Mihir Jagtap
df2cfbeefc
Update rotary_embedding.py
...
# For rotate_half style RoPE (used by HunyuanVideo, GameCraft),
# we need to expand cos/sin to full head_dim using repeat_interleave.
# The rotate_half operation works on consecutive PAIRS: (x0,x1), (x2,x3)...
# so cos/sin must be interleaved: [c0,c0,c1,c1,...] to match the pairing.
# Using torch.cat would produce [c0,c1,...,c0,c1,...] which is WRONG.
2026-02-08 21:23:13 -08:00
Mihir Jagtap
423a10afcc
Add GameCraft VAE (AutoencoderKLCausal3D) for Hunyuan-GameCraft port
...
- Add GameCraftVAE and EncoderCausal3D/DecoderCausal3D in gamecraftvae.py
- Add gamecraftvae_blocks.py with DownEncoderBlockCausal3D, UpDecoderBlockCausal3D,
and GameCraftVAEAttention (softmax upcast to float32 for bf16/fp16 stability)
- Add GameCraftVAEConfig with scaling_factor, temporal_compression_ratio=4,
mid_block_causal_attn=True, and block_out_channels layout
- Register AutoencoderKLCausal3D -> GameCraftVAE in models registry for checkpoint loading
2026-02-08 12:24:36 -08:00
Mihir Jagtap
db92ff7ba8
add the stages scripts [denoising and image encoding]
2026-02-08 12:20:45 -08:00
Mihir Jagtap
820cde2667
Update pipeline_batch_info.py
...
Camera/action control inputs (GameCraft)
2026-02-08 12:19:22 -08:00
Mihir Jagtap
5b196efb87
Update registry.py
...
Hunyuan (excludes gamecraft, hyworld, and versioned models)
2026-02-08 12:18:34 -08:00
Mihir Jagtap
938374a2e5
add conversion files full and text encoder
2026-02-08 12:16:12 -08:00
Mihir Jagtap
2285728de4
debug the vae and encoder tests
2026-02-08 12:13:41 -08:00
Mihir Jagtap
b7b11f2b4e
Create test_gamecraft_pipeline_parity.py
2026-02-07 14:11:35 -08:00
Mihir Jagtap
8efeee8da2
update rotary embedding and registry for gamecraft
2026-02-07 14:04:26 -08:00
Mihir Jagtap
66679d8ab3
add tests for text_encoders and vae
2026-02-07 13:59:47 -08:00
Mihir Jagtap
5744a23087
Create convert_gamecraft_vae.py
2026-02-07 13:53:35 -08:00
Mihir Jagtap
5e7caa8cb2
update the denoising stage
2026-02-07 13:52:17 -08:00
Mihir Jagtap
0c41d1dab5
update pipeline files
2026-02-07 13:50:38 -08:00
Mihir Jagtap
59e75f81d3
update sample files
2026-02-07 13:48:12 -08:00
Mihir Jagtap
cf452eea54
add config and pipeline code for gamecraft
2026-02-07 13:46:49 -08:00
Mihir Jagtap
ffc54c962f
add gamecraft vae blocks and script
2026-02-07 13:44:58 -08:00
Mihir Jagtap
2979b4681d
update the files (will be skipped)
2026-02-07 12:58:28 -08:00
Mihir Jagtap
2b4f3a231e
update the conversion script
2026-02-07 12:57:18 -08:00
Mihir Jagtap
b5f375faf6
Update registry.py
2026-02-07 12:56:03 -08:00
Mihir Jagtap
1b119943cd
fix(gamecraft): correct the parity test
2026-02-07 12:55:15 -08:00
Mihir Jagtap
3f30a69cce
fix(gamecraft): add GameCraftFinalLayer with correct shift/scale order
...
Introduce GameCraftFinalLayer to match official implementation (shift, scale)
2026-02-07 12:53:51 -08:00
Mihir Jagtap
0f3ff22190
Merge branch 'hao-ai-lab:main' into hunyuan-gamecraft-v2-port
2026-02-06 19:17:21 -08:00
XOR-op
f76efe798e
[bugfix]: _compile_conditions regression ( #1077 )
2026-02-06 18:57:43 -08:00
Zhang Peiyuan
09f455233e
[misc] readme small fix ( #1076 )
2026-02-06 16:57:48 -08:00
XOR-op
aea300f690
[perf]: use CUDA IPC in multiproc executor to avoid serialization overhead ( #1061 )
2026-02-06 19:43:18 -05:00
Hao Zhang
b92219f6a6
more fix and relocate STA arguments to pipeline config ( #1073 )
2026-02-06 13:38:35 -08:00
Mihir Jagtap
8b09b3de74
update the tests and add checkpoint conversion code
2026-02-06 08:54:18 -08:00
Mihir Jagtap
d0c6073074
update gamecraft pipeline code
2026-02-06 08:50:12 -08:00
Mihir Jagtap
debaa56f04
update gamecraft
2026-02-06 08:43:08 -08:00
Jinzhe Pan
a321b95a8a
[Fix] remove video ratio limitation ( #1069 )
2026-02-05 20:33:08 -08:00
Mihir Jagtap
c6df89a0de
Merge branch 'main' into hunyuan-gamecraft-v2-port
2026-02-06 09:47:33 +05:30
Mihir Jagtap
82f96eeb6b
fix merge conflicts
...
