From 667569698096425e3bd6a7eb57dc7a2459d2f066 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 18 Apr 2026 18:25:27 -0400
Subject: [PATCH] Add MI355X config: glm5-fp4-sglang-mtp

New GLM-5 MXFP4 MI355X SGLang MTP benchmark. Uses the
lmsysorg/sglang-rocm:v0.5.10rc0-rocm700-mi35x-20260417 image against
the amd/GLM-5-MXFP4 model, with TP=8 conc 4-64 for 1k1k and 8k1k.

Unlike glm5-fp8-mi355x-sglang, the image already ships transformers
with glm_moe_dsa support, so no pip install -U transformers is needed
at benchmark time.

Launch flags per the request: --trust-remote-code, --tp $TP,
--chunked-prefill-size 131072, --disable-radix-cache,
--mem-fraction-static 0.85, --model-loader-extra-config
'{"enable_multithread_load": true}', --watchdog-timeout 1200,
--reasoning-parser glm45, --tool-call-parser glm47, plus EAGLE
speculative decoding (num-steps=3, eagle-topk=1, num-draft-tokens=4)
behind SGLANG_ENABLE_SPEC_V2=1. Client passes --use-chat-template as
required by AGENTS.md for MTP.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---
 .github/configs/amd-master.yaml               | 18 +++++
 benchmarks/single_node/glm5_fp4_mi355x_mtp.sh | 78 +++++++++++++++++++
 perf-changelog.yaml                           | 11 +++
 3 files changed, 107 insertions(+)
 create mode 100755 benchmarks/single_node/glm5_fp4_mi355x_mtp.sh

diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml
index 9e1f9834e..6a2b3e9a3 100644
--- a/.github/configs/amd-master.yaml
+++ b/.github/configs/amd-master.yaml
@@ -319,6 +319,24 @@ glm5-fp8-mi355x-sglang-mtp:
     search-space:
     - { tp: 8, conc-start: 4, conc-end: 64, spec-decoding: mtp }
 
+glm5-fp4-mi355x-sglang-mtp:
+  image: lmsysorg/sglang-rocm:v0.5.10rc0-rocm700-mi35x-20260422
+  model: amd/GLM-5-MXFP4
+  model-prefix: glm5
+  runner: mi355x
+  precision: fp4
+  framework: sglang
+  multinode: false
+  seq-len-configs:
+  - isl: 1024
+    osl: 1024
+    search-space:
+    - { tp: 8, conc-start: 4, conc-end: 64, spec-decoding: mtp }
+  - isl: 8192
+    osl: 1024
+    search-space:
+    - { tp: 8, conc-start: 4, conc-end: 64, spec-decoding: mtp }
+
 glm5-fp8-mi355x-atom:
   image: rocm/atom:rocm7.2.1-ubuntu24.04-pytorch2.9.1-atom0.1.2.post
   model: zai-org/GLM-5-FP8
diff --git a/benchmarks/single_node/glm5_fp4_mi355x_mtp.sh b/benchmarks/single_node/glm5_fp4_mi355x_mtp.sh
new file mode 100755
index 000000000..bdb2452ad
--- /dev/null
+++ b/benchmarks/single_node/glm5_fp4_mi355x_mtp.sh
@@ -0,0 +1,78 @@
+#!/usr/bin/env bash
+
+source "$(dirname "$0")/../benchmark_lib.sh"
+
+check_env_vars \
+    MODEL \
+    TP \
+    CONC \
+    ISL \
+    OSL \
+    RANDOM_RANGE_RATIO \
+    RESULT_FILENAME
+
+if [[ -n "$SLURM_JOB_ID" ]]; then
+  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
+fi
+
+hf download "$MODEL"
+
+export SGLANG_ENABLE_SPEC_V2=1
+
+SERVER_LOG=/workspace/server.log
+PORT=${PORT:-8888}
+
+EVAL_CONTEXT_ARGS=""
+if [ "${EVAL_ONLY}" = "true" ]; then
+    setup_eval_context
+    EVAL_CONTEXT_ARGS="--context-length $EVAL_MAX_MODEL_LEN"
+fi
+# Start GPU monitoring (power, temperature, clocks every second)
+start_gpu_monitor
+
+python3 -m sglang.launch_server \
+    --model-path $MODEL \
+    --host=0.0.0.0 \
+    --port $PORT \
+    --trust-remote-code \
+    --tp $TP \
+    --chunked-prefill-size 131072 \
+    --disable-radix-cache \
+    --mem-fraction-static 0.85 \
+    --model-loader-extra-config '{"enable_multithread_load": true}' \
+    --watchdog-timeout 1200 \
+    --reasoning-parser glm45 \
+    --tool-call-parser glm47 \
+    --speculative-algorithm EAGLE \
+    --speculative-num-steps 3 \
+    --speculative-eagle-topk 1 \
+    --speculative-num-draft-tokens 4 \
+    $EVAL_CONTEXT_ARGS > $SERVER_LOG 2>&1 &
+
+SERVER_PID=$!
+
+# Wait for server to be ready
+wait_for_server_ready --port "$PORT" --server-log "$SERVER_LOG" --server-pid "$SERVER_PID"
+
+run_benchmark_serving \
+    --model "$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 "$RESULT_FILENAME" \
+    --result-dir /workspace/ \
+    --use-chat-template
+
+# After throughput, run evaluation only if RUN_EVAL is true
+if [ "${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 4a539e2c0..835ba0451 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -1702,3 +1702,14 @@
   description:
     - "Add VLLM_FLOAT32_MATMUL_PRECISION=high"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1069
+
+- config-keys:
+    - glm5-fp4-mi355x-sglang-mtp
+  description:
+    - "Add GLM-5 MXFP4 MI355X SGLang MTP benchmark"
+    - "Image: lmsysorg/sglang-rocm:v0.5.10rc0-rocm700-mi35x-20260422"
+    - "Model: amd/GLM-5-MXFP4"
+    - "EAGLE speculative decoding (num-steps=3, eagle-topk=1, num-draft-tokens=4) behind SGLANG_ENABLE_SPEC_V2=1"
+    - "Image ships transformers with glm_moe_dsa support, so no extra pip install is needed (unlike glm5-fp8-mi355x-sglang)"
+    - "Configs: 1k1k and 8k1k, TP=8 conc 4-64 with spec-decoding=mtp"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXXX