From a82882f392d95f0d26fb1ddcf3189472a01ffcff Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:48:26 -0400
Subject: [PATCH] [Klaud Cold] Delete benchmarks/multi_node/deprecated

Nothing resolves or sources these 12 retired multi-node scripts. Drop the
configuration-procedures link to the llmd-vllm wrapper (it pointed at a path
that no longer existed) and reword the ATOM agentic comment that named its
retired SGLang sibling.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 .../agentic/dsv4_fp4_mi355x_atom-disagg.sh    |  4 +-
 .../dsv4_fp4_gb200_llmd-vllm-disagg.sh        | 60 -------------
 .../deprecated/dsv4_fp4_mi355x_atom-disagg.sh | 83 ------------------
 .../dsv4_fp4_mi355x_sglang-disagg.sh          | 83 ------------------
 .../glm5_fp8_mi355x_sglang-disagg.sh          | 84 -------------------
 .../deprecated/gptoss_fp4_gb200_dynamo-trt.sh | 81 ------------------
 .../kimik2.5_fp4_mi355x_atom-disagg.sh        | 78 -----------------
 .../kimik2.5_fp4_mi355x_vllm-disagg.sh        | 80 ------------------
 .../minimaxm2.5_fp8_mi355x_vllm-disagg.sh     | 78 -----------------
 .../minimaxm3_fp4_mi355x_atom-disagg.sh       | 78 -----------------
 .../minimaxm3_fp4_mi355x_vllm-disagg.sh       | 78 -----------------
 .../minimaxm3_fp8_mi355x_atom-disagg.sh       | 78 -----------------
 .../minimaxm3_fp8_mi355x_vllm-disagg.sh       | 83 ------------------
 docs/configuration-procedures.md              |  2 +-
 docs/configuration-procedures_zh.md           |  2 +-
 15 files changed, 4 insertions(+), 948 deletions(-)
 delete mode 100755 benchmarks/multi_node/deprecated/dsv4_fp4_gb200_llmd-vllm-disagg.sh
 delete mode 100644 benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_atom-disagg.sh
 delete mode 100755 benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_sglang-disagg.sh
 delete mode 100755 benchmarks/multi_node/deprecated/glm5_fp8_mi355x_sglang-disagg.sh
 delete mode 100644 benchmarks/multi_node/deprecated/gptoss_fp4_gb200_dynamo-trt.sh
 delete mode 100644 benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_atom-disagg.sh
 delete mode 100755 benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_vllm-disagg.sh
 delete mode 100644 benchmarks/multi_node/deprecated/minimaxm2.5_fp8_mi355x_vllm-disagg.sh
 delete mode 100644 benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_atom-disagg.sh
 delete mode 100755 benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_vllm-disagg.sh
 delete mode 100644 benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_atom-disagg.sh
 delete mode 100644 benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_vllm-disagg.sh

diff --git a/benchmarks/multi_node/agentic/dsv4_fp4_mi355x_atom-disagg.sh b/benchmarks/multi_node/agentic/dsv4_fp4_mi355x_atom-disagg.sh
index 21a3a854c..931434e1a 100755
--- a/benchmarks/multi_node/agentic/dsv4_fp4_mi355x_atom-disagg.sh
+++ b/benchmarks/multi_node/agentic/dsv4_fp4_mi355x_atom-disagg.sh
@@ -3,8 +3,8 @@
 # Agentic trace-replay recipe for a disaggregated ATOM server on MI355X
 # (DeepSeek-V4-Pro FP4, 1P1D TP8), mooncake RDMA KV transfer + atomesh router.
 #
-# CI-style sibling of dsv4_fp4_mi355x_sglang-disagg.sh (same agentic trace
-# workload, same submit.sh path), but drives the ATOM engine instead of SGLang.
+# CI-style sibling of the former SGLang dsv4_fp4_mi355x_sglang-disagg.sh (same
+# agentic trace workload, same submit.sh path), but drives the ATOM engine.
