3467f8befb
[feat] Add end-to-end MMAudio V2A training, inference, and evaluation ( #1648 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 04:25:00 -07:00
Kun Lin and Claude Opus 5.5
7320d9d350
[feat] Add HunyuanVideo 1.5 embedding preprocessing pipeline ( #1663 )
...
Adds preprocess_hunyuan15_overfit.py, which writes VAE latents plus Qwen and ByT5 text embeddings in the dual-text parquet schema (pyarrow_schema_t2v_dual_text) that main's HunyuanVideo 1.5 training (#1662 ) reads; main's guide and overfit config pointed at this script, but nothing in-tree produced text_embedding_2 before.
Reconciled with #1662 during maintainer review (Swarm orders #723 and #777 ): main's schema and collate are the base. On top of them, both text streams now share one CFG-dropout decision per row also when _sample_index is absent, and a row missing the primary text_embedding raises instead of training unconditioned. The PR's widening of the shared t2v schema and its writer changes were dropped. docs/training/hunyuan15_overfit.md now describes the in-tree script.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 04:20:58 -07:00
af78c759f1
[new-model] Add Wan2.2-Animate-14B character animation/replacement inference ( #1765 )
...
Co-authored-by: SuhaanCommits <suhaan@cedar.build >
Co-authored-by: Claude Fable 5 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-10-07 04:19:37 -07:00
f82c190cbb
[kernel] Build + allow attn_qat_infer FP4 attention on sm_121a (DGX Spark) ( #1598 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 04:09:39 -07:00
e722023bd6
[perf] Wan kernel fusion: Triton-fused residual+LayerNorm+modulate inference path ( #1708 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 04:08:07 -07:00
bc0191d775
[perf] Harden Ulysses agreement and tune bounded H3 transfers ( #1807 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 04:01:53 -07:00
2d6e7d9b6b
[feat]: MiniMax-H3 encoder split: dedicated text-encoder node group for 720p on DGX Spark ( #1877 )
...
Co-authored-by: magicbear <magicbear@users.noreply.github.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:46:33 -07:00
Kevin Lin and Claude Opus 5.5
5e61accc37
[misc] Clean up QAD 5090 example inference scripts ( #1496 )
...
Split the FastWan-QAD Wan2.1-1.3B examples into per-recipe scripts (fp8, nvfp4_qat, nvfp4_sa2) with a shared _qad_common.py and a standalone multi-run benchmark harness, replacing FastWan_QAD_TAEHV.py.
The PR's fastvideo-kernel build changes (sm_121a arch probe and build.sh mapping) were dropped during maintainer review (Swarm order #664 ) in favour of #1598 , which owns the kernel arch policy; the RTX 5090 (sm_120a) recipes need nothing beyond main's kernels.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:44:49 -07:00
f60514183c
[ci] Separate performance behavior tests ( #1620 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:43:12 -07:00
William Lin and Claude Opus 5.5
156611bab3
[ci]: run the full attention test directory in the unit lane ( #1658 )
...
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:37:46 -07:00
William Lin and Claude Opus 5.5
6d6195a9f1
[feat]: fp8 PV mode for the FA4-FP4 attention path ( #1654 )
...
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:36:54 -07:00
8348e83e80
[feat] Import MiniMax H3 Comfy NVFP4-AWQ text encoder ( #1865 )
...
Co-authored-by: 武垚乐 <wuyaole@mininglamp.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:34:40 -07:00
2452022f15
[new-model] Add LTX-2.5 inference support ( #1704 )
...
Co-authored-by: Aryan Kumar <aryan5v@users.noreply.github.com >
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Co-authored-by: CodeRabbit <noreply@coderabbit.ai >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:34:18 -07:00
4b3f99b224
[new-model] Added LingBot-World-Fast image-to-video support ( #1665 )
...
