Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,191 @@
base:
schema: 2
name: minimaxm31flash-fp4-b300-sglang-agentic
resources:
gpu_type: b300
gpus_per_node: 8
frontend:
type: sglang
enable_multiple_frontends: false
observability:
enabled: false
tachometer:
enabled: false
engine: sglang
# Cold loads of the 250 GB checkpoint plus JIT MSA kernel builds and graph capture outlast the default.
health_check:
max_attempts: 1440
interval_seconds: 5
roles:
agg:
nodes: 1
workers: 1
gpus: 4
# Serve flags follow the cookbook's B300 cells at
# https://docs.sglang.io/cookbook/autoregressive/MiniMax/MiniMax-M3.1.
args:
served-model-name: MiniMaxAI/MiniMax-M3.1-Flash
trust-remote-code: true
reasoning-parser: minimax-m3
tool-call-parser: minimax-m3
# Seven-token DSpark, the bundled draft's dspark_block_size. Throughput runs
# switch to synthetic acceptance at the golden acceptance length.
speculative-algorithm: DSPARK
speculative-dspark-block-size: 7
speculative-use-block-verification: true
cuda-graph-backend-prefill: breakable
tokenizer-worker-num: 8
watchdog-timeout: 3600
enable-metrics: true
env:
PYTHONNOUSERSITE: '1'
PYTHONUNBUFFERED: '1'
SGLANG_TIMEOUT_KEEP_ALIVE: '1800'
FMHA_SM100_ALLOW_JIT: '1'
benchmark:
env:
# Warmup can leave bytes unacknowledged past AIPerf's 30 s default.
AIPERF_HTTP_TCP_USER_TIMEOUT: '900000'
AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'

# TP4 Balanced cell (TRT-LLM routed MoE). The scheduler batch is 2x CONC (AgentX
# subagent fan-out), capped at the cookbook's 128; the decode graph batch matches it.

override_tp4_c1:
roles:
agg:
args:
tensor-parallel-size: 4
moe-runner-backend: flashinfer_trtllm_routed
chunked-prefill-size: 32768
max-running-requests: 2
cuda-graph-max-bs-decode: 2
mem-fraction-static: 0.8
benchmark:
env:
CONC: '1'
KV_OFFLOADING: none

override_tp4_c4:
roles:
agg:
args:
tensor-parallel-size: 4
moe-runner-backend: flashinfer_trtllm_routed
chunked-prefill-size: 32768
max-running-requests: 8
cuda-graph-max-bs-decode: 8
mem-fraction-static: 0.8
benchmark:
env:
CONC: '4'
KV_OFFLOADING: none

override_tp4_c8:
roles:
agg:
args:
tensor-parallel-size: 4
moe-runner-backend: flashinfer_trtllm_routed
chunked-prefill-size: 32768
max-running-requests: 16
cuda-graph-max-bs-decode: 16
mem-fraction-static: 0.8
benchmark:
env:
CONC: '8'
KV_OFFLOADING: none

override_tp4_c16:
roles:
agg:
args:
tensor-parallel-size: 4
moe-runner-backend: flashinfer_trtllm_routed
chunked-prefill-size: 32768
max-running-requests: 32
cuda-graph-max-bs-decode: 32
mem-fraction-static: 0.8
benchmark:
env:
CONC: '16'
KV_OFFLOADING: none

override_tp4_c32:
roles:
agg:
args:
tensor-parallel-size: 4
moe-runner-backend: flashinfer_trtllm_routed
chunked-prefill-size: 32768
max-running-requests: 64
cuda-graph-max-bs-decode: 64
mem-fraction-static: 0.8
benchmark:
env:
CONC: '32'
KV_OFFLOADING: none

# High-Throughput cell: TP8 with attention DP2 and EP8 (MegaMoE). The scheduler
# batch is 2x CONC up to the cookbook's 256, and each of the two attention-DP
# ranks captures decode graphs for half of it.

override_tp8_ep8_c64:
roles:
agg:
gpus: 8
args:
tensor-parallel-size: 8
attn-dp-size: 2
enable-dp-lm-head: true
expert-parallel-size: 8
chunked-prefill-size: 65536
max-running-requests: 128
cuda-graph-max-bs-decode: 64
mem-fraction-static: 0.8
env:
SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: '32768'
benchmark:
env:
CONC: '64'
KV_OFFLOADING: none

override_tp8_ep8_c128:
roles:
agg:
gpus: 8
args:
tensor-parallel-size: 8
attn-dp-size: 2
enable-dp-lm-head: true
expert-parallel-size: 8
chunked-prefill-size: 65536
max-running-requests: 256
cuda-graph-max-bs-decode: 128
mem-fraction-static: 0.8
env:
SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: '32768'
benchmark:
env:
CONC: '128'
KV_OFFLOADING: none

override_tp8_ep8_c256:
roles:
agg:
gpus: 8
args:
tensor-parallel-size: 8
attn-dp-size: 2
enable-dp-lm-head: true
expert-parallel-size: 8
chunked-prefill-size: 65536
max-running-requests: 256
cuda-graph-max-bs-decode: 128
mem-fraction-static: 0.8
env:
SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: '32768'
benchmark:
env:
CONC: '256'
KV_OFFLOADING: none
24 changes: 24 additions & 0 deletions inferencex-e2e/configs/nvidia-master.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -9444,6 +9444,30 @@ dsv41flash-fp4-b300-sglang-agentic-dspark:
- { tp: 2, ep: 2, dp-attn: true, kv-offloading: dram, kv-offload-backend: { name: hicache }, spec-decoding: mtp, conc-list: [64, 256], router: { name: sglang-router, version: "0.3.2" }, srt-recipe: agentic.yaml }
- { tp: 4, ep: 4, dp-attn: true, kv-offloading: dram, kv-offload-backend: { name: hicache }, spec-decoding: mtp, conc-list: [256], router: { name: sglang-router, version: "0.3.2" }, srt-recipe: agentic.yaml }

