wxiwnd 01d6ceff8f fix(nsa): guard cp split to extend batches

2026-06-26 23:53:26 +08:00

..

test_conv_layer.py

ci: unify PR test suite naming (#21187)

2026-03-23 00:18:45 -07:00

test_deep_gemm_jit_cache.py

Reuse DeepGEMM JIT cache across restarts

2026-06-11 02:55:04 +08:00

test_mamba_state_scatter_triton.py

ci: unify PR test suite naming (#21187)

2026-03-23 00:18:45 -07:00

test_nsa_cp_utils.py

fix(nsa): guard cp split to extend batches

2026-06-26 23:53:26 +08:00

test_nsa_dequant_only_topk.py

Skip unused FP8 NSA K-cache dequant work

2026-06-10 07:00:43 +08:00

test_nsa_topk_transform.py

Protect CP shared-KV cache-hit correctness under batched FP8 reuse

2026-06-07 13:26:49 +08:00

test_nsa_transform_index.py

Expand prefill CP KV capacity by sharding persistent NSA KV

2026-04-26 04:11:17 +08:00