 # Modeled on ATOM recipes/DeepSeek-V4-Agentic-PD-Max.md: three concurrency
 # tiers selected by the search space -- TP (conc 1-32), DP-attention
 # (conc 64-128, no offload), and DP-attention + CPU KV offload (conc 256,
diff --git a/benchmarks/multi_node/deprecated/dsv4_fp4_gb200_llmd-vllm-disagg.sh b/benchmarks/multi_node/deprecated/dsv4_fp4_gb200_llmd-vllm-disagg.sh
deleted file mode 100755
index f3eca3f27..000000000
--- a/benchmarks/multi_node/deprecated/dsv4_fp4_gb200_llmd-vllm-disagg.sh
+++ /dev/null
@@ -1,60 +0,0 @@
-#!/usr/bin/env bash
-#
-# Wrapper for the DeepSeek-V4-Pro GB200 llmd-vllm P/D disagg benchmark
-# (mid-curve 1P1D and high-tpt 2P1D). Sibling of gptoss_fp4_h200_llmd-vllm.sh -
-# same shape, different topology (GB200 = 4 GPUs/node, role spans 2 nodes;
-# H200 = 8 GPUs/node, role on a single node). The runner resolves this script via
-#   SCRIPT_NAME="${EXP_NAME%%_*}_${PRECISION}_gb200_llmd-vllm-disagg.sh"
-# from launch_gb200-nv.sh.
-
-set -euo pipefail
-
-source "$(dirname "$0")/../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    MODEL_PATH \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO
-
-if [[ -n "${SLURM_JOB_ID:-}" ]]; then
-    echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/llm-d" || exit 1
-
-# GB200 = 4 GPUs per node (Grace+Blackwell). The shared submit.sh
-# defaults GPUS_PER_NODE to 8, which is wrong for this SKU and would
-# overshoot DP_SIZE = nodes * 8.
-export GPUS_PER_NODE="${GPUS_PER_NODE:-4}"
-
-export TIME_LIMIT="${TIME_LIMIT:-08:00:00}"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-# Worker count per role (Option B multi-engine). Prefer an explicit
-# PREFILL_WORKERS/DECODE_WORKERS from the matrix additional-settings; else fall
-# back to the matrix num-worker fields (PREFILL_NUM_WORKERS/DECODE_NUM_WORKERS);
-# else 1 (single engine = unchanged 1P+1D / mid-curve). submit.sh reads these.
-export PREFILL_WORKERS="${PREFILL_WORKERS:-${PREFILL_NUM_WORKERS:-1}}"
-export DECODE_WORKERS="${DECODE_WORKERS:-${DECODE_NUM_WORKERS:-1}}"
-
-JOB_ID=$(bash ./submit.sh \
-    "$PREFILL_NODES" \
-    "$DECODE_NODES" \
-    "$ISL" "$OSL" "${CONC_LIST// /x}" inf \
-    "$RANDOM_RANGE_RATIO")
-
-if [[ -z "$JOB_ID" ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_atom-disagg.sh b/benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_atom-disagg.sh
deleted file mode 100644
index d17d1a323..000000000
--- a/benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_atom-disagg.sh
+++ /dev/null
@@ -1,83 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-# Use upstreamed multi_node scripts (no external clone needed)
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-# Set up SGL launch script-specific environment variables
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-export PREFILL_ENABLE_EP=false
-else
-export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-export PREFILL_ENABLE_DP=true
-else
-export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-export DECODE_ENABLE_EP=false
-else
-export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-export DECODE_ENABLE_DP=true
-else
-export DECODE_ENABLE_DP=false
-fi
-
-# Launch jobs based on ISL/OSL
-# Replace ' ' in CONC_LIST with 'x' such that the concurrency list is represented
-# by a list of numbers delimited by 'x'. This is because of how the underlying launch script
-# expects the concurrencies.
