Kohya S. fa53f71ec0

fix: improve numerical stability by conditionally using float32 in Anima (#2302) ...

* fix: improve numerical stability by conditionally using float32 in block computations

* doc: update README for improvement stability for fp16 training on Anima in version 0.10.3

2026-04-02 12:36:29 +09:00

..

ipex

…

__init__.py

…

adafactor_fused.py

…

anima_models.py

fix: improve numerical stability by conditionally using float32 in Anima (#2302)

2026-04-02 12:36:29 +09:00

anima_train_utils.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

anima_utils.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

attention_processors.py

…

attention.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

chroma_models.py

…

config_util.py

Add skip_image_resolution to deduplicate multi-resolution dataset (#2273)

2026-03-19 08:43:39 +09:00

custom_offloading_utils.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

custom_train_functions.py

…

deepspeed_utils.py

fix: str is not "no" to str != "no"

2026-02-16 07:58:15 +09:00

device_utils.py

Merge branch 'sd3' into feat-hunyuan-image-2.1-inference

2025-09-13 20:13:58 +09:00

flux_models.py

…

flux_train_utils.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

flux_utils.py

…

fp8_optimization_utils.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

huggingface_util.py

…

hunyuan_image_models.py

feat: add multi backend attention and related update for HI2.1 models and scripts

2025-09-20 19:45:33 +09:00

hunyuan_image_modules.py

feat: add multi backend attention and related update for HI2.1 models and scripts

2025-09-20 19:45:33 +09:00

hunyuan_image_text_encoder.py

fix: --fp8_vl to work

2025-09-18 23:46:18 +09:00

hunyuan_image_utils.py

fix: remove print statement for guidance rescale in AdaptiveProjectedGuidance

2025-09-21 13:09:38 +09:00

hunyuan_image_vae.py

feat: add vae_chunk_size argument for memory-efficient VAE decoding and processing

2025-09-21 11:09:37 +09:00

hypernetwork.py

…

jpeg_xl_util.py

…

leco_train_util.py

Fix/leco cleanup (#2294)

2026-03-29 20:41:43 +09:00

lora_utils.py

feat: Support LoKr/LoHa for SDXL and Anima (#2275)

2026-02-23 22:09:00 +09:00

lpw_stable_diffusion.py

…

lumina_models.py

Fix bug and optimization Lumina training

2026-02-12 22:52:28 +07:00

lumina_train_util.py

Fix bug and optimization Lumina training

2026-02-12 22:52:28 +07:00

lumina_util.py

…

model_util.py

…

original_unet.py

…

qwen_image_autoencoder_kl.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

safetensors_utils.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

sai_model_spec.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

sd3_models.py

…

sd3_train_utils.py

…

sd3_utils.py

…

sdxl_lpw_stable_diffusion.py

…

sdxl_model_util.py

…

sdxl_original_control_net.py

…

sdxl_original_unet.py

…

sdxl_train_util.py

Merge branch 'dev' into sd3

2025-09-29 21:02:02 +09:00

slicing_vae.py

…

strategy_anima.py

Add/modify some implementation for anima (#2261)

2026-02-13 08:15:06 +09:00

strategy_base.py

feat: backward compatibility for SD/SDXL latent cache (#2276)

2026-02-23 21:44:51 +09:00

strategy_flux.py

…

strategy_hunyuan_image.py

…

strategy_lumina.py

…

strategy_sd3.py

…

strategy_sd.py

feat: backward compatibility for SD/SDXL latent cache (#2276)

2026-02-23 21:44:51 +09:00

strategy_sdxl.py

…

train_util.py

Fix/leco cleanup (#2294)

2026-03-29 20:41:43 +09:00

utils.py

…