diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp16dp2ep32-latency-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp16dp2ep32-latency-agentic.yaml index d871a1c498..4e27959e2f 100644 --- a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp16dp2ep32-latency-agentic.yaml +++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp16dp2ep32-latency-agentic.yaml @@ -2,7 +2,7 @@ name: "kimik3-vllm-agg-h200-tp16dp2ep32-latency-agentic" model: path: "kimik3" - container: "vllm/vllm-openai:kimi-k3" + container: "vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1" precision: "fp4" identity: @@ -10,9 +10,9 @@ identity: repo: "moonshotai/Kimi-K3" revision: "9f62e4e9fffbd0a83ddd60e1c209d828994b3569" container: - image: "vllm/vllm-openai:kimi-k3" + image: "vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1" frameworks: - vllm: "0.1.dev19262+gb6bbf29dd.d20260727" + vllm: "0.29.0" dynamo: install: false diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-balanced-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-balanced-agentic.yaml index ba2c7f226d..b9530459d0 100644 --- a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-balanced-agentic.yaml +++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-balanced-agentic.yaml @@ -2,7 +2,7 @@ name: "kimik3-vllm-agg-h200-tp8dp4ep32-balanced-agentic" model: path: "kimik3" - container: "vllm/vllm-openai:kimi-k3" + container: "vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1" precision: "fp4" identity: @@ -10,9 +10,9 @@ identity: repo: "moonshotai/Kimi-K3" revision: "9f62e4e9fffbd0a83ddd60e1c209d828994b3569" container: - image: "vllm/vllm-openai:kimi-k3" + image: "vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1" frameworks: - vllm: "0.1.dev19262+gb6bbf29dd.d20260727" + vllm: "0.29.0" dynamo: install: false diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-vllm-simple-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-vllm-simple-agentic.yaml index 8f76304fbb..665566da95 100644 --- a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-vllm-simple-agentic.yaml +++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k3/agentic/agg-h200-tp8dp4ep32-vllm-simple-agentic.yaml @@ -2,7 +2,7 @@ name: "kimik3-vllm-agg-h200-tp8dp4ep32-vllm-simple-agentic" model: path: "kimik3" - container: "vllm/vllm-openai:kimi-k3" + container: "vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1" precision: "fp4" identity: @@ -10,9 +10,9 @@ identity: repo: "moonshotai/Kimi-K3" revision: "9f62e4e9fffbd0a83ddd60e1c209d828994b3569" container: - image: "vllm/vllm-openai:kimi-k3" + image: "vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1" frameworks: - vllm: "0.1.dev19262+gb6bbf29dd.d20260727" + vllm: "0.29.0" dynamo: install: false diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml index 1d0144da48..ce2cbd6a97 100644 --- a/configs/nvidia-master.yaml +++ b/configs/nvidia-master.yaml @@ -1392,7 +1392,7 @@ qwen3.5-fp4-b300-sglang-mtp: # cluster paths while srt-slurm owns allocation, direct vLLM startup, # readiness, and log capture. P/D disaggregation is intentionally deferred. kimik3-fp4-h200-vllm-agentic: - image: vllm/vllm-openai:kimi-k3 + image: vllm/vllm-openai:v0.29.0@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1 model: moonshotai/Kimi-K3 model-prefix: kimik3 runner: cluster:h200-dgxc