003 80GB of virtual memory needed to train Klein 9B

#1013 Crash training video when offload active

#1012 torch.compile fails with record_stream after commit d184c6c

#1005 Ideogram4 model "Stop signal received; terminating process" while generating samples bug

#794 A possible solution to prevent learning process slowdown

#818 Pretrained LORA Bug

#994 torchcodec version 0.9.1 is no longer compatible with PyTorch version 2.13.0. I hope the author fixes this and also checks for potential issues with environment dependencies for other specified versions.

#991 H3 loRA fine-tuning but generated videos are pure noise

#1006 Just very tiny thing: "Cache Text Embeddings" Info balloon needs an update

#990 Severe VRAM regression / OOM with Krea 2 training after PyTorch 2.13 update

#1001 MiniMax-H3: official non-pruned transformer weights fail to load on 0.12.8

#798 diffusers issue

#820 [Docker] Container fails to initialize SQLite schema on first run — `main.Queue` does not exist

#821 Dataset not visible

#824 Ace Step 1.5 error on Batch size > 1 and slider training

#642 Sleepmode

#781 LTX-2.3 IC-LoRA