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO})
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_sglang-disagg.sh b/benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_sglang-disagg.sh
deleted file mode 100755
index d17d1a323..000000000
--- a/benchmarks/multi_node/deprecated/dsv4_fp4_mi355x_sglang-disagg.sh
+++ /dev/null
@@ -1,83 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-# Use upstreamed multi_node scripts (no external clone needed)
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-# Set up SGL launch script-specific environment variables
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-export PREFILL_ENABLE_EP=false
-else
-export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-export PREFILL_ENABLE_DP=true
-else
-export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-export DECODE_ENABLE_EP=false
-else
-export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-export DECODE_ENABLE_DP=true
-else
-export DECODE_ENABLE_DP=false
-fi
-
-# Launch jobs based on ISL/OSL
-# Replace ' ' in CONC_LIST with 'x' such that the concurrency list is represented
-# by a list of numbers delimited by 'x'. This is because of how the underlying launch script
-# expects the concurrencies.
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO})
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/glm5_fp8_mi355x_sglang-disagg.sh b/benchmarks/multi_node/deprecated/glm5_fp8_mi355x_sglang-disagg.sh
deleted file mode 100755
index d92dad750..000000000
--- a/benchmarks/multi_node/deprecated/glm5_fp8_mi355x_sglang-disagg.sh
+++ /dev/null
@@ -1,84 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-# Use upstreamed multi_node scripts (no external clone needed)
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-# Set up SGL launch script-specific environment variables
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-export PREFILL_ENABLE_EP=false
-else
-export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-export PREFILL_ENABLE_DP=true
-else
-export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-export DECODE_ENABLE_EP=false
-else
-export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-export DECODE_ENABLE_DP=true
-else
-export DECODE_ENABLE_DP=false
-fi
-
-# Launch jobs based on ISL/OSL
-# Replace ' ' in CONC_LIST with 'x' such that the concurrency list is represented
-# by a list of numbers delimited by 'x'. This is because of how the underlying launch script
-# expects the concurrencies.
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO} \
-    "${NODELIST:-}")
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/gptoss_fp4_gb200_dynamo-trt.sh b/benchmarks/multi_node/deprecated/gptoss_fp4_gb200_dynamo-trt.sh
deleted file mode 100644
index efa587bed..000000000
--- a/benchmarks/multi_node/deprecated/gptoss_fp4_gb200_dynamo-trt.sh
+++ /dev/null
@@ -1,81 +0,0 @@
-#!/usr/bin/bash
-
-set -x
-
-source "$(dirname "$0")/../../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_MAX_NUM_TOKENS \
-    PREFILL_MAX_BATCH_SIZE \
-    DECODE_MAX_NUM_TOKENS \
-    DECODE_MAX_BATCH_SIZE \
-    DECODE_GPU_MEM_FRACTION \
-    MODEL_PATH \
-    SERVED_MODEL_NAME \
-    RUNNER_NAME
-
-if [[ "$SPEC_DECODING" == "mtp" ]]; then
-    check_env_vars DECODE_MTP_SIZE
-else
-    DECODE_MTP_SIZE="0"
-fi
-
-PERFORMANCE_SWEEPS_PATH="components/backends/trtllm/performance_sweeps"
-
-echo "Cloning Dynamo repository..."
