[NV] llm-d: enforce-eager on H200 prefill to skip cudagraph OOM

ezrasilvera · ezrasilvera · commit f284d7d33269 · 2026-06-05T13:27:55.000+03:00
Signed-off-by: Ezra Silvera &lt;ezra@il.ibm.com&gt;
diff --git a/benchmarks/multi_node/llm-d-recipes/dsr1-fp8-h200-1p1d-simple.yaml b/benchmarks/multi_node/llm-d-recipes/dsr1-fp8-h200-1p1d-simple.yaml
@@ -98,6 +98,7 @@ prefill:
     --max-model-len 16384
     --block-size 256
     --no-enable-prefix-caching
+    --enforce-eager
   env: {}
 
 decode:
