Aryan Kumar 91200ea61e [bugfix]: Keep model discovery answering for unresolved paths ...

Resolve the served family with allow_remote=False so /v1/model_info answers null instead of downloading a manifest and 500ing on an unresolved path. Drop the unreachable canvas clamp so both H3 families advertise one geometry, and pin that invariant in the tests.

2026-11-13 20:16:56 +00:00

..

server_contracts

[bugfix]: Keep model discovery answering for unresolved paths

2026-11-13 20:16:56 +00:00

apple_silicon_fast_mode.md

[bugfix] FastMetal-QAD MLX support: refuse CUDA QAD trees, use packed mlx_dit config, stream loads (#1736) (#1758)

2026-08-25 14:51:40 -07:00

inference_schema_parity_inventory.yaml

[perf]: training-free token reduction for MiniMax H3 (#2004)

2026-11-08 04:06:38 +00:00

overview.md

[refactor] Simplify Wan sampling and tests (#1825)

2026-09-07 15:57:19 -07:00

training_architecture.md

[docs] Migrate documentation and cookbook to Blume (#1969)

2026-10-16 16:09:04 -07:00