-git clone https://github.com/ai-dynamo/dynamo.git
-cd dynamo
-git checkout release/0.5.1-rc0.20260105
-git submodule update --init --recursive
-
-cd "$PERFORMANCE_SWEEPS_PATH"
-
-# Set up environment variables based on ISL/OSL
-if [ "$ISL" = "1024" ] && [ "$OSL" = "1024" ]; then
-    export CACHE_TRANSCEIVER_MAX_NUM_TOKENS=1024
-elif [ "$ISL" = "8192" ] && [ "$OSL" = "1024" ]; then
-    export CACHE_TRANSCEIVER_MAX_NUM_TOKENS=8448
-else
-    echo "Unsupported ISL/OSL combination: $ISL/$OSL"
-    exit 1
-fi
-
-kind=dynamo_disagg
-additional_slurm_args="--time=04:00:00"
-ntasks_per_node=4
-
-gen_nodes=$(((DECODE_TP + 3)/4 * DECODE_NUM_WORKERS))
-total_nodes=$((PREFILL_NUM_WORKERS + gen_nodes))
-total_tasks=$((total_nodes * ntasks_per_node))
-
-decode_eplb_num_slots=0
-
-sbatch --nodes=${total_nodes} \
-    --ntasks=${total_tasks} \
-    --ntasks-per-node=${ntasks_per_node} \
-    --job-name="${RUNNER_NAME}" \
-    --segment=${total_nodes} ${additional_slurm_args} \
-    benchmark_disagg.slurm \
-    ${PREFILL_NUM_WORKERS} ${PREFILL_TP} \
-    ${PREFILL_MAX_BATCH_SIZE} ${PREFILL_MAX_NUM_TOKENS} \
-    ${PREFILL_DP_ATTN} ${DECODE_NUM_WORKERS} \
-    ${DECODE_TP} ${DECODE_EP} ${DECODE_MAX_BATCH_SIZE} \
-    ${DECODE_MAX_NUM_TOKENS} ${DECODE_DP_ATTN} \
-    ${DECODE_GPU_MEM_FRACTION} ${decode_eplb_num_slots} \
-    ${DECODE_MTP_SIZE} "${CONC_LIST}" \
-    ${gen_nodes} ${kind} \
-    ${MODEL_PATH} ${SERVED_MODEL_NAME} \
-    ${IMAGE} ${ISL} ${OSL}
diff --git a/benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_atom-disagg.sh b/benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_atom-disagg.sh
deleted file mode 100644
index 1505b905d..000000000
--- a/benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_atom-disagg.sh
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-# Use upstreamed multi_node scripts (no external clone needed)
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-# Set up SGL launch script-specific environment variables
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-export PREFILL_ENABLE_EP=false
-else
-export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-export PREFILL_ENABLE_DP=true
-else
-export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-export DECODE_ENABLE_EP=false
-else
-export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-export DECODE_ENABLE_DP=true
-else
-export DECODE_ENABLE_DP=false
-fi
-
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO})
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_vllm-disagg.sh b/benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_vllm-disagg.sh
deleted file mode 100755
index d7995fb25..000000000
--- a/benchmarks/multi_node/deprecated/kimik2.5_fp4_mi355x_vllm-disagg.sh
+++ /dev/null
@@ -1,80 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-# Same EP/DP booleans as dsr1_fp8_mi355x_sglang-disagg.sh → amd_utils/submit.sh
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-    export PREFILL_ENABLE_EP=false
-else
-    export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-    export PREFILL_ENABLE_DP=true
-else
-    export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-    export DECODE_ENABLE_EP=false
-else
-    export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-    export DECODE_ENABLE_DP=true
-else
-    export DECODE_ENABLE_DP=false
-fi
-
-# Parameter order matches SGLang disagg submit.sh; arg 16 is optional NODELIST.
