Skip to content

[ATOM-SGL/VLLM-CI] Modify testing cases for SGLang and vLLM. #1530

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
junyyang-amd wants to merge 2 commits into
base: main
Choose a base branch
from
+41 −17
Open

[ATOM-SGL/VLLM-CI] Modify testing cases for SGLang and vLLM. #1530

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3f618b4
Modify arom-sglang-test cases
junyyang-amd Jul 9, 2026
ba8ae6a
Modify atom-vllm-test cases
junyyang-amd Jul 9, 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
39 changes: 27 additions & 12 deletions .github/workflows/atom-sglang-test.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -240,31 +240,30 @@ jobs:
fail-fast: false
matrix:
include:
- model_name: "DeepSeek-R1-FP8 TP4"
model_path: "deepseek-ai/DeepSeek-R1-0528"
extra_args: "--trust-remote-code --tensor-parallel-size 4 --attention-backend aiter --kv-cache-dtype fp8_e4m3 --mem-fraction-static 0.85 --page-size 1 --disable-radix-cache"
- model_name: "GLM-5.1-FP8 TP8"
model_path: "zai-org/GLM-5.1-FP8"
extra_args: "--trust-remote-code --tensor-parallel-size 8 --attention-backend aiter --kv-cache-dtype fp8_e4m3 --mem-fraction-static 0.85 --page-size 1 --disable-radix-cache"
env_vars: |
SGLANG_DEFAULT_SERVER_ARGS=
SGLANG_AITER_FP8_PREFILL_ATTN=0
SGLANG_USE_AITER=1
ATOM_ENABLE_DS_QKNORM_QUANT_FUSION=1
SGLANG_EXTERNAL_MODEL_PACKAGE=atom.plugin.sglang.models
SGLANG_ENABLE_TORCH_COMPILE=1
TORCHINDUCTOR_COMPILE_THREADS=128
accuracy_test_threshold: 0.91
lm_eval_num_fewshot: 5
accuracy_test_threshold: 0.93
runner: atom-mi355-8gpu-vllm-sgl-ci
- model_name: "DeepSeek-R1-FP4 TP4"
- model_name: "DeepSeek-R1-FP4-V2 TP8 MTP3"
model_path: "amd/DeepSeek-R1-0528-MXFP4-v2"
extra_args: "--trust-remote-code --tensor-parallel-size 4 --attention-backend aiter --kv-cache-dtype fp8_e4m3 --mem-fraction-static 0.85 --page-size 1 --disable-radix-cache"
extra_args: "--trust-remote-code --tensor-parallel-size 8 --attention-backend aiter --kv-cache-dtype fp8_e4m3 --mem-fraction-static 0.85 --page-size 1 --disable-radix-cache --speculative-draft-model-path SGLang/DeepSeek-R1-NextN --speculative-algorithm NEXTN --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4 --max-running-requests 256 --cuda-graph-bs 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 160 192 224 256"
env_vars: |
SGLANG_DEFAULT_SERVER_ARGS=
ATOM_ENABLE_DS_QKNORM_QUANT_FUSION=1
SGLANG_AITER_FP8_PREFILL_ATTN=0
SGLANG_USE_AITER=1
ATOM_ENABLE_DS_QKNORM_QUANT_FUSION=1
SGLANG_EXTERNAL_MODEL_PACKAGE=atom.plugin.sglang.models
SGLANG_ENABLE_SPEC_V2=1
SGLANG_ENABLE_TORCH_COMPILE=1
SGLANG_EXTERNAL_MODEL_PACKAGE=atom.plugin.sglang.models
TORCHINDUCTOR_COMPILE_THREADS=128
accuracy_test_threshold: 0.91
accuracy_test_threshold: 0.93
runner: atom-mi355-8gpu-vllm-sgl-ci
- model_name: "Qwen3.5-35B-A3B-FP8 TP2"
model_path: "Qwen/Qwen3.5-35B-A3B-FP8"
