hao-ai-lab-FastVideo

Files

T

Davids048andClaude Opus 5.5 61dffbe904 [refactor]: read typed config paths in the LTX-2, MiniMax-H3, Wan, MMAudio, MagiHuman, and Stable Audio pipelines

- LTX-2: the refine settings are read from pipeline.ltx2.refine.* and
  pipeline.components.upsampler_weights. The checkpoint's model_index.json
  defaults fill the fields that are still None, and the remaining switches
  get their defaults, in one recorded override that rebinds the pipeline's
  config. An explicit input value now always wins over the checkpoint
  default; the flat sentinels (the generic refine_* names and the class
  default of the step count) are gone.
- MiniMax-H3: the sequential load, decode backend, TAEH3, parallel VAE,
  offload, and checkpoint schedule settings are read from typed paths.
- Wan, MMAudio, MagiHuman, Stable Audio: flow_shift, dmd_denoising_steps,
  output_type, workload_type, offload, and autocast settings are read from
  typed paths. MagiHuman no longer reads the never-set log_level_progress
  name.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7

2026-10-03 06:06:25 +00:00

__init__.py

[refactor]: name the model-family config partitions *Options

2026-10-02 17:49:25 +00:00

compat.py

[refactor]: run the inference path on the resolved config

2026-10-03 05:35:14 +00:00

device_policy.py

[refactor]: run the inference path on the resolved config

2026-10-03 05:35:14 +00:00

errors.py

[feat] [1/n] API improvements: add intial files for new fastvideo public API (#1218 )