kohya-ss/sd-scripts

Issues

• Has anyone already done finetuning a flux model（not lora)? Why is the image quality of the model generated by my finetune so poor?

  #1681 opened 2 months ago
  5

• Flux Results Not quite good after updating

  #1680 opened 2 months ago
  0

• Having NotImplementedError: Cannot copy out of meta tensor; no data! Issue

  #1679 opened 2 months ago
  4

• I have a few experiments I'd like to run training FLUX using weighted captions.

  #1678 opened 2 months ago
  1

• Having training issue with FLUX-schnell

  #1676 opened 2 months ago
  4

• Training not starting when path name includes spaces

  #1674 opened 2 months ago
  1

• flux finetune always fills shared memory

  #1673 opened 2 months ago
  5

• Failure with Flux ALL MERGED model !!

  #1672 opened 2 months ago
  0

• Cannot save Flux Dev checkpoint merged with LoRAs

  #1670 opened 2 months ago
  0

• Some flux fp8 models is returning an error, others don't why? Error "Cannot copy out of meta tensor..."

  #1669 opened 2 months ago
  1

• Not in train mode! error when resuming from schedulefree state

  #1667 opened 2 months ago
  1

• NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

  #1665 opened 2 months ago
  4

• Dataset Maker - Could not find TensorRT

  #1663 opened 2 months ago
  0

• Lycoris issue

  #1662 opened 2 months ago
  2

• The “Windows Installation” procedure in the readme is out of date and causes an error.

  #1659 opened 2 months ago
  4

• fatal: could not get a repository handle for submodule 'sd-scripts'

  #1658 opened 2 months ago
  1

• using '--alpha_mask' option seems not training text encoder

  #1657 opened 2 months ago
  1

• --blocks_to_swap is supported on linux?

  #1656 opened 2 months ago
  4

• The new WSD scheduler seems to have an issue.

  #1653 opened 2 months ago
  0

• ⚡Add Support for LoRA Training on Flux-Dev2Pro Models

  #1652 opened 2 months ago
  3

• "TypeError: get_cosine_schedule_with_warmup() got an unexpected keyword argument 'num_decay_steps'" Error occurs in dev branch

  #1651 opened 2 months ago
  1

• D:\AI\pinokio\api\fluxgym.git\env\lib\site-packages\torch\utils\checkpoint.py:1399: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead. [2024-09-28 15:15:24] [INFO] with device_autocast_ctx, torch.cpu.amp.autocast(**cpu_autocast_kwargs), recompute_context: # type: ignore[attr-defined] [2024-09-28 15:21:07] [INFO] D:\AI\pinokio\api\fluxgym.git\env\lib\site-packages\torch\utils\checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead. [2024-09-28 15:21:07] [INFO] with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]请问卡住不动怎么解决

  #1649 opened 2 months ago
  1

• cant sample with flux (fine-tuning)

  #1647 opened 2 months ago
  4

• Dataset maker sd1.5 - wd-tagger module stopped working on colab

  #1645 opened 2 months ago
  0

• Please add AdEMAMix to the scripts.

  #1644 opened 2 months ago
  2

• Regarding the Multi-Level Caption

  #1643 opened 2 months ago
  3

• about flux image normalization

  #1642 opened 2 months ago
  2

• Kohya_ss native trainer colab broken yesterday. Captured before/ after install log. Jax, Key array.

  #1641 opened 2 months ago
  10

• Was "Validation Loss" ever successfully implemented?

  #1638 opened 2 months ago
  0

• Specify blocks to Train w/ Finetuning?

  #1636 opened 2 months ago
  3

• Drop text encoder weights

  #1635 opened 2 months ago
  1

• FLUX LoRA Training - Text Encoder active training indicated when it should be disabled with --network_train_unet_only

  #1634 opened 2 months ago
  2

• FLUX: Bleeding?

  #1633 opened 2 months ago
  0

• Recommended settings for ScheduleFree.

  #1631 opened 2 months ago
  4

• Issue Reminder

  #1630 opened 2 months ago
  0

• Extract Flux Lora not working after 93/304

  #1629 opened 2 months ago
  0

• When I train with multiple GPUs, this error will be reported. This has been bothering me for a long time. How can I solve it

  #1627 opened 2 months ago
  1

• how to use flux full-finetune safetensors in diffusers?

  #1626 opened 2 months ago
  3

• Strange behavior with --network_args parameter

  #1624 opened 2 months ago
  0

• neither 'setup.py' nor 'pyproject.toml' found.

  #1623 opened 2 months ago
  0

• Inverted extraction for FLUX.

  #1622 opened 2 months ago
  4

• lora to original flux lora

  #1621 opened 2 months ago
  1

• TypeError: get_cosine_schedule_with_warmup() got an unexpected keyword argument 'num_decay_steps'

  #1618 opened 2 months ago
  6

• AttributeError: 'LycorisNetworkKohya' object has no attribute 'train_t5xxl'

  #1617 opened 2 months ago
  4

• ValueError: invalid literal for int() with base 10: 'still' when using facecrop

  #1616 opened 2 months ago
  1

• Issue for convert_flux_lora.py: RuntimeError: Some tensors share memory...

  #1614 opened 2 months ago
  5

• [Feature Request] Decouple `linear1` and `linear2` Flux layers in network_args

  #1613 opened 2 months ago
  8

• Flux merge tool does not properly handle LoRA keys from AI-Toolkit when merging w/ Flux.dev model - All keys "Unused" / Unchanged model file outputted

  #1612 opened 3 months ago
  9

• Does the FLUX/SD3 branch not work with SDXL yet?

  #1611 opened 3 months ago
  0

• FLUX LoRA Extraction Resemblance Issue

  #1610 opened 3 months ago
  1