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO} \
-    "${NODELIST:-}")
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/minimaxm2.5_fp8_mi355x_vllm-disagg.sh b/benchmarks/multi_node/deprecated/minimaxm2.5_fp8_mi355x_vllm-disagg.sh
deleted file mode 100644
index a9a28d889..000000000
--- a/benchmarks/multi_node/deprecated/minimaxm2.5_fp8_mi355x_vllm-disagg.sh
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-    export PREFILL_ENABLE_EP=false
-else
-    export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-    export PREFILL_ENABLE_DP=true
-else
-    export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-    export DECODE_ENABLE_EP=false
-else
-    export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-    export DECODE_ENABLE_DP=true
-else
-    export DECODE_ENABLE_DP=false
-fi
-
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO} \
-    "${NODELIST:-}")
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_atom-disagg.sh b/benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_atom-disagg.sh
deleted file mode 100644
index e86e044a4..000000000
--- a/benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_atom-disagg.sh
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-# Use upstreamed multi_node scripts (no external clone needed)
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-# Set up SGL launch script-specific environment variables
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-export PREFILL_ENABLE_EP=false
-else
-export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-export PREFILL_ENABLE_DP=true
-else
-export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-export DECODE_ENABLE_EP=false
-else
-export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-export DECODE_ENABLE_DP=true
-else
-export DECODE_ENABLE_DP=false
-fi
-
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO})
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_vllm-disagg.sh b/benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_vllm-disagg.sh
deleted file mode 100755
index ac0f0a27d..000000000
--- a/benchmarks/multi_node/deprecated/minimaxm3_fp4_mi355x_vllm-disagg.sh
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-    export PREFILL_ENABLE_EP=false
-else
-    export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-    export PREFILL_ENABLE_DP=true
-else
-    export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-    export DECODE_ENABLE_EP=false
-else
-    export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-    export DECODE_ENABLE_DP=true
-else
-    export DECODE_ENABLE_DP=false
-fi
-
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO} \
-    "${NODE_LIST:-}")
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_atom-disagg.sh b/benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_atom-disagg.sh
deleted file mode 100644
index e86e044a4..000000000
--- a/benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_atom-disagg.sh
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-# Use upstreamed multi_node scripts (no external clone needed)
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-# Set up SGL launch script-specific environment variables
-export TIME_LIMIT="08:00:00"
-export MODEL_PATH=$MODEL_PATH
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-export PREFILL_ENABLE_EP=false
-else
-export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-export PREFILL_ENABLE_DP=true
-else
-export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-export DECODE_ENABLE_EP=false
-else
-export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-export DECODE_ENABLE_DP=true
-else
-export DECODE_ENABLE_DP=false
-fi
-
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO})
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_vllm-disagg.sh b/benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_vllm-disagg.sh
deleted file mode 100644
index 91628d632..000000000
--- a/benchmarks/multi_node/deprecated/minimaxm3_fp8_mi355x_vllm-disagg.sh
+++ /dev/null
@@ -1,83 +0,0 @@
-#!/usr/bin/env bash
-
-source "$(dirname "$0")/../../benchmark_lib.sh"
-
-check_env_vars \
-    CONC_LIST \
-    ISL \
-    OSL \
-    IMAGE \
-    SPEC_DECODING \
-    MODEL_PATH \
-    PREFILL_NUM_WORKERS \
-    PREFILL_TP \
-    PREFILL_EP \
-    PREFILL_DP_ATTN \
-    DECODE_NUM_WORKERS \
-    DECODE_TP \
-    DECODE_EP \
-    DECODE_DP_ATTN \
-    PREFILL_NODES \
-    DECODE_NODES \
-    RANDOM_RANGE_RATIO \
-    FRAMEWORK
-
-if [[ -n "$SLURM_JOB_ID" ]]; then
-  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
-fi
-
-set -x
-
-cd "$GITHUB_WORKSPACE/benchmarks/multi_node/amd_utils" || exit 1
-
-export TIME_LIMIT="08:00:00"
-# MiniMax-M3 MXFP8 (~414 GB) is pre-staged in this cluster's shared HF cache
-# (/it-share/hf-hub-cache/models--MiniMaxAI--MiniMax-M3-MXFP8), not the default
-# /it-share/data the launcher sets. Point the disagg model dir there for M3 only;
-# submit.sh exports MODEL_DIR=$MODEL_PATH and job.slurm resolves the snapshot under
-# it and bind-mounts MODEL_DIR into the prefill/decode serving containers.
