Zijie Tian ac1ccbceaa feat: add XAttention sparse policy integration ...

Integrate COMPASS XAttention algorithm into nano-vllm's CPU offload
execution path. Uses FlashAttention with native GQA support for
offload mode.

New files:
- nanovllm/kvcache/sparse/utils.py: find_blocks_chunked() utility
- nanovllm/kvcache/sparse/kernels.py: Triton kernels for XAttention
- nanovllm/kvcache/sparse/xattn.py: XAttentionPolicy implementation

Modified:
- nanovllm/config.py: Add XATTN configuration parameters
- nanovllm/engine/model_runner.py: Support XATTN policy
- nanovllm/kvcache/sparse/__init__.py: Register XAttentionPolicy
- tests/test_ruler.py: Add --sparse-policy parameter

Test results (32k ruler):
- NIAH tasks: 12/12 (100%)
- QA/Recall tasks: 11/15 (73%)
- Overall: 23/27 (85%)

Co-Authored-By: Claude <noreply@anthropic.com>

2026-01-14 10:04:46 +08:00

..

comm

[WIP] Added sgDMA operator for scatter kvcache communication.

2025-12-24 23:48:52 +08:00

debug

[refactor] Refactor the kvcache offload.

2026-01-04 19:37:03 +08:00

engine

feat: add XAttention sparse policy integration

2026-01-14 10:04:46 +08:00

kvcache

feat: add XAttention sparse policy integration

2026-01-14 10:04:46 +08:00

layers

🐛 fix: remove torch.compile from add_rms_forward to avoid recompilation

2026-01-14 07:02:02 +08:00

models

[tests] Added test_niah_standalone.py.

2026-01-12 00:16:37 +08:00

utils

[claudesquad] update from 'lw-offload-2' on 08 Jan 26 20:53 CST

2026-01-08 20:53:08 +08:00

__init__.py

better

2025-06-15 10:36:45 +08:00

config.py

feat: add XAttention sparse policy integration

2026-01-14 10:04:46 +08:00

llm.py

support tensor parallel

2025-06-15 01:31:24 +08:00

sampling_params.py

compile random sampling

2025-08-31 22:55:34 +08:00