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

[AMD] CI - Fix AMD CI (multimodal test, move flaky test to non-deterministic group) (#20815)

Browse Source
This commit is contained in:
YC Yen-Ching Tseng
2026-03-19 11:50:19 -07:00
committed by GitHub
parent 3b89c6d3ea
commit 9e629d31fd
4 changed files with 7 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/pr-test-amd-rocm720.yml
+2 -2
View File
@@ -533,7 +533,7 @@ jobs:
max-parallel: 1 # Run one at a time to avoid eviction from resource exhaustion during AITER kernel JIT max-parallel: 1 # Run one at a time to avoid eviction from resource exhaustion during AITER kernel JIT
matrix: matrix:
runner: [linux-mi325-1gpu-sglang] runner: [linux-mi325-1gpu-sglang]
part: [0, 1] # 2 partitions: 11 tests ÷ 2 = ~5-6 tests each part: [0, 1, 2, 3]
runs-on: ${{matrix.runner}} runs-on: ${{matrix.runner}}
steps: steps:
- name: Checkout code - name: Checkout code
@@ -639,7 +639,7 @@ jobs:
ci_sglang python3 sglang/multimodal_gen/test/run_suite.py \ ci_sglang python3 sglang/multimodal_gen/test/run_suite.py \
--suite 1-gpu \ --suite 1-gpu \
--partition-id ${{ matrix.part }} \ --partition-id ${{ matrix.part }} \
--total-partitions 2 \ --total-partitions 4 \
-k "not flux_2" -k "not flux_2"
# Post-test diagnostics # Post-test diagnostics
.github/workflows/pr-test-amd.yml
+3 -3
View File
@@ -539,7 +539,7 @@ jobs:
max-parallel: 1 # Run one at a time to avoid eviction from resource exhaustion during AITER kernel JIT max-parallel: 1 # Run one at a time to avoid eviction from resource exhaustion during AITER kernel JIT
matrix: matrix:
runner: [linux-mi325-1gpu-sglang] runner: [linux-mi325-1gpu-sglang]
part: [0, 1] # 2 partitions: 11 tests ÷ 2 = ~5-6 tests each part: [0, 1, 2, 3] # 2 partitions: 11 tests ÷ 2 = ~5-6 tests each
runs-on: ${{matrix.runner}} runs-on: ${{matrix.runner}}
steps: steps:
- name: Checkout code - name: Checkout code
@@ -621,7 +621,7 @@ jobs:
docker exec ci_sglang rocm-smi --showmeminfo vram 2>/dev/null || echo "rocm-smi not available" docker exec ci_sglang rocm-smi --showmeminfo vram 2>/dev/null || echo "rocm-smi not available"
- name: Run diffusion server tests (1-GPU) - name: Run diffusion server tests (1-GPU)
timeout-minutes: 70 timeout-minutes: 90
run: | run: |
# AMD CI: All 1-GPU tests except FLUX.2 (FLUX.1 covers same code path) # AMD CI: All 1-GPU tests except FLUX.2 (FLUX.1 covers same code path)
# Tests: T2V, T2I, I2V, LoRA # Tests: T2V, T2I, I2V, LoRA
@@ -644,7 +644,7 @@ jobs:
ci_sglang python3 sglang/multimodal_gen/test/run_suite.py \ ci_sglang python3 sglang/multimodal_gen/test/run_suite.py \
--suite 1-gpu \ --suite 1-gpu \
--partition-id ${{ matrix.part }} \ --partition-id ${{ matrix.part }} \
--total-partitions 2 \ --total-partitions 4 \
-k "not flux_2" -k "not flux_2"
# Post-test diagnostics # Post-test diagnostics
test/registered/lora/test_multi_lora_backend.py
+1 -1
View File
@@ -26,7 +26,7 @@ from sglang.test.lora_utils import (
from sglang.test.test_utils import CustomTestCase, is_in_ci from sglang.test.test_utils import CustomTestCase, is_in_ci
register_cuda_ci(est_time=100, suite="stage-b-test-large-1-gpu") register_cuda_ci(est_time=100, suite="stage-b-test-large-1-gpu")
register_amd_ci(est_time=100, suite="stage-b-test-small-1-gpu-amd-nondeterministic") register_amd_ci(est_time=100, suite="stage-b-test-small-1-gpu-amd")
class TestMultiLoRABackend(CustomTestCase): class TestMultiLoRABackend(CustomTestCase):
test/registered/models/test_vlm_models.py
+1 -1
View File
@@ -17,7 +17,7 @@ from sglang.test.test_utils import is_in_ci
register_cuda_ci(est_time=228, suite="stage-b-test-large-1-gpu") register_cuda_ci(est_time=228, suite="stage-b-test-large-1-gpu")
register_amd_ci(est_time=850, suite="stage-b-test-small-1-gpu-amd") register_amd_ci(est_time=850, suite="stage-b-test-small-1-gpu-amd-nondeterministic")
_is_hip = is_hip() _is_hip = is_hip()
# VLM models for testing # VLM models for testing