Davids048
a520eceed7
document vendored reward runtimes in agent memory
2026-06-08 20:22:13 +00:00
Davids048
b2c75534ac
fix mypy issues for vendored reward runtimes
2026-06-08 20:07:46 +00:00
Davids048
0b9c325984
fix mypy issues in RL integration files
2026-06-08 19:54:35 +00:00
Davids048
2afa837d96
handle yapf and ruff issues
2026-06-08 19:22:11 +00:00
Davids048
83bb602836
apply yapf formatting to reward runtimes
2026-06-06 20:59:50 +00:00
Davids048
606b42b6f9
fix spelling and markdown pre-commit issues
2026-06-06 20:58:49 +00:00
Davids048
e9a7a128c2
patch: resolve VideoAlign checkpoint and FA2 fallback
...
Resolve the default VideoAlign checkpoint path by downloading the KlingTeam/VideoReward Hugging Face snapshot and passing the returned local snapshot directory into VideoVLMRewardInference. Explicit checkpoint_path values still work as local path overrides.
Make the VideoAlign FlashAttention fallback check for classic FlashAttention-2 metadata/API instead of only checking for a flash_attn namespace. FastVideo may have FlashAttention-4/CuTe installed, but Transformers' flash_attention_2 path requires classic FlashAttention-2. When classic FA2 is unavailable, warn and use SDPA for the VideoAlign reward model.
2026-06-05 08:54:37 +00:00
Davids048
78b6995e13
WIP: vendor HPSv3 and VideoAlign reward runtimes
...
Adds vendored runtime code under fastvideo/train/methods/rl/reward/HPSv3 and fastvideo/train/methods/rl/reward/VideoAlign, replacing the broken gitlinks with normal tracked files.
The provenance and porting rules live in the vendor package markers: fastvideo/train/methods/rl/reward/HPSv3/__init__.py, fastvideo/train/methods/rl/reward/HPSv3/hpsv3/__init__.py, and fastvideo/train/methods/rl/reward/VideoAlign/__init__.py.
The FastVideo wrapper scripts hpsv3.py and videoalign.py depend on these vendored packages through explicit package imports. The copied runtime files are ported faithfully from upstream, with only import-path changes needed for package importability; this means some third-party code is unrelated to FastVideo internals but remains faithful to the source.
Known follow-ups: this vendor code does not currently pass pre-commit. VideoAlign also still assumes the checkpoint artifacts already exist under its checkpoints path or the configured VIDEOALIGN_CHECKPOINT_PATH; the missing/downloaded checkpoint resolution is not addressed in this commit.
2026-06-05 07:34:58 +00:00
Davids048
d858270708
patch formatting based on ruff hints
2026-06-05 01:46:09 +00:00
Davids048
298845ccd9
apply pre-commit formatting
2026-06-05 01:45:03 +00:00
Adam
9c2fa5aa9d
[feat] GenRL: add explicit HPSv3 VideoAlign recipes ( #1405 )
2026-06-05 01:14:43 +00:00
Adam
73e0fa727f
[feat] GenRL: add Wan LoRA adapter support ( #1404 )
2026-06-05 01:14:43 +00:00
Adam and Davids048
1fbe5a868c
[feat] GenRL: fix PPO loop cadence and diagnostics ( #1403 )
...
Co-authored-by: Davids048 <jundasu@ucsd.edu >
2026-06-05 01:14:43 +00:00
Adam and Davids048
e1be068c46
[feat] GenRL: add runtime and memory stability helpers ( #1402 )
...
Co-authored-by: Davids048 <jundasu@ucsd.edu >
2026-06-05 01:14:43 +00:00
Adam
cd69575095
[feat] GenRL: keep repeated prompt samples on one rank ( #1401 )
2026-06-05 01:14:43 +00:00
Adam and Davids048
2c19cbdebd
[feat] GenRL: stabilize reward model compatibility ( #1400 )
...
