From 019ae525ca7800b87ebef81a398410951ee49aa5 Mon Sep 17 00:00:00 2001
From: Cam Quilici <cjquilici@gmail.com>
Date: Fri, 17 Apr 2026 16:59:24 -0500
Subject: [PATCH 1/2] Add B300 config: kimi-k2.5-int4-vllm

- Add kimik2.5-int4-b300-vllm benchmark config and the corresponding benchmarks/single_node/kimik2.5_int4_b300.sh launch script
- At the time of submission, the vLLM Kimi-K2.5 recipes page does not have a B300-specific recipe, so this reuses the existing Kimi-K2.5 INT4 B200 vLLM recipe as-is until B300-specific tuning is available
- Image: vllm/vllm-openai:v0.15.1 (same as B200), runner: b300, same TP=8 and concurrency 4-64 search-space as B200
---
 .github/configs/nvidia-master.yaml           | 21 +++++
 benchmarks/single_node/kimik2.5_int4_b300.sh | 81 ++++++++++++++++++++
 perf-changelog.yaml                          |  8 ++
 3 files changed, 110 insertions(+)
 create mode 100755 benchmarks/single_node/kimik2.5_int4_b300.sh

diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml
index d6202608d..0e31d7d85 100644
--- a/.github/configs/nvidia-master.yaml
+++ b/.github/configs/nvidia-master.yaml
@@ -2199,6 +2199,27 @@ kimik2.5-int4-b200-vllm:
     search-space:
     - { tp: 8, conc-start: 4, conc-end: 64 }
 
+# NOTE: At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html
+# does not have a B300-specific recipe, so this config reuses the existing
+# Kimi-K2.5 INT4 B200 vLLM recipe as-is until B300-specific tuning is available.
+kimik2.5-int4-b300-vllm:
+  image: vllm/vllm-openai:v0.19.0-cu130
+  model: moonshotai/Kimi-K2.5
+  model-prefix: kimik2.5
+  runner: b300
+  precision: int4
+  framework: vllm
+  multinode: false
+  seq-len-configs:
+  - isl: 1024
+    osl: 1024
+    search-space:
+    - { tp: 8, conc-start: 4, conc-end: 64 }
+  - isl: 8192
+    osl: 1024
+    search-space:
+    - { tp: 8, conc-start: 4, conc-end: 64 }
+
 kimik2.5-int4-h200-vllm:
   image: vllm/vllm-openai:v0.16.0
   model: moonshotai/Kimi-K2.5
diff --git a/benchmarks/single_node/kimik2.5_int4_b300.sh b/benchmarks/single_node/kimik2.5_int4_b300.sh
new file mode 100755
index 000000000..9f428f9e9
--- /dev/null
+++ b/benchmarks/single_node/kimik2.5_int4_b300.sh
@@ -0,0 +1,81 @@
+#!/usr/bin/env bash
+
+# NOTE: At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html
+# does not have a B300-specific recipe, so this script reuses the existing
+# Kimi-K2.5 INT4 B200 vLLM recipe as-is until B300-specific tuning is available.
+
+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 PYTHONNOUSERSITE=1
+export VLLM_USE_FLASHINFER_MOE_INT4=1
+
+SERVER_LOG=/workspace/server.log
+PORT=${PORT:-8888}
+
+if [ "${EVAL_ONLY}" = "true" ]; then
+    setup_eval_context
+    MAX_MODEL_LEN="$EVAL_MAX_MODEL_LEN"
+fi
+# Start GPU monitoring (power, temperature, clocks every second)
+start_gpu_monitor
+
+set -x
+vllm serve $MODEL --host 0.0.0.0 --port $PORT \
+--gpu-memory-utilization 0.95 \
+--tensor-parallel-size $TP \
+--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 \
+--no-enable-prefix-caching > $SERVER_LOG 2>&1 &
+
+SERVER_PID=$!
+
+# 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"
+    append_lm_eval_summary
+fi
+
+# Stop GPU monitoring
+stop_gpu_monitor
+set +x
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 75986a0eb..7b67be987 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -1479,6 +1479,14 @@
     - "At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html does not have a B300-specific recipe, so this reuses the existing Kimi-K2.5 FP4 B200 vLLM recipe as-is"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1056
 
+- config-keys:
+    - kimik2.5-int4-b300-vllm
+  description:
+    - "Add Kimi-K2.5 INT4 B300 vLLM benchmark"
+    - "Image: vllm/vllm-openai:v0.19.0-cu130"
+    - "At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html does not have a B300-specific recipe, so this reuses the existing Kimi-K2.5 INT4 B200 vLLM recipe as-is"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1057
+
 - config-keys:
     - gptoss-fp4-mi300x-vllm
   description:

From ce47b76574238b8672c7ea00ea23898f5e3c7459 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Mon, 20 Apr 2026 19:24:58 -0400
Subject: [PATCH 2/2] Drop --disable-log-requests from kimik2.5-int4-b300
 launch

The B300 script pins vllm/vllm-openai:v0.19.0-cu130, which removed the
--disable-log-requests CLI flag. Passing it causes the server to exit
immediately:

  vllm: error: unrecognized arguments: --disable-log-requests
  Server died before becoming healthy. Exiting.

Logging is off by default in newer vLLM releases, so removing the flag
is sufficient.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---
 benchmarks/single_node/kimik2.5_int4_b300.sh | 1 -
 1 file changed, 1 deletion(-)

diff --git a/benchmarks/single_node/kimik2.5_int4_b300.sh b/benchmarks/single_node/kimik2.5_int4_b300.sh
index 9f428f9e9..93644d879 100755
--- a/benchmarks/single_node/kimik2.5_int4_b300.sh
+++ b/benchmarks/single_node/kimik2.5_int4_b300.sh
@@ -47,7 +47,6 @@ vllm serve $MODEL --host 0.0.0.0 --port $PORT \
 --tool-call-parser kimi_k2 \
 --compilation_config.pass_config.fuse_allreduce_rms true \
 --trust-remote-code \
---disable-log-requests \
 --no-enable-prefix-caching > $SERVER_LOG 2>&1 &
 
 SERVER_PID=$!