This branch has conflicts that must be resolved
Changes can be cleanly merged.
fastvideo/configs/pipelines/registry.py
fastvideo/pipelines/pipeline_registry.py
fastvideo/pipelines/stages/denoising.py
2026-02-05 20:15:40 -08:00
Mihir Jagtap
aac33e3876
update the denoising scripts
2026-02-05 19:37:20 -08:00
Mihir Jagtap
8ff3ab01a5
Update denoising.py
2026-02-05 19:32:35 -08:00
Mihir Jagtap
62f0866a46
Create gamecraft_denoising.py
2026-02-05 19:32:10 -08:00
MihirJagtap
1194120fd2
[feature] Add Hunyuan-GameCraft model support
...
Add support for Hunyuan-GameCraft video generation model with camera pose
conditioning via Plücker coordinates.
Components added:
- CameraNet module for camera pose encoding (fastvideo/models/cameranet.py)
- HunyuanGameCraftTransformer3DModel extending HunyuanVideo with CameraNet
- Model and pipeline configs with param_names_mapping for weight loading
- Weight conversion script for official PyTorch -> Diffusers safetensors
- Numerical parity test for DiT alignment validation
The implementation reuses existing HunyuanVideo components (VAE, text encoders,
transformer blocks) and adds CameraNet for game-specific camera control.
2026-02-05 19:00:38 -08:00
Wei Zhou
98308db7e0
[Feature] [Hy1.5] Support HY1.5 super-resolution pipeline for 1080p videos ( #1046 )
2026-02-05 16:39:20 -08:00
Hao Zhang
c1e18f6722
Some minor fixes ( #1068 )
2026-02-05 16:28:00 -08:00
William Lin
75e193a2c9
[core] Refactor and centralize our registry for models, pipelines, and sampling params ( #1066 )
2026-02-05 14:40:30 -08:00
XOR-op
d6e0a7d0dd
[refactor] Action module ( #1065 )
2026-02-05 13:56:35 -08:00
William Lin and gemini-code-assist[bot]
7fc5f241da
[misc] Fix naming instruction in runpod.md ( #1067 )
...
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2026-02-05 13:45:49 -08:00
Mingjia Huo
aae48a7e90
[feat] HYworld VAE with cache ( #1057 )
2026-02-05 04:18:17 -08:00
William Lin
88f38eb0f4
[misc] upgrade torch to 2.10 ( #1048 )
2026-02-05 04:15:22 -08:00
Shao Duan
d750b463dc
Added Sequence Parallelism for LTX-2 Distilled ( #1036 )
2026-02-04 15:25:30 -08:00
KyleShao and Will Lin
e10b26a3d8
[feat] Add Cosmos 2.5 I2W/V2W support (staged pipeline + examples) ( #1021 )
...
Co-authored-by: Will Lin <wlsaidhi@gmail.com >
2026-02-04 14:19:30 -08:00
XOR-op
74636ba246
[chore]: use higher precision timestamp in logging ( #1062 )
2026-02-04 13:22:59 -08:00
Wei Zhou
7e2f3f14e7
[Bugfix] [Wan I2V] Fix CLIP Image encoder config ( #1063 )
2026-02-04 13:20:19 -08:00
Kaiqin Kong
caa1c402ba
[bugfix] Double Normalization in Preprocessing Dataset ( #1055 )
2026-01-31 11:11:59 -08:00
XOR-op
38a6bd93d3
[chore]: update sageattn3 installation instructions ( #1050 )
2026-01-29 15:45:52 -08:00
alexzms
b867ef7e7c
[SP Sharding] Fix SP loss sharding on token axis (thw) with padding; add distributed correctness tests ( #1045 )
...
Fixes the sequence parallel sharding on t, now SP shards on t*h*w
2026-01-27 22:53:04 -08:00