Ethanfel/ComfyUI-SelVA
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
78 Commits 6 Branches 0 Tags
28229d62ce95b51d45e9d6c6cf10b2ebd76a90e5
Commit Graph

8 Commits

Author SHA1 Message Date
Ethanfel f99d2666e8 fix: interpolate sync_cond to match audio sequence length in transformer 
Sync_MLP interpolates sync features based on video duration, but audio
latent length depends on the user-set audio duration. When video != audio
duration, the sequences diverge. Resample sync_cond to x's length before
the gated addition so any video/audio duration combo works.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-27 21:21:39 +01:00
Ethanfel 9b1cb71b2a fix: remove MMDiTWrapper import and dead code paths from factory.py 
MMDiTWrapper was removed from diffusion.py during cleanup but the import
in factory.py was missed, causing ImportError on every model load.
Also stub wavelet and diffusion_prior paths that reference deleted modules.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-27 19:12:40 +01:00
Ethanfel 8b634923dd fix: remove unused tqdm import from sampling.py 
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-27 18:01:29 +01:00
Ethanfel 87bea21d49 feat: extract prismaudio_core inference with callback-enabled sampling 
Add inference subpackage with:
- sampling.py: sample_discrete_euler modified from upstream to add callback
  parameter for ComfyUI progress bars (uses enumerate for step index)
- utils.py: set_audio_channels and prepare_audio for audio preprocessing

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-27 17:59:37 +01:00
Ethanfel 30e85f0f99 fix: resolve critical bugs and quality issues in prismaudio_core/models 
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-27 17:56:02 +01:00
Ethanfel 6e1186d5bd fix: clean up dead code paths and debug artifacts in prismaudio_core/models 
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-27 17:49:57 +01:00
Ethanfel 84c81e0e55 feat: extract prismaudio_core model modules (DiT, conditioners, VAE, diffusion) 
Fetch and adapt inference-critical model modules from upstream PrismAudio repo:
- dit.py: DiffusionTransformer with debug prints removed
- diffusion.py: ConditionedDiffusionModelWrapper, DiTWrapper, MMDiTWrapper
- conditioners.py: Cond_MLP, Sync_MLP, MultiConditioner with stubbed training imports
- autoencoders.py: AudioAutoencoder, OobleckEncoder/Decoder
- transformer.py: ContinuousTransformer, Attention with flash_attn fallback to SDPA
- blocks.py, utils.py, bottleneck.py, pretransforms.py, local_attention.py, pqmf.py
- adp.py: UNetCFG1d, UNet1d, NumberEmbedder
- mmmodules/model/low_level.py: MLP, ChannelLastConv1d, ConvMLP

All internal imports rewritten from PrismAudio.* to prismaudio_core.*,
training-only imports stubbed, flash_attn made optional with HAS_FLASH_ATTN flag.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-27 17:31:22 +01:00
Ethanfel b60ff4111b feat: extract prismaudio_core config and model factory 
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-27 17:05:57 +01:00