[CI][Bugfix] Fix `Hybrid SSM NixlConnector PD prefix cache test (2 GPUs)` (#47157)
Signed-off-by: NickLucche <nicolo.lucchesi@mistral.ai>
Nicolò Lucchesi committed
Jun 30, 2026 at 17:14 UTC
dc148dc4d7633f749409f4b724a25669fd455d3d
1 file changed
+2
tests/v1/kv_connector/nixl_integration/run_mamba_prefix_cache_test.sh
+2
@@ -51,6 +51,7 @@ vllm serve $MODEL \
51
--trust-remote-code \
52
--enable-prefix-caching \
53
--mamba-cache-mode all \
54
+ --attention-backend FLASHINFER \
55
--kv-transfer-config "$KV_CONFIG" &
56
57
# Start decode instance
@@ -68,6 +69,7 @@ vllm serve $MODEL \
69
--trust-remote-code \
70
--enable-prefix-caching \
71
--mamba-cache-mode all \
72
+ --attention-backend FLASHINFER \
73
--kv-transfer-config "$KV_CONFIG" &
74
75
echo "Waiting for prefill instance on port $PREFILL_PORT..."