-export MODEL_PATH=/it-share/hf-hub-cache
-export MODEL_NAME=$MODEL_NAME
-export CONTAINER_IMAGE=$IMAGE
-
-if [[ "${PREFILL_EP:-1}" -eq 1 ]]; then
-    export PREFILL_ENABLE_EP=false
-else
-    export PREFILL_ENABLE_EP=true
-fi
-
-if [[ "$PREFILL_DP_ATTN" == "true" ]]; then
-    export PREFILL_ENABLE_DP=true
-else
-    export PREFILL_ENABLE_DP=false
-fi
-
-if [[ "${DECODE_EP:-1}" -eq 1 ]]; then
-    export DECODE_ENABLE_EP=false
-else
-    export DECODE_ENABLE_EP=true
-fi
-
-if [[ "$DECODE_DP_ATTN" == "true" ]]; then
-    export DECODE_ENABLE_DP=true
-else
-    export DECODE_ENABLE_DP=false
-fi
-
-JOB_ID=$(bash ./submit.sh $PREFILL_NODES \
-    $PREFILL_NUM_WORKERS \
-    $DECODE_NODES \
-    $DECODE_NUM_WORKERS \
-    $ISL $OSL "${CONC_LIST// /x}" inf \
-    ${PREFILL_ENABLE_EP} ${PREFILL_ENABLE_DP} \
-    ${DECODE_ENABLE_EP} ${DECODE_ENABLE_DP} \
-    ${PREFILL_TP} ${DECODE_TP} \
-    ${RANDOM_RANGE_RATIO} \
-    "${NODELIST:-}")
-
-if [[ $? -ne 0 ]]; then
-    echo "Failed to submit job" >&2
-    exit 1
-fi
-
-echo "$JOB_ID"
diff --git a/docs/configuration-procedures.md b/docs/configuration-procedures.md
index 779db276b..9b4f8f2d3 100644
--- a/docs/configuration-procedures.md
+++ b/docs/configuration-procedures.md
@@ -220,7 +220,7 @@ Do not ship one side alone. `srtctl` reads the recipe, while matrix generation r
 
 ## Register an llm-d recipe
 
-Sources: [`benchmarks/llm-d/README.md`](../benchmarks/llm-d/README.md), [`benchmarks/multi_node/llm-d/README.md`](../benchmarks/multi_node/llm-d/README.md), [`llm-d-recipes/`](../benchmarks/multi_node/llm-d-recipes/), and the current [`llmd-vllm` benchmark wrapper](../benchmarks/multi_node/dsv4_fp4_gb200_llmd-vllm-disagg.sh).
+Sources: [`benchmarks/llm-d/README.md`](../benchmarks/llm-d/README.md), [`benchmarks/multi_node/llm-d/README.md`](../benchmarks/multi_node/llm-d/README.md), and [`llm-d-recipes/`](../benchmarks/multi_node/llm-d-recipes/).
 
 llm-d is not the srt-slurm path: InferenceX owns the Slurm allocation and starts one container per node.
 
diff --git a/docs/configuration-procedures_zh.md b/docs/configuration-procedures_zh.md
index 624307090..ae3f9a6ab 100644
--- a/docs/configuration-procedures_zh.md
+++ b/docs/configuration-procedures_zh.md
@@ -169,7 +169,7 @@ B200 Nscale 的 GLM-5.1 可用 `MODEL_PATH` 指定已有共享权重，覆盖默
 
 ## 注册 llm-d 配方
 
-来源：[`benchmarks/llm-d/README.md`](../benchmarks/llm-d/README.md)、[`benchmarks/multi_node/llm-d/README.md`](../benchmarks/multi_node/llm-d/README.md)、[`llm-d-recipes/`](../benchmarks/multi_node/llm-d-recipes/) 和当前 [`llmd-vllm` 基准 wrapper](../benchmarks/multi_node/dsv4_fp4_gb200_llmd-vllm-disagg.sh)。
+来源：[`benchmarks/llm-d/README.md`](../benchmarks/llm-d/README.md)、[`benchmarks/multi_node/llm-d/README.md`](../benchmarks/multi_node/llm-d/README.md)、和 [`llm-d-recipes/`](../benchmarks/multi_node/llm-d-recipes/)。
 
 llm-d 不是 srt-slurm 路径：InferenceX 自己持有 Slurm allocation，并在每个节点启动一个容器。