# MiniMax-M3.1-Flash AgentX on B300 with SGLang: day-zero image from the SGLang
# cookbook (https://docs.sglang.io/cookbook/autoregressive/MiniMax/MiniMax-M3.1),
# manifest digest verified 2026-10-24. Serve flags follow the cookbook's
# B300 Balanced (TP4) and High-Throughput (TP8, attention DP2, EP8).
# Seven-token DSpark (bundled draft); synthetic acceptance follows
# infx/golden_al_distribution/minimaxm31flash_dspark.yaml.
Comment on lines +9447 to +9452

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 nit (optional): The new config and recipe add header and provenance comments, which the mandatory comment rules in AGENTS.md:64-65 forbid. sweep:^# (MiniMax-M3.1-Flash AgentX|TP4 Balanced cell|High-Throughput cell)|# Serve flags follow the cookbook For example, nvidia-master.yaml:9362-9367 describes the config and says "manifest digest verified 2026-10-24". It is also dated a day after the PR was opened. Fix: delete these block, section and provenance comments. Keep only short per-line technical reasons, such as the dp-attn note. The cookbook URL and verification details already live in the perf-changelog entry.

Why this was flagged

AGENTS.md:62-66 says the comment rules cover every file, recipes and configs included. It bans comments above a recipe, override or section, and comments that describe the config. It also bans comments saying where something came from. The diff adds a 6-line block above minimaxm31flash-fp4-b300-sglang-agentic-dspark (nvidia-master.yaml:9362-9367) that describes the topology and records provenance ("day-zero image from the SGLang cookbook", "manifest digest verified 2026-10-24"). The recipe adds section comments above the override groups (agentic.yaml:51-52 and 129-131) and a cookbook-source comment (agentic.yaml:24-25). Reviewers and the CODEOWNER checklist have to flag these, and the comments are new rule violations that the base does not have. The verification date in the master config is after the PR's open date (2026-10-23), so that provenance claim is also unreliable.

Verification: nit: this happens on every merge of these files as written. AGENTS.md "Code comments (mandatory)" says "No header comments" and "No history". The diff adds a 6-line block right above minimaxm31flash-fp4-b300-sglang-agentic-dspark: in nvidia-master.yaml that describes the config and its provenance. The recipe adds the same kind of comments. Nothing breaks at runtime.

minimaxm31flash-fp4-b300-sglang-agentic-dspark:
image: lmsysorg/sglang:dev-minimax-m31-flash@sha256:614b1509bcb729c653528d448bff9850f8977fcad9d12ba56bd6dbe3d7985e87
model: MiniMaxAI/MiniMax-M3.1-Flash
model-prefix: minimaxm31flash
runner: cluster:b300-dsxe
precision: fp4
framework: sglang
multinode: false
srt-recipe-dir: minimaxm31flash/sglang/b300-fp4-mtp
scenarios:
agentic-coding:
- search-space:
# Balanced: TP4.
- { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 16, 32], srt-recipe: agentic.yaml }
# High-Throughput: TP8 + EP8 with attention DP2 inside the server (--attn-dp-size 2,
# not full DP attention, so dp-attn stays false).
- { tp: 8, ep: 8, spec-decoding: mtp, kv-offloading: none, conc-list: [64, 128, 256], srt-recipe: agentic.yaml }

# Official SGLang nightly; TP4 C1 uses GPU Engram, other points use host DRAM.
dsv41flash-fp4-gb300-sglang-agentic-dspark:
image: lmsysorg/sglang:nightly-dev-cu13-20261021-c1c287a4@sha256:b0473834763b2753e069cb96c1acaf235ba7fd632ccd9bbeb44da809c437fdeb
Expand Down
9 changes: 9 additions & 0 deletions inferencex-e2e/perf-changelog.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -11082,6 +11082,15 @@
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4444
scenario-type: [agentic-coding]

- config-keys:
- minimaxm31flash-fp4-b300-sglang-agentic-dspark
description:
- "Add day-zero MiniMax-M3.1-Flash (MiniMaxAI/MiniMax-M3.1-Flash, NVFP4) AgentX on B300 with SGLang image lmsysorg/sglang:dev-minimax-m31-flash (https://docs.sglang.io/cookbook/autoregressive/MiniMax/MiniMax-M3.1)."
- "Serve flags follow the cookbook's B300 Balanced (TP4) and High-Throughput (TP8, attention DP2, EP8): DSpark with block verification, breakable prefill graphs, minimax-m3 reasoning and tool parsers, 8 tokenizer workers."
- "Seven-token DSpark (the bundled draft's dspark_block_size); throughput runs use synthetic acceptance from the minimaxm31flash_dspark golden curve (AL 4.04 at 7 draft tokens, #4425)."
- "Search space: Balanced TP4 at conc 1-32; TP8 + EP8 with attention DP2 at conc 64-256."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4441

- config-keys:
- minimaxm31flash-fp4-mi355x-sglang-agentic-dspark
description:
Expand Down
Loading