d-flex flex-column width-full">

fix: reduce GB300 DEP CUDA graph capture sizes

cad3e01

Select commit

Loading

Failed to load commit list.

Uh oh!

There was an error while loading. Please reload this page.

Cursor

Cursor Bugbot

CodeQL - Code Quality on: dynamic

Analyze (python)

CodeQL on: dynamic

Analyze (actions)

Analyze (python)

Run Sweep on: pull_request

check-changelog

reuse-sweep-gate

setup

canary-select

reuse-ingest-artifacts

canary /

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp8/ep8/dpfalse/nw1) | disagg-true spec-none conc-4x16x64x128x4096

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp4/ep1/dpfalse/nw1) | disagg-true spec-none conc-1x4x8x16

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp4/ep4/dptrue/nw2) | disagg-true spec-none conc-2048

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dptrue/nw1) | disagg-true spec-none conc-512x4096

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dpfalse/nw1) | disagg-true spec-none conc-32

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-16

multi-node 1k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_1k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw3) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-4

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp4/ep1/dpfalse/nw1) | disagg-true spec-none conc-1x4x8x16

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-128

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-256x512

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-16

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw3) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-512

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw3) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-32

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw4) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-4096

multi-node 8k1k / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw5) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-4x64

single-node 1k1k /

single-node 8k1k /

agentic /

multi-node agentic /

eval /

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp4/ep1/dpfalse/nw1) | disagg-true spec-none conc-1x4x8x16 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw1) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-128 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-256x512 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw2) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-16 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw3) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-512 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw3) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-32 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw4) D(tp8/ep8/dptrue/nw2) | disagg-true spec-none conc-4096 | eval-only

multi-node eval / (vllm/vllm-openai:nightly-aarch64, MiniMaxAI/MiniMax-M3-MXFP8, minimaxm3, fp8, ... / minimaxm3_8k1k fp8 gb300-nv dynamo-vllm | P(tp2/ep2/dptrue/nw5) D(tp8/ep8/dpfalse/nw2) | disagg-true spec-none conc-4x64 | eval-only

collect-results / collect-results

collect-evals / collect-evals

upload-changelog-metadata

calc-success-rate

compare-results

trigger-ingest

comment-unofficial-run-visualizer

Code scanning results

CodeQL

fix: reduce GB300 DEP CUDA graph capture sizes

Uh oh!

Select a check to view from the sidebar