Co-authored-by: Davids048 <jundasu@ucsd.edu >
2026-06-05 01:14:43 +00:00
Peiyuan Zhang
5ae633a551
mv to utils
2026-06-05 01:14:42 +00:00
Peiyuan Zhang
6e663214a7
sampled video looks correct
2026-06-05 01:14:42 +00:00
Peiyuan Zhang
8b0ba24deb
time profiling and log sampled videos
2026-06-05 01:14:42 +00:00
Peiyuan Zhang
ef83574e44
gen RL runing
2026-06-05 01:14:14 +00:00
Peiyuan Zhang
9f2fc2f303
improve ema
2026-06-05 01:13:50 +00:00
Peiyuan Zhang
c52de7e9df
first edit
2026-06-05 01:07:44 +00:00
William Lin
c62a8514b0
[chore]: release v0.2.0 ( #1432 )
2026-06-04 14:22:49 -07:00
William Lin
3eb8081801
[chore]: unpin runtime deps in pyproject.toml ( #1431 )
2026-06-04 13:46:49 -07:00
Raghav K
3505d09564
[bugfix] tests: include ltx2_3_base in expected LTX2 preset set ( #1427 ) ( #1428 )
2026-06-04 12:17:18 -07:00
Shao Duan and SolitaryThinker
570607945c
[feat] dreamverse: sequence parallelism for serving ( #1424 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-06-01 23:13:49 -07:00
Kaiqin Kong and SolitaryThinker
d3a821cdcf
[feat] LoRA controls and integration for Dreamverse ( #1420 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-06-01 18:54:11 -07:00
Kaiqin Kong
3f24578139
[bugfix] LTX2: honor video_position_offset_sec in the DiT ( #1422 )
2026-06-01 16:59:15 -07:00
Kevin Lin
89fcf08378
[bugfix] Fix STFT dtype mismatch ( #1419 )
2026-05-31 20:56:07 -07:00
c2930b2aa1
[ci] Add additional Dreamverse UI tests ( #1417 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
2026-05-31 19:53:29 -07:00
KUAN-HAO HUANG and SolitaryThinker
019239690b
[perf] Add Adaptive Guidance (CFG gating) for stale-uncond reuse ( #1372 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-05-30 13:58:55 -07:00
alexzms
d6119c1f82
[bugfix]: dreamverse modal bypasses ENTRYPOINT — set ffmpeg env + key check ( #1413 )
2026-05-29 16:50:01 -07:00
alexzms
84214c80bb
[feat] LTX-2.3 audio: BWE vocoder path ( #1398 )
2026-05-29 16:43:26 -07:00
alexzms and mergify[bot]
c2d7143c72
[feat] LTX-2.3 transformer support (config-gated extension of LTX-2) ( #1397 )
...
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
2026-05-29 16:42:37 -07:00
Kaiqin Kong and SolitaryThinker
afdb6fbfa5
[feat] Add MatrixGame3.0 ( #1201 )
...
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-05-27 21:07:01 +00:00
alexzms
ba4c02d883
[docs]: highlight Dreamverse deployment paths + add Server B200 (SSH) guide ( #1409 )
2026-05-27 09:30:58 -07:00
Junda Su
2c137931f3
[bugfix] Fix Dreamverse Modal compile warmup latency ( #1394 )
2026-05-26 17:58:38 -07:00
Shao Duan
36682797a0
[feat] eval: input ergonomics + Evaluator features + bug fixes ( #1392 )
2026-05-26 13:59:45 -07:00
William Lin
0ef1357a77
[docs]: surface activation-trace utility in add-model skills ( #1399 )
2026-05-26 13:45:12 -07:00
William Lin
a75d19786a
[docs]: Wire activation trace into mkdocs nav + perf/troubleshooting ( #1304 )
2026-05-26 13:14:36 -07:00
alexzms
be548a78ea
[feat] VSA-256 fastpath on Blackwell via FA4 CuTe block-sparse attention ( #1354 )
2026-05-26 12:58:29 -07:00
alexzms
6a610e2bc9
[ci] add per-method single-step training tests for fastvideo.train ( #1343 )
2026-05-25 17:09:54 -07:00
Shao Duan and abaghyangor
321d5112b4
[refactor] eval: consolidate FVD into common.fvd, remove benchmarks/fvd ( #1380 )
...
Co-authored-by: abaghyangor <abaghyangor@gmail.com >
2026-05-24 12:09:47 -07:00
William Lin
ba75ad82db
[refactor]: shared attention infra additions for QAT-compat (Attn-QAT 5/12) ( #1383 )
2026-05-23 16:19:24 -07:00
Junda Su
58caa5109f
[ci] Add DreamVerse app CI tests ( #1386 )
2026-05-23 14:28:26 -07:00
2f3ca8aaad
[perf]: register FA2/FA3 default flash_attn_func as a torch.library custom op ( #1373 )
...
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com >
Co-authored-by: SolitaryThinker <wlsaidhi@gmail.com >
2026-05-23 00:39:06 -07:00
Junda Su
3a67319cb6
[infra] Use npm for Dreamverse web builds ( #1385 )
2026-05-22 18:50:10 -07:00
Junda Su
266fa044b3
[infra] Add Dreamverse Modal UI image build ( #1381 )
2026-05-22 11:36:53 -07:00
William Lin
f3398db868
chore: pin dreamverse npm deps to address Dependabot alerts ( #1359 )
2026-05-22 11:29:33 -07:00
fda02036bc
[feat]: Attn-QAT inference + training backends (deadcode) (Attn-QAT 4/12) ( #1358 )
...
Co-authored-by: jzhang38 <42993249+jzhang38@users.noreply.github.com >
Co-authored-by: RandNMR73 <99706358+RandNMR73@users.noreply.github.com >
2026-05-22 02:59:25 -07:00
Davids048