Skip to content

Yeswanth/minimax fp4 gb300 b300 dynamo vllm disagg #1560

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
yeswanthk-26 wants to merge 9 commits into
base: main
Choose a base branch
from
Open

Yeswanth/minimax fp4 gb300 b300 dynamo vllm disagg #1560

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
5aa4520
[GB300][B300][vLLM] Add MiniMax-M2.5 FP4 disagg Dynamo configs
yeswanthk-26 May 22, 2026
b887433
Use SA default account for b300 minimax run
yeswanthk-26 May 23, 2026
6d14c33
Merge branch 'main' into yeswanth/minimax-fp4-gb300-b300-dynamo-vllm-…
Ankur-singh May 23, 2026
2f1c47c
Simplify descriptions for MiniMax benchmarks
Ankur-singh May 23, 2026
6d9c522
Remove b300 8k1k conc512 outlier point
yeswanthk-26 May 25, 2026
a135a6a
Pin minimax gb300 runs to gb300-nv runner
yeswanthk-26 May 26, 2026
0f6e7bc
Fix gb300 minimax model path to scratch
yeswanthk-26 May 26, 2026
00b1ca0
Restore squash validation path on gb300-nv
yeswanthk-26 May 26, 2026
bea3877
Merge main into minimax submission branch
yeswanthk-26 May 26, 2026
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
387 changes: 387 additions & 0 deletions .github/configs/nvidia-master.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9428,3 +9428,390 @@ glm5-fp4-gb300-dynamo-sglang:
tp: 4
ep: 1
dp-attn: false

minimaxm2.5-fp4-gb300-dynamo-vllm:
image: vllm/vllm-openai:v0.20.1
model: nvidia/MiniMax-M2.5-NVFP4
model-prefix: minimaxm2.5
runner: gb300-nv
precision: fp4
framework: dynamo-vllm
multinode: true
disagg: true
scenarios:
fixed-seq-len:
- isl: 1024
osl: 1024
search-space:
- conc-list: [2, 4, 16]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 1
dp-attn: false
- conc-list: [4, 8, 16, 64]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p2d.yaml"
decode:
num-worker: 2
tp: 4
ep: 1
dp-attn: false
- conc-list: [32, 64, 128]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: false
- conc-list: [64, 128, 256, 512, 1024]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p3d.yaml"
decode:
num-worker: 3
tp: 4
ep: 4
dp-attn: false
- conc-list: [2048]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d.yaml"
decode:
num-worker: 3
tp: 2
ep: 2
dp-attn: true
- conc-list: [6144, 8192]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d-c6144.yaml"
decode:
num-worker: 3
tp: 2
ep: 2
dp-attn: true
- conc-list: [1024, 2048, 4096]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep8-2p1d.yaml"
decode:
num-worker: 1
tp: 8
ep: 8
dp-attn: true
- isl: 8192
osl: 1024
search-space:
- conc-list: [2, 4, 8, 16]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 1
dp-attn: false
- conc-list: [32, 64, 128, 256]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: false
- conc-list: [64, 128]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-2p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: false
- conc-list: [256]
prefill:
num-worker: 4
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep4-4p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: true
- conc-list: [1024, 2048]
prefill:
num-worker: 4
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep8-4p1d.yaml"
decode:
num-worker: 1
tp: 8
ep: 8
dp-attn: true

minimaxm2.5-fp4-b300-dynamo-vllm:
image: vllm/vllm-openai:v0.20.1
model: nvidia/MiniMax-M2.5-NVFP4
model-prefix: minimaxm2.5
runner: b300
precision: fp4
framework: dynamo-vllm
multinode: true
disagg: true
scenarios:
fixed-seq-len:
- isl: 1024
osl: 1024
search-space:
- conc-list: [4, 16]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 1
dp-attn: false
- conc-list: [4]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp8-1p1d.yaml"
decode:
num-worker: 1
tp: 8
ep: 1
dp-attn: false
- conc-list: [8, 16]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p2d.yaml"
decode:
num-worker: 2
tp: 4
ep: 1
dp-attn: false
- conc-list: [32, 64, 128]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: false
- conc-list: [64, 128, 256, 1024]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p3d.yaml"
decode:
num-worker: 3
tp: 4
ep: 4
dp-attn: false
- conc-list: [4096]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-1p2d.yaml"
decode:
num-worker: 2
tp: 2
ep: 2
dp-attn: true
- conc-list: [2048, 4096]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d.yaml"
decode:
num-worker: 3
tp: 2
ep: 2
dp-attn: true
- conc-list: [6144, 8192]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d-c6144.yaml"
decode:
num-worker: 3
tp: 2
ep: 2
dp-attn: true
- conc-list: [1024, 1536, 2048, 4096]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep8-2p1d.yaml"
decode:
num-worker: 1
tp: 8
ep: 8
dp-attn: true
- isl: 8192
osl: 1024
search-space:
- conc-list: [2, 4, 8, 16]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 1
dp-attn: false
- conc-list: [4]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp8-1p1d.yaml"
decode:
num-worker: 1
tp: 8
ep: 1
dp-attn: false
- conc-list: [32, 128]
prefill:
num-worker: 1
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-1p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: false
- conc-list: [64, 128, 256, 512]
prefill:
num-worker: 2
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-2p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: false
- conc-list: [384]
prefill:
num-worker: 4
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep4-4p1d.yaml"
decode:
num-worker: 1
tp: 4
ep: 4
dp-attn: true
- conc-list: [384]
prefill:
num-worker: 4
tp: 1
ep: 1
dp-attn: false
additional-settings:
- "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep8-4p1d.yaml"
decode:
num-worker: 1
tp: 8
ep: 8
dp-attn: true
Loading
Loading