[WIP] [NV] Kimi fp4 configs

.github/configs/nvidia-master.yaml

@@ -1873,6 +1873,31 @@ kimik2.5-int4-h200-vllm:
     search-space:
     - { tp: 8, conc-start: 4, conc-end: 64 }
 
+kimik2.5-fp4-b200-vllm:
+  image: vllm/vllm-openai:v0.16.0
+  model: nvidia/Kimi-K2.5-NVFP4
+  model-prefix: kimik2.5
+  runner: b200
+  precision: fp4
+  framework: vllm
+  multinode: false
+  seq-len-configs:
+  - isl: 1024
+    osl: 1024
+    search-space:
+    - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 }
+    - { tp: 4, ep: 4, conc-start: 4, conc-end: 64 }
+  - isl: 1024
+    osl: 8192
+    search-space:
+    - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 }
+    - { tp: 4, ep: 4, conc-start: 4, conc-end: 64 }
+  - isl: 8192
+    osl: 1024
+    search-space:
+    - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 }
+    - { tp: 4, ep: 4, conc-start: 4, conc-end: 64 }
+
 dsr1-fp8-b200-sglang-mtp:
   image: lmsysorg/sglang:v0.5.8-cu130-amd64
   model: deepseek-ai/DeepSeek-R1-0528

benchmarks/single_node/kimik2.5_fp4_b200.sh

@@ -0,0 +1,66 @@
+#!/usr/bin/env bash
+
+source "$(dirname "$0")/../benchmark_lib.sh"
+
+check_env_vars \
+    MODEL \
+    TP \
+    CONC \
+    ISL \
+    OSL \
+    MAX_MODEL_LEN \
+    RANDOM_RANGE_RATIO \
+    RESULT_FILENAME
+
+if [[ -n "$SLURM_JOB_ID" ]]; then
+  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
+fi
+
+hf download "$MODEL"
+
+nvidia-smi
+
+export TORCH_CUDA_ARCH_LIST="10.0"
+export PYTHONNOUSERSITE=1
+
+SERVER_LOG=/workspace/server.log
+PORT=${PORT:-8888}
+
+set -x
+vllm serve $MODEL --host 0.0.0.0 --port $PORT \
+--tensor-parallel-size=$TP \
+--gpu-memory-utilization 0.90 \
+--max-model-len $MAX_MODEL_LEN \
+--max-num-seqs $CONC \
+--reasoning-parser kimi_k2 \
+--tool-call-parser kimi_k2 \
+--compilation_config.pass_config.fuse_allreduce_rms true \
+--trust-remote-code \
+--disable-log-requests > $SERVER_LOG 2>&1 &
+
+SERVER_PID=$!
+
+# Wait for server to be ready
+wait_for_server_ready --port "$PORT" --server-log "$SERVER_LOG" --server-pid "$SERVER_PID"
+
+pip install -q datasets pandas
+
+run_benchmark_serving \
+    --model "$MODEL" \
+    --port "$PORT" \
+    --backend vllm \
+    --input-len "$ISL" \
+    --output-len "$OSL" \
+    --random-range-ratio "$RANDOM_RANGE_RATIO" \
+    --num-prompts $(( CONC * 10 )) \
+    --max-concurrency "$CONC" \
+    --result-filename "$RESULT_FILENAME" \
+    --result-dir /workspace/ \
+    --trust-remote-code
+
+# After throughput, run evaluation only if RUN_EVAL is true
+if [ "${RUN_EVAL}" = "true" ]; then
+    run_eval --framework lm-eval --port "$PORT" --concurrent-requests $CONC
+    append_lm_eval_summary
+fi
+set +x

perf-changelog.yaml

@@ -875,3 +875,10 @@
     - "Server: reasoning-parser qwen3, tool-call-parser qwen3_coder, enable-flashinfer-allreduce-fusion, mem-fraction-static 0.8"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/855
 
+- config-keys:
+    - kimik2.5-fp4-b200-vllm
+  description:
+    - "Add Kimi K2.5 NVFP4 vLLM benchmark configuration for B200"
+    - "Model: nvidia/Kimi-K2.5-NVFP4 with --reasoning-parser kimi_k2 and --trust-remote-code"
+    - "Image: vllm/vllm-openai:v0.16.0"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/862