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-10-07 03:32:38 -07:00
William Lin and Claude Opus 5.5
d967b99928
[bugfix]: registry ambiguity warning by specificity + QAD weights file path ( #1649 )
...
Registry: when several model detectors match a path, the first registered match still wins (main's pinned precedence, e.g. test_local_manifest_detectors_keep_first_match). Path specificity against each entry's registered HF paths now only decides whether to warn: the resolution is silent when the first match is strictly more specific than every other match (e.g. LTX-2.3 distilled directories that the LTX-2 base detector also claims through the shared pipeline class name), and warns otherwise, including when a later match is more specific than the one chosen. The PR originally re-ranked matches by specificity; that changed main's tested first-match resolution for local manifest directories, so it was narrowed during maintainer review (Swarm order #720 ).
Also fixes the QAD weights file path.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:30:55 -07:00
8f1eb2992c
[bugfix]: retain NVFP4 weights for LTX2 refinement LoRA ( #1705 )
...
Co-authored-by: Aryan Kumar <aryan5v@users.noreply.github.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:29:46 -07:00
pkisfaludi-nv and Claude Opus 5.5
683689de4f
[bugfix] Fix Ulysses sequence-parallel crash in self-forcing causal Wan DiT ( #1627 )
...
Retargeted onto fastvideo/models/wan/causal_transformer.py, where main moved the causal Wan implementation (fastvideo/models/dits/causal_wanvideo.py is now a compatibility shim), with the guarded flex-attention trim prepared during the maintainer review (Swarm order #657 ).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:29:22 -07:00
d5287f8352
[feat]: add MiniMax H3 Ref2VA and LoRA training support ( #1757 )
...
Co-authored-by: William Lin <8941107+SolitaryThinker@users.noreply.github.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:21:45 -07:00
cdfbd64b04
[perf] FP8: fuse the dynamic activation quantization into Triton kernels ( #1861 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:20:04 -07:00
Haochen Jiang and Claude Opus 5.5
299d0f6838
[bugfix] wrap the generic DenoisingStage in the inference_denoising profiler region ( #1707 )
...
profiler_region_inference_denoising was wrapped in exactly one place (minimax_h3_denoising.py), so every model using the generic DenoisingStage (Wan among them) exported a trace with no denoising region. Wrap the denoising loops in DenoisingStage and DmdDenoisingStage with profiler_region("inference_denoising"); the loop bodies are untouched.
These are the two classes STAGE_METRIC_MAP names directly. The perf harness also reports dit_time_s for every stage that inherits performance_component_metric, and those forward overrides (Cosmos*, LongCat*, Gen3C, GameCraft, HYWorld, MatrixGame2/3, Causal*, DreamXWorldAR, LingBot, GlmImage) remain outside the region for now. With no profiler configured the wrap is a no-op; SSIM is exactly 1.0 on FLASH_ATTN and TORCH_SDPA.
Caveat: the region span now appears in exported traces, but a pre-existing profiler behaviour (region exit toggles CUDA collection off, fastvideo/profiler.py) may leave GPU kernel rows out of the trace.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:19:42 -07:00
c18436125a
[misc]: Refactor separate training pipeline runner ( #1696 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:19:21 -07:00
5d515ee617
[perf]: remove per-step host syncs and repeated work from causal inference ( #1687 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:18:56 -07:00
37b9dc1836
[new-model] Add native Helios-Distilled T2V pipeline ( #1670 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 03:18:24 -07:00
a0291a57c1
[ci] Add Wan2.2-TI2V-5B SSIM regression test ( #1772 )
...
Co-authored-by: SuhaanCommits <suhaan@cedar.build >
Co-authored-by: Claude Fable 5 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-10-07 02:44:24 -07:00
6102fac00d
[bugfix]: fix modular ops activation checkpoint recomputation ( #1716 )
...
