3 people f26d9f432c [Bugfix][GLM-5.3-Flash] Honor --mamba-ssm-cache-dtype for the KDA recurrent state (#59147) ...

Signed-off-by: Jared Wen <jaredwen@inferact.ai>
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

2026-10-11 17:55:00 +00:00

..

__init__.py

[Perf][GLM-5.3-Flash] Decode hot-path cleanups: strided KDA recurrent inputs, NoPE MQA query without concat, no duplicate router GEMM (#55736)

2026-09-10 16:28:33 +08:00

test_kda_recurrent.py

[Bugfix][GLM-5.3-Flash] Keep batch x heads out of gridDim.z in the GLM-5.3-Flash fused recurrent KDA kernel (#56974)

2026-10-05 09:31:45 +00:00

test_kda_state_dtype.py

[Bugfix][GLM-5.3-Flash] Honor --mamba-ssm-cache-dtype for the KDA recurrent state (#59147)

2026-10-11 17:55:00 +00:00

test_sequence_parallel.py

[Model] Use upstream GLM-5.3 and Qwen4-Exp configs and processor (#57387)

2026-10-02 02:25:18 +00:00

test_sparse_indexer_topk_dispatch.py

[ROCm][Perf][GLM-5.3-Flash] Add AITER topk backend for decodes (#58167)

2026-10-02 11:47:41 +01:00