[KV Connector] Fix DecodeBench fp8 fill values and add a startup fill mode (#58472)

Signed-off-by: Zijing Liu <liuzijing2014@gmail.com>
Signed-off-by: Yifan Qiao <yifanqiao@inferact.ai>
Co-authored-by: Yifan Qiao <yifanqiao@inferact.ai>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Zijing Liu
2026-09-24 23:55:47 +00:00
committed by GitHub
co-authored by Yifan Qiao Codex mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 71891bdb94
commit 7a9406690e
3 changed files with 549 additions and 38 deletions
+1 -1
View File
@@ -315,6 +315,6 @@ except Exception as e:
### Benchmarking
- To simulate the decode deployment of disaggregated serving, pass `--kv-transfer-config '{"kv_connector":"DecodeBenchConnector","kv_role":"kv_both"}'` to the `vllm serve` invocation. The connector populates KV cache with random values so decode can be profiled in isolation.
- To simulate the decode deployment of disaggregated serving, pass `--kv-transfer-config '{"kv_connector":"DecodeBenchConnector","kv_role":"kv_both"}'` to the `vllm serve` invocation. The connector populates KV cache with random values so decode can be profiled in isolation. Add `"kv_connector_extra_config":{"startup_fill":true}` to fill the whole KV cache once at startup instead of in the decode steps that admit new requests (see the connector docstring for its limitations).
- **CUDAGraph capture**: Use `--compilation_config '{"cudagraph_mode": "FULL_DECODE_ONLY"}'` to enable CUDA graph capture for decode only and save KV cache.