A high-throughput and memory-efficient inference and serving engine for LLMs
[CI/Docs] Remove stale disagg prefill links (#44918)
Signed-off-by: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com>
Mohammad Miadh Angkad committed
Jun 9, 2026 at 18:05 UTC
fff9210b2a5f0acbfbd6e972c985e81f04ae17c8
1 file changed
-4
docs/features/disagg_prefill.md
-4
@@ -47,10 +47,6 @@ Now supports 9 types of connectors:
47
--kv-transfer-config '{"kv_connector":"FlexKVConnectorV1","kv_role":"kv_both"}'
48
```
49
50
-## Benchmarks
51
-
52
-Please refer to [benchmarks/disagg_benchmarks](../../benchmarks/disagg_benchmarks) for disaggregated prefilling benchmarks.
53
-
50
## Development
51
52
We implement disaggregated prefilling by running 2 vLLM instances. One for prefill (we call it prefill instance) and one for decode (we call it decode instance), and then use a connector to transfer the prefill KV caches and results from prefill instance to decode instance.