Co-authored-by: William Lin <8941107+SolitaryThinker@users.noreply.github.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:41:12 -07:00
1ad33c1832
[feat] Support training_cfg_rate > 0 for LTX-2 ( #1752 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:38:42 -07:00
da87973412
[ci] Clarify dashboard cohort grouping for legacy v1 records ( #1613 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:33:18 -07:00
William Lin and Claude Opus 5.5
a3fd1d0bab
[ci]: prevent direct tests from minting suite gates ( #1615 )
...
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:32:48 -07:00
3e87d6f4ae
[feat] ROCm: route the video sparse attention backends to their Triton kernels ( #1851 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:32:27 -07:00
4bb5627fa8
[feat] MiniMax H3: block-FP8 text encoder on Hopper, its converter, and a 4xH100 FastH3 profile ( #1829 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:31:52 -07:00
08ab8b5c00
[bugfix] ROCm: detect the platform from a HIP torch build when amdsmi is unavailable ( #1849 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:20:47 -07:00
fb5efebdab
[feat] Add HunyuanVideo 1.5 T2V training support ( #1662 )
...
Co-authored-by: William Lin <willlin@nvidia.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:19:22 -07:00
6e4813f8b4
[bugfix] use in-process executor for num_gpus=1 ( #1755 ) ( #1879 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:18:23 -07:00
13ef81e6da
[ci] Surface worker process logs on perf hard regressions ( #1701 )
...
Co-authored-by: Claude Fable 5 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-10-07 02:14:02 -07:00
13b3b36b8b
[feat] Add Infinite Livestream app ( #1878 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 02:07:37 -07:00
72795d0af8
[perf] avoid full-sequence materialization in VSA coarse/sparse combine ( #1813 )
...
Co-authored-by: Claude Opus 5 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-10-07 01:54:35 -07:00
af63a92f6a
[bugfix] Shut down Inductor workers gracefully ( #1786 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:54:03 -07:00
a4e7322ba6
[bugfix]: synchronize post-FSDP LoRA replicas ( #1666 )
...
Co-authored-by: Suckl <Suckl@users.noreply.github.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:53:30 -07:00
d6e13f9cc8
[feat] Add VQeval long-video quality metric ( #1727 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:39:11 -07:00
6c4c18690e
[bugfix] Enable SP for LongCat BSA refinement ( #1862 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:38:43 -07:00
a8688dddd9
[bugfix]: requantize MXFP8 weights after LoRA unmerge ( #1876 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:38:02 -07:00
cacc8cfcb3
[bugfix]: condition HunyuanVideo 1.5 i2v on the reference image ( #1693 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:36:35 -07:00
ef62ac48d1
[new-model] Add Wan2.2-S2V-14B speech-to-video (model port + checkpoint converter) ( #1683 )
...
Co-authored-by: SuhaanCommits <suhaan@cedar.build >
Co-authored-by: Claude Fable 5 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-10-07 01:36:01 -07:00
00cb7d0ba2
[ci] Add DGX Spark (GB10) single-GPU perf config + gpu_types gating ( #1679 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:30:42 -07:00
69a6215e7f
[feat] Add Dreamverse multimodal inputs and H3 routing ( #1835 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:30:26 -07:00
abe98e87f5
[feat] Add phase-aware modular training performance metrics ( #1826 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:30:11 -07:00
274b922d39
[misc]: surface GenerationResult.peak_memory_mb in the MiniMax H3 examples ( #1785 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:29:56 -07:00
38966056d0
[ci]: make merge-gate startup reliable ( #1762 )
...
Co-authored-by: William Lin <8941107+SolitaryThinker@users.noreply.github.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:29:42 -07:00
d8ca702dfd
[bugfix] registry: register magi_human presets ( #1751 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com >
2026-10-07 01:28:53 -07:00
William Lin
33d81730e9
[feat]: Kandinsky 6 Lite: register the 3.2B checkpoints and add docs + cookbook recipes ( #1927 )
2026-10-06 17:09:53 -07:00
