[Docs] Add ECMooncakeConnector usage example for EPD (#57742)
Signed-off-by: jiangkuaixue123 <jiangxiaozhou111@163.com> Co-authored-by: Codex <noreply@openai.com>
jiangkuaixue123 committed
Sep 21, 2026 at 09:15 UTC
7ddd461a284d2c318410f29110f1604028b961ee
1 file changed
+22
-2
docs/features/disagg_encoder.md
+22
-2
@@ -32,8 +32,9 @@ Design doc: <https://docs.google.com/document/d/1aed8KtC6XkXtdoV87pWT0a8OJlZ-Cpn
32
33
## 2 Usage Example
34
35
-The current reference pathway is **ExampleConnector**.
36
-Below ready-to-run scripts shows the workflow:
35
+### ExampleConnector
36
+
37
+The following scripts demonstrate the workflow with **ExampleConnector**:
38
39
1 Encoder instance + 1 PD instance:
40
`examples/disaggregated/disaggregated_encoder/disagg_1e1pd_example.sh`
@@ -41,6 +42,25 @@ Below ready-to-run scripts shows the workflow:
42
1 Encoder instance + 1 Prefill instance + 1 Decode instance:
43
`examples/disaggregated/disaggregated_encoder/disagg_1e1p1d_example.sh`
44
45
+### ECMooncakeConnector
46
+
47
+**ECMooncakeConnector** transfers encoder outputs using the Mooncake TransferEngine.
48
+See the [Mooncake integration example](../../tests/v1/ec_connector/integration/run_epd_mooncake_ec_full_pipeline.sh)
49
+for a complete 1 Encoder instance + 1 PD instance setup, including the producer and
50
+consumer `--ec-transfer-config` settings and proxy configuration.
51
+
52
+With vLLM and Mooncake installed, run the example from the repository root:
53
+
54
+```bash
55
+GPU_E=0 GPU_PD=1 MOONCAKE_EC_PROTOCOL=tcp \
56
+ bash tests/v1/ec_connector/integration/run_epd_mooncake_ec_full_pipeline.sh
57
+```
58
+
59
+The script uses `Qwen/Qwen2.5-VL-3B-Instruct` by default, runs a single-GPU baseline,
60
+and compares the disaggregated outputs against it. Set `MODEL` to use another model
61
+or `MOONCAKE_EC_PROTOCOL=rdma` to use RDMA on supported hardware. See the script for
62
+additional configuration options.
63
+
64
---
65
66
## 3 Test Script