SolitaryThinker d81f1f166b [train] VSA-H3 DMD2 overfit: per-role backends, VSA metadata wiring, wandb guard ...

- MiniMaxH3Model: allow per-role attention backends (TORCH_SDPA/FLASH_ATTN/
  VIDEO_SPARSE_ATTN_H3); predict_noise routes attn_metadata_vsa for
  attn_kind='vsa' (dense roles keep None).
- MiniMaxH3DMDModel.prepare_batch builds real VSA-H3 metadata (packed
  text|cond|audio prefix segments via _h3_vsa_prefix_segments, sparsity from
  training.vsa.sparsity) when the role runs VIDEO_SPARSE_ATTN_H3.
- VSA-H3 tile(): fresh buffer for grad-tracking forwards; shared-buffer
  reuse trips autograd's in-place version check at student backward.
- build_tracker: degrade to DummyTracker when wandb is unimportable or has
  no credentials (modular stack already logs per-step loss dicts via the
  wandb tracker; no new callback needed).
- dmd2_vsa0_overfit.yaml: student VSA-H3 @ sparsity 0.0, teacher/critic
  FLASH_ATTN, 3-step DMD schedule, wandb project h3-dmd2-vsa.
- preprocess_minimax_h3_overfit: --video/--prompt/--output-dir/--model-path
  CLI; overfit_vsa0.sh: preprocess + 4-GPU torchrun with FASTVIDEO_FA4=1.
- CPU tests: VSA metadata build/routing, per-role backend override,
  wandb guard + loss-dict logging, tile-under-grad regression.

2026-08-09 14:46:58 -07:00

..

api

[feat] Add MiniMax H3 (#1674)

2026-08-04 13:54:39 -07:00

attention

[train] VSA-H3 DMD2 overfit: per-role backends, VSA metadata wiring, wandb guard

2026-08-09 14:46:58 -07:00

configs

[feat] VSA for MiniMax H3: packed mixed-modality sparse attention (FA4 CuTe sm10x), exempt/compete mask policies, gate-compress branch, per-request sparsity/schedule knobs, attention-mass probe

2026-08-09 12:39:48 -07:00

dataset

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

distributed

[bugfix]: free CUDA memory between train-framework model tests (LongCat OOM on L40S) (#1565)

2026-07-06 21:27:30 -07:00

entrypoints

[feat] VSA for MiniMax H3: packed mixed-modality sparse attention (FA4 CuTe sm10x), exempt/compete mask policies, gate-compress branch, per-request sparsity/schedule knobs, attention-mass probe

2026-08-09 12:39:48 -07:00

eval

[feat] World model training using third person games (#1443)

2026-07-10 05:09:07 +00:00

hooks

[perf] Mark LayerwiseOffloadHook entry points torch.compiler.disable (remove per-layer graph break) (#1365)

2026-05-20 03:09:23 +00:00

layers

[feat]: FA4-FP4 ATTN_QAT_INFER on sm_100/sm_103 + NVFP4 weight purge (#1647)

2026-07-27 15:46:52 -07:00

logging_utils

[Feature] Remove V1 folder (#642)

2025-07-24 22:43:12 -07:00

models

[feat] VSA for MiniMax H3: packed mixed-modality sparse attention (FA4 CuTe sm10x), exempt/compete mask policies, gate-compress branch, per-request sparsity/schedule knobs, attention-mass probe

2026-08-09 12:39:48 -07:00

performance

[ci] Add exact identity performance statuses (#1560)

2026-07-15 14:59:09 -07:00

performance_dashboard

[ci] Add exact identity performance statuses (#1560)

2026-07-15 14:59:09 -07:00

pipelines

[train] VSA-H3 DMD2 overfit: per-role backends, VSA metadata wiring, wandb guard

2026-08-09 14:46:58 -07:00

platforms

[misc] VSA-H3: /simplify pass — dedupe with the Wan VSA backend, drop dead state

2026-08-09 12:58:59 -07:00

tests

[train] VSA-H3 DMD2 overfit: per-role backends, VSA metadata wiring, wandb guard

2026-08-09 14:46:58 -07:00

third_party

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

train

[train] VSA-H3 DMD2 overfit: per-role backends, VSA metadata wiring, wandb guard

2026-08-09 14:46:58 -07:00

training

[feat] Add Kandinsky5 QAD training pipeline: data preprocessing, QAT finetune, QAT-aware DMD distillation (#1601)

2026-07-26 02:59:21 -07:00

worker

[bugfix]: skip unused output materialization (#1567)

2026-07-16 04:03:07 +00:00

workflow

[misc]: consolidate dataset download scripts under examples/datasets/ (#1667)

2026-08-02 14:21:24 -07:00

__init__.py

[feat] [4/n] Improve API: refactor sampling param and merge with presets (#1234)

2026-04-16 14:10:02 -07:00

AGENTS.md

[docs] add hierarchical AGENTS.md per-directory guidance (#1278)

2026-05-02 03:28:18 -07:00

envs.py

[attn] Make FA4 explicit opt-in via FASTVIDEO_FA4 and delete the runtime fallback machinery (#1540)

2026-07-04 14:51:00 -07:00

fastvideo_args.py

[refactor]: resolve attention backend once per component at load time (#1657)

2026-08-02 15:07:42 -07:00

forward_context.py

[feat] pre-commit support 120 col num (#1167)

2026-03-11 20:19:30 -07:00

image_processor.py

[feat] pre-commit support 120 col num (#1167)

2026-03-11 20:19:30 -07:00

logger.py

[feat] pre-commit support 120 col num (#1167)

2026-03-11 20:19:30 -07:00

profiler.py

[perf]: H3 torch.compile + CUDA graphs (1.2-1.3x) with denoising step marking (#1689)

2026-08-06 16:23:33 -07:00

registry.py

[misc] Support partial Hugging Face model downloads (#1684)

2026-08-06 16:09:46 -07:00

utils.py

[misc]: re-run yapf on main so pre-commit passes again (#1700)

2026-08-07 16:17:52 -07:00

version.py

[chore]: release v0.2.0 (#1432)

2026-06-04 14:22:49 -07:00