ViperEkura 3fb4b8ab13 perf: use int32 paged KV indices ...

- store page-table, request-row, and cache-location indices as int32
- preserve CUDA graph replay with bit-exact logits and KV cache coverage
- improve B=1 decode latency by 1-6% across 1K-32K contexts on L20

2026-08-15 13:17:06 +08:00

..

conftest.py

fix: use max_context_len for stable num_splits in paged decode

2026-08-07 14:42:53 +08:00

test_backend_equivalence.py

perf: use int32 paged KV indices

2026-08-15 13:17:06 +08:00

test_backend.py

feat: propagate attention backend across scheduler threads

2026-08-09 13:32:40 +08:00

test_kernel_mask.py

fix: broadcast ref/old model state_dict for FSDP

2026-07-31 08:32:22 +08:00