Skip to content

Commit dc999ef

Browse files
seungrokjclaude
andcommitted
[AMD] agentx-v0.4: fix configs for MiniMax/Kimi/Qwen agentic entries
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
1 parent d2b2826 commit dc999ef

1 file changed

Lines changed: 12 additions & 5 deletions

File tree

.github/configs/amd-master.yaml

Lines changed: 12 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -872,6 +872,7 @@ minimaxm2.5-fp4-mi355x-atom:
872872
- { tp: 4, conc-start: 4, conc-end: 128 }
873873
- { tp: 8, conc-start: 4, conc-end: 16 }
874874

875+
# target
875876
minimaxm2.5-fp4-mi355x-vllm-agentic-lmcache:
876877
image: vllm/vllm-openai-rocm:v0.22.0
877878
model: amd/MiniMax-M2.5-MXFP4
@@ -884,8 +885,8 @@ minimaxm2.5-fp4-mi355x-vllm-agentic-lmcache:
884885
agentic-coding:
885886
- duration: 1800
886887
search-space:
887-
- { tp: 1, ep: 1, offloading: none, conc-list: [4, 8, 16, 32, 40, 48] }
888-
- { tp: 1, ep: 1, offloading: lmcache, conc-list: [4, 8, 16, 32, 40, 48] }
888+
- { tp: 1, ep: 1, offloading: none, conc-list: [1, 2, 4, 8, 16, 32, 40, 48] }
889+
- { tp: 1, ep: 1, offloading: lmcache, conc-list: [1, 2, 4, 8, 16, 32, 40, 48] }
889890

890891
minimaxm2.5-fp4-mi355x-vllm:
891892
image: vllm/vllm-openai-rocm:v0.22.0
@@ -2533,6 +2534,7 @@ kimik2.5-fp4-mi355x-vllm-agentic:
25332534
- { tp: 4, offloading: none, conc-list: [16, 24, 32, 40] }
25342535
- { tp: 4, offloading: cpu, conc-list: [16, 24, 32, 40] }
25352536

2537+
# target
25362538
kimik2.5-fp4-mi355x-vllm-agentic-lmcache:
25372539
image: vllm/vllm-openai-rocm:v0.21.0
25382540
model: amd/Kimi-K2.5-MXFP4
@@ -2603,16 +2605,21 @@ minimaxm2.5-fp8-mi325x-vllm-agentic:
26032605
- { tp: 4, offloading: none, conc-list: [1, 2, 4, 8, 16, 20, 24, 28, 32, 40, 48] }
26042606
- { tp: 4, offloading: cpu, conc-list: [16, 20, 24, 28, 32] }
26052607

2608+
# target
26062609
qwen3.5-fp8-mi355x-sglang-agentic-hicache:
26072610
image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260531
26082611
model: Qwen/Qwen3.5-397B-A17B-FP8
26092612
model-prefix: qwen3.5
2610-
runner: mi355x
2613+
runner: b300
2614+
precision: fp8
2615+
framework: sglang
2616+
multinode: false
2617+
scenarios:
26112618
agentic-coding:
26122619
- duration: 1800
26132620
search-space:
2614-
- { tp: 4, ep: 1, offloading: none, conc-list: [4, 8, 16, 32, 40, 48, 56, 64, 128] }
2615-
- { tp: 4, ep: 1, offloading: hicache, conc-list: [4, 8, 16, 32, 40, 48, 56, 64, 128] }
2621+
- { tp: 4, ep: 1, offloading: none, conc-list: [1, 2, 4, 8, 16, 32, 40, 48] }
2622+
- { tp: 4, ep: 1, offloading: hicache, conc-list: [1, 2, 4, 8, 16, 32, 40, 48] }
26162623

26172624
dsv4-fp4-mi355x-vllm-agentic:
26182625
image: vllm/vllm-openai-rocm:v0.22.0

0 commit comments

Comments
 (0)