Repository Issue Activity (beta)

#929 CUDA error: out of memory

#934 Why does Krea2-Turbo training automatically download the Qwen3-VL-4B model?

#612 Training z-image lora to start the task, always stuck at Step 0 of 3000

#806 Models are not downloading.

#905 offline work fails

#574 CUDA error: out of memory

#577 Samples view become from full width view, how to fix to grid view

#776 flux_train_ui fails on Gradio 6.x due to removed Image kwargs

#799 Stuck at Quantizing Qwen 3 when training Flux2 Klein Base 9B

#924 krea2 model download stoped - fetching transformer weights

#621 Error running job: database is locked

#504 Training speed degrades significantly with GPU power and temperature drop

#591 Font colours are bad

#925 Krea2 LoRA trained locally on Mac with AI-toolkit 0.10.19 has all lora_B tensors at zero

#907 Stuck in Loading Transformers - Flux Klein Base 9b

#45 Does it also support cloud?

#592 For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

#918 Is there any way to speed up the photo upload?

#921 Feature request: AceStep 1.5 XL audio LM context (lm_hints) for quality parity with ComfyUI

#735 No way to close the application

#872 Regression: unable to train Flux.2 D. OOM on mistral quantisation (32GB VRAM, 128GB RAM). Worked in 75781fb.

#656 Flux.2 klein base 4b/9b Edit mode training GPU has no load

#853 Bug Report: torch.histc called with Int tensor on MPS backend causes NotImplementedError in MoE models

#737 Dataset loader silently ignores unmatched control images when filenames differ

#857 # Bug Report: `load_te()` Ignores Local `text_encoder` Directory, Always Downloads from HuggingFace