New model: Anima - #1487

Merged

dxqb merged 57 commits into

Nerogar:mergefrom

Jul 4, 2026

Commits on Mar 24, 2026

torch 2.11
dxqb
committed
0c869b8

Commits on May 24, 2026

Upgrade transformers to 5.9 and huggingface-hub to 1.16

dxqb
and
claude
committed
271f51e

Commits on May 25, 2026

Merge branch 'master' into torch_2_11
dxqb
authored
5a8745c
Upgrade torch to 2.12.0 (CUDA 13.0 / ROCm 7.2)

dxqb
and
claude
committed
6913fc6
Switch nccl pin to cu13 (matches torch 2.12+cu130 dependency)

dxqb
and
claude
committed
6e4c2e5
Update triton-windows to 3.7.0.post26 (matches triton 3.7.0)

dxqb
and
claude
committed
b3c6462
Merge branch 'transformers5' into anima_base
dxqb
committed
8c10d1c

Commits on May 30, 2026

Merge branch 'upstream' into anima_base
dxqb
committed
477f940
anima: Anima model support (LoRA + Fine-Tune)
dxqb
committed
ea59c93

Commits on Jun 4, 2026

fix: call init_compile() in training thread to apply dynamo config

dxqb
and
claude
committed
7117833
Merge branch 'fix/compiled-optimizer-thread-local' into anima
dxqb
committed
67169e8
Merge branch 'master' into anima
dxqb
authored
03b7156

Commits on Jun 13, 2026

Add shared pipeline-dtype-copy helper to DtypeModelSaverMixin

dxqb
and
claude
committed
efd85f7
Generalize _copy_pipeline_to_dtype and adopt it in all model savers

dxqb
and
claude
committed
e0f4a72

Commits on Jun 17, 2026

Commits on Jun 19, 2026

Commits on Jun 29, 2026

Commits on Jul 1, 2026

Merge branch 'save-pipeline-dtype-util' into anima_base
dxqb
committed
e3f32ef
Simplify _copy_pipeline_to_dtype API: pass tokenizer objects directly
dxqb
committed
c1c8ae8
Merge branch 'save-pipeline-dtype-util' into anima_base
dxqb
committed
8cb37e7
Merge branch 'anima_base' into anima
dxqb
committed
cf5c965
Use shared _copy_pipeline_to_dtype in AnimaModelSaver
dxqb
committed
b2137e2
Merge branch 'merge' into lora-output-formats
dxqb
committed
cc43b19
Merge branch 'lora-output-formats' of origin into lora-output-formats
dxqb
committed
54a8c2c
Merge branch 'lora-output-formats' into anima_base
dxqb
committed
065ec03
Merge branch 'anima_base' into anima
dxqb
committed
a5f914e
Merge remote-tracking branch 'Nerogar/merge' into anima_base
dxqb
committed
7d87b24
Merge branch 'anima_base' into anima
dxqb
committed
a0c5cae
Bump diffusers pin and switch Anima LoRA preset to KOHYA_LORA output format
dxqb
committed
3df8a7c

Commits on Jul 2, 2026

Avoid N^2 base recompute in fused qkv LoRA/LoHa adapters

dxqb
and
claude
committed
3a59057
Merge branch 'lora-output-formats' into anima_base
dxqb
committed
a97c0de
Merge branch 'anima_base' into anima
dxqb
committed
f188523
Merge branch 'merge' into anima
dxqb
authored
9c2f7b8
Address self-review comments on lora-output-formats

dxqb
and
claude
committed
09ee0e3

Commits on Jul 3, 2026

LoRA loader: default LEGACY conversion to None instead of a generic guess

dxqb
and
claude
committed
6a13051
Merge branch 'lora-output-formats' into anima_base
dxqb
committed
824daad
Merge branch 'anima_base' into anima
dxqb
committed
8b52b3b
Address PR review comments on lora-output-formats
dxqb
committed
6faaa0e
Merge branch 'anima' of origin into anima
dxqb
committed
7ae73c7
Fix convert tool base-model handling and Flux2 LEGACY LoRA output regression
dxqb
committed
b4f1a92
Address PR #76 review comments on lora-output-formats

dxqb
and
claude
committed
8559988
Merge branch 'lora-output-formats' into anima_base
dxqb
committed
26e0b98
Merge branch 'anima_base' into anima
dxqb
committed
3fd9a2e
Fix ORIGINAL_TRANSFORMER allowlist missing Anima after lora-output-formats merge
dxqb
committed
097a4d1
Merge branch 'merge' of https://github.com/Nerogar/OneTrainer into anima
dxqb
committed
e03912b

Commits on Jul 4, 2026

Merge remote-tracking branch 'Nerogar/merge' into anima_base
dxqb
committed
ac2dad7
Merge branch 'anima_base' into anima
dxqb
committed
cf02ca4
Address anima_base PR #78 review comments

dxqb
and
claude
committed
f397a75