minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

add NPU GSM8K accuracy tests for 7 models (#32649)

Browse Source
This commit is contained in:
Sugar920
2026-08-01 14:22:57 +08:00
committed by GitHub
parent c0d06a6547
commit fd96a35fb0
8 changed files with 486 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/pr-test-npu.yml
+25
View File
@@ -463,6 +463,31 @@ jobs:
runner: linux-aarch64-a3-2
test_case: test/registered/ascend/performance/qwen3_6_27b/test_npu_qwen3_6_27b_w8a8_1p_in64k_out1k_50ms.py
test_type: 'perf'
# NPU accuracy tests
- name: glm4_7_flash_1p_gsm8k
runner: linux-aarch64-a3-2
test_case: test/registered/ascend/accuracy/glm4_7_flash/test_npu_glm4_7_flash_1p_gsm8k.py
test_type: 'accuracy'
- name: qwen3_vl_30b_a3b_bf16_2p_gsm8k
runner: linux-aarch64-a3-4
test_case: test/registered/ascend/accuracy/qwen3_vl_30b_a3b/test_npu_qwen3_vl_30b_a3b_bf16_2p_gsm8k.py
test_type: 'accuracy'
- name: glm5_top64_pruned_bf16_8p_gsm8k
runner: linux-aarch64-a3-16
test_case: test/registered/ascend/accuracy/glm5_top64_pruned/test_npu_glm5_top64_pruned_bf16_8p_gsm8k.py
test_type: 'accuracy'
- name: moonshotai_moonlight_16b_a3b_bf16_1p_gsm8k
runner: linux-aarch64-a3-2
test_case: test/registered/ascend/accuracy/moonshotai_moonlight_16b_a3b/test_npu_moonlight_16b_a3b_bf16_1p_gsm8k.py
test_type: 'accuracy'
- name: qwen3_5_9b_bf16_1p_gsm8k
runner: linux-aarch64-a3-2
test_case: test/registered/ascend/accuracy/qwen3_5_9b/test_npu_qwen3_5_9b_bf16_1p_gsm8k.py
test_type: 'accuracy'
- name: qwen3_vl_8b_bf16_2p_gsm8k
runner: linux-aarch64-a3-4
test_case: test/registered/ascend/accuracy/qwen3_vl_8b/test_npu_qwen3_vl_8b_bf16_2p_gsm8k.py
test_type: 'accuracy'
uses: ./.github/workflows/nightly-test-npu-e2e-single-node.yml
with:
runner: ${{ matrix.test_config.runner }}