Skip to content

[CI] Reorg test file and fix moonlight oom #593

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

Merged
guapisolo merged 4 commits into from
Feb 17, 2026
+71 −2,299
Merged

[CI] Reorg test file and fix moonlight oom #593

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
2f8c8ac
1st
guapisolo Feb 13, 2026
8c50455
fix 2
guapisolo Feb 13, 2026
bd78fd0
better ci-image impl
guapisolo Feb 16, 2026
8ef732e
format
guapisolo Feb 16, 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
17 changes: 15 additions & 2 deletions .github/workflows/pr-test.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,19 @@
################################################################################















name: PR Test

on:
Expand Down Expand Up @@ -414,7 +427,7 @@ jobs:
strategy:
fail-fast: false
matrix:
info: [{"num_gpus": 8, "test_file": "e2e/precision/test_qwen3_0.6B_parallel_check.py"}, {"num_gpus": 8, "test_file": "e2e/precision/test_qwen3_0.6B_megatron_fsdp_align.py"}]
info: [{"num_gpus": 8, "test_file": "e2e/precision/test_qwen3_0.6B_parallel_check.py"}]
defaults:
run:
working-directory: ${{ github.workspace }}
Expand Down Expand Up @@ -597,7 +610,7 @@ jobs:
strategy:
fail-fast: false
matrix:
info: [{"num_gpus": 8, "test_file": "e2e/image/test_qwen2.5_0.5B_gsm8k_async_short.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen2.5_0.5B_gsm8k_short.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_0.6B_fsdp_colocated_2xGPU.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_4B_fsdp_true_on_policy.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_vl_4B_fsdp.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_0.6B_fsdp_distributed.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_quick_start_glm4_9B.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_30B_A3B.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_4B_ppo.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_moonlight_16B_A3B.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_mimo_7B_mtp_only_grad.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_0.6B_parallel_check.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_0.6B_megatron_fsdp_align.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_4B_ckpt.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen3_4B_ckpt.py --async-save"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen2.5_0.5B_gsm8k.py"}, {"num_gpus": 8, "test_file": "e2e/image/test_qwen2.5_0.5B_gsm8k_async.py"}]
info: [{"num_gpus": 8, "test_file": "e2e/fsdp/test_qwen3_4B_fsdp_true_on_policy.py"}, {"num_gpus": 8, "test_file": "e2e/fsdp/test_qwen3_vl_4B_fsdp.py"}, {"num_gpus": 8, "test_file": "e2e/fsdp/test_qwen3_0.6B_fsdp_distributed.py"}, {"num_gpus": 8, "test_file": "e2e/fsdp/test_qwen3_0.6B_megatron_fsdp_align.py"}, {"num_gpus": 8, "test_file": "e2e/megatron/test_quick_start_glm4_9B.py"}, {"num_gpus": 8, "test_file": "e2e/megatron/test_qwen3_30B_A3B.py", "use_deepep": "1", "use_fp8_rollout": "1"}, {"enable_eval": "0", "num_gpus": 8, "test_file": "e2e/megatron/test_qwen3_30B_A3B_r3.py", "use_deepep": "1", "use_fp8_rollout": "1"}, {"enable_eval": "0", "num_gpus": 8, "test_file": "e2e/megatron/test_qwen3_30B_A3B_r3.py"}, {"num_gpus": 8, "test_file": "e2e/megatron/test_qwen3_4B_ppo.py"}, {"num_gpus": 8, "test_file": "e2e/megatron/test_moonlight_16B_A3B.py"}, {"enable_eval": "0", "num_gpus": 8, "test_file": "e2e/megatron/test_moonlight_16B_A3B_r3.py"}, {"num_gpus": 8, "test_file": "e2e/megatron/test_mimo_7B_mtp_only_grad.py"}, {"num_gpus": 8, "test_file": "e2e/short/test_qwen2.5_0.5B_gsm8k_async_short.py"}, {"num_gpus": 8, "test_file": "e2e/short/test_qwen2.5_0.5B_gsm8k_short.py"}, {"num_gpus": 8, "test_file": "e2e/short/test_qwen3_0.6B_fsdp_colocated_2xGPU.py"}, {"num_gpus": 8, "test_file": "e2e/precision/test_qwen3_0.6B_parallel_check.py"}, {"num_gpus": 8, "test_file": "e2e/ckpt/test_qwen3_4B_ckpt.py"}, {"num_gpus": 8, "test_file": "e2e/ckpt/test_qwen3_4B_ckpt.py --async-save"}, {"num_gpus": 8, "test_file": "e2e/long/test_qwen2.5_0.5B_gsm8k.py"}, {"num_gpus": 8, "test_file": "e2e/long/test_qwen2.5_0.5B_gsm8k_async.py"}]
defaults:
run:
working-directory: ${{ github.workspace }}
Expand Down
98 changes: 46 additions & 52 deletions .github/workflows/pr-test.yml.j2
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,43 @@
<% set default_image = 'radixark/miles:dev' %>