Expand All @@ -275,6 +274,22 @@ jobs:
ATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=0
accuracy_test_threshold: 0.76
runner: atom-mi355-8gpu-vllm-sgl-ci
- model_name: "DeepSeek-V4-Flash TP8"
model_path: "deepseek-ai/DeepSeek-V4-Flash"
extra_args: "--trust-remote-code --tensor-parallel-size 8 --kv-cache-dtype fp8_e4m3 --mem-fraction-static 0.9 --swa-full-tokens-ratio 0.1 --max-running-requests 256 --page-size 256 --disable-radix-cache --disable-shared-experts-fusion --tool-call-parser deepseekv4 --reasoning-parser deepseek-v4"
env_vars: |
SGLANG_DEFAULT_SERVER_ARGS=
AITER_BF16_FP8_MOE_BOUND=0
ATOM_MOE_GU_ITLV=1
SGLANG_DEFAULT_THINKING=1
SGLANG_DSV4_REASONING_EFFORT=max
SGLANG_USE_AITER=1
SGLANG_EXTERNAL_MODEL_PACKAGE=atom.plugin.sglang.models
TORCHINDUCTOR_COMPILE_THREADS=128
lm_eval_num_fewshot: 5
lm_eval_num_concurrent: 8
accuracy_test_threshold: 0.94
runner: atom-mi355-8gpu-vllm-sgl-ci
- model_name: "DeepSeek-V4-Pro TP8"
model_path: "deepseek-ai/DeepSeek-V4-Pro"
extra_args: "--trust-remote-code --tensor-parallel-size 8 --kv-cache-dtype fp8_e4m3 --mem-fraction-static 0.9 --swa-full-tokens-ratio 0.1 --max-running-requests 256 --page-size 256 --disable-radix-cache --disable-shared-experts-fusion --tool-call-parser deepseekv4 --reasoning-parser deepseek-v4"
Expand Down
19 changes: 14 additions & 5 deletions .github/workflows/atom-vllm-test.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -260,6 +260,14 @@ jobs:
lm_eval_num_fewshot: 3
accuracy_test_threshold: 0.88
runner: atom-mi355-8gpu-vllm-sgl-ci
- display_name: "GLM-5.1-FP8 TP8"
model_name: "GLM-5.1-FP8"
model_path: "zai-org/GLM-5.1-FP8"
extra_args: "--trust-remote-code --tensor-parallel-size 8 --default-chat-template-kwargs '{\"enable_thinking\":false}' --max-num-batched-tokens 16384 --max-model-len 16384"
env_vars: "AITER_QUICK_REDUCE_QUANTIZATION=INT4"
lm_eval_num_fewshot: 20
accuracy_test_threshold: 0.88
runner: atom-mi355-8gpu-vllm-sgl-ci
- display_name: "Kimi-K2.5-MXFP4 TP4"
model_name: "Kimi-K2.5-MXFP4"
model_path: "amd/Kimi-K2.5-MXFP4-AttnFP8"
Expand All @@ -268,15 +276,16 @@ jobs:
lm_eval_num_fewshot: 3
accuracy_test_threshold: 0.92
runner: atom-mi355-8gpu-vllm-sgl-ci
- display_name: "Qwen3.5-35B-A3B-FP8 TP2"
model_name: "Qwen3.5-35B-A3B-FP8"
model_path: "Qwen/Qwen3.5-35B-A3B-FP8"
extra_args: "--tensor-parallel-size 2 --attention-backend ROCM_AITER_FA"
- display_name: "Qwen3.5-397B-A17B-MXFP4 TP4"
model_name: "Qwen3.5-397B-A17B-MXFP4"
model_path: "amd/Qwen3.5-397B-A17B-MXFP4"
extra_args: "--tensor-parallel-size 4"
env_vars: |
ATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1
ATOM_USE_CUSTOM_ALL_GATHER=0
ATOM_FP8_BLOCKSCALE_WEIGHT_PRESHUFFLE=0
lm_eval_num_fewshot: 3
accuracy_test_threshold: 0.77
accuracy_test_threshold: 0.83
runner: atom-mi355-8gpu-vllm-sgl-ci
runs-on: ${{ matrix.runner }}
timeout-minutes: 180
Expand Down
Loading