Files

T

SolitaryThinker 5b5608cb37 [docs] magi-human: Wave 8 broader CFG/preset/fallback audit; OQ-6 production resolved

Documents Wave 8 audit findings and targeted fixes in
tests/local_tests/magi-human.md:

- New subsection 'Wave 8 (2026-05-01)' under Numerical-alignment
  investigation: 4-item audit table (items #1, #3, #10, #7-stale),
  per-test parity numbers post-Wave-8, and OQ-6 status update.
- Phase 11 status table: pipeline parity row updated to reflect Wave 7.5
  real-prompt encoding and Wave 8 production-fix context.
- Last verified line updated to 2026-05-01 (Wave 8 broader audit + fixes).
- OQ-6 status updated to RESOLVED-PRODUCTION; bf16 amplification floor
  TRACKED separately.

Key conclusion: all production-facing root causes for OQ-6 are now
identified and fixed (incomplete neg prompt Wave 7, tokenizer pre-padding
Wave 8 #1, resolution defaults Wave 8 #3, silent audio fallback Wave 8
#10). Residual parity-test drift is the inherent bf16+CFG amplification
floor through the multistep FlowUniPC scheduler — not a code bug.

2026-05-05 10:55:49 -07:00

local_tests

[docs] magi-human: Wave 8 broader CFG/preset/fallback audit; OQ-6 production resolved

2026-05-05 10:55:49 -07:00

__init__.py

Added LTX-2 Distilled T2V Generation (#1016 )

2026-01-21 14:11:39 -08:00

README.md

[refactor] tests/local_tests: organize by model family (#1269 )

2026-05-01 01:49:54 -07:00

test_bsa.py

[Feature] Add BSA (Bidirectional Sparse Attention) inference backend (#1174 )

2026-04-05 05:00:33 +00:00

README.md

Tests

tests/local_tests/ holds local-only parity, smoke, and component tests organized by model family (e.g. tests/local_tests/ltx2/, tests/local_tests/stable_audio/). They typically require a checked-out reference repo and/or model weights and are skipped in CI; see each family's README.md for setup. The top-level tests/local_tests/README.md is the family index.
The CI-backed test suite still lives in fastvideo/tests/.
Eventually, all tests will move under tests/.