<% set fsdp_tests = [
{'test_file': 'e2e/fsdp/test_qwen3_4B_fsdp_true_on_policy.py', 'num_gpus': 8},
{'test_file': 'e2e/fsdp/test_qwen3_vl_4B_fsdp.py', 'num_gpus': 8},
{'test_file': 'e2e/fsdp/test_qwen3_0.6B_fsdp_distributed.py', 'num_gpus': 8},
{'test_file': 'e2e/fsdp/test_qwen3_0.6B_megatron_fsdp_align.py', 'num_gpus': 8},
] %>

<% set megatron_tests = [
{'test_file': 'e2e/megatron/test_quick_start_glm4_9B.py', 'num_gpus': 8},
{'test_file': 'e2e/megatron/test_qwen3_30B_A3B.py', 'num_gpus': 8, 'use_deepep': '1', 'use_fp8_rollout': '1'},
{'test_file': 'e2e/megatron/test_qwen3_30B_A3B_r3.py', 'num_gpus': 8, 'use_deepep': '1', 'use_fp8_rollout': '1', 'enable_eval': '0'},
{'test_file': 'e2e/megatron/test_qwen3_30B_A3B_r3.py', 'num_gpus': 8, 'enable_eval': '0'},
{'test_file': 'e2e/megatron/test_qwen3_4B_ppo.py', 'num_gpus': 8},
{'test_file': 'e2e/megatron/test_moonlight_16B_A3B.py', 'num_gpus': 8},
{'test_file': 'e2e/megatron/test_moonlight_16B_A3B_r3.py', 'num_gpus': 8, 'enable_eval': '0'},
{'test_file': 'e2e/megatron/test_mimo_7B_mtp_only_grad.py', 'num_gpus': 8},
] %>

<% set short_tests = [
{'test_file': 'e2e/short/test_qwen2.5_0.5B_gsm8k_async_short.py', 'num_gpus': 8},
{'test_file': 'e2e/short/test_qwen2.5_0.5B_gsm8k_short.py', 'num_gpus': 8},
{'test_file': 'e2e/short/test_qwen3_0.6B_fsdp_colocated_2xGPU.py', 'num_gpus': 8},
] %>

<% set precision_tests = [
{'test_file': 'e2e/precision/test_qwen3_0.6B_parallel_check.py', 'num_gpus': 8},
] %>

<% set ckpt_tests = [
{'test_file': 'e2e/ckpt/test_qwen3_4B_ckpt.py', 'num_gpus': 8},
{'test_file': 'e2e/ckpt/test_qwen3_4B_ckpt.py --async-save', 'num_gpus': 8},
] %>

<% set long_tests = [
{'test_file': 'e2e/long/test_qwen2.5_0.5B_gsm8k.py', 'num_gpus': 8},
{'test_file': 'e2e/long/test_qwen2.5_0.5B_gsm8k_async.py', 'num_gpus': 8},
] %>

<% set jobs = {
'fast': {
'test_executor': 'pytest',
Expand All @@ -21,76 +60,31 @@
},
'e2e-test-short': {
'label': 'run-ci-short',
'tests': [
{'test_file': 'e2e/short/test_qwen2.5_0.5B_gsm8k_async_short.py', 'num_gpus': 8},
{'test_file': 'e2e/short/test_qwen2.5_0.5B_gsm8k_short.py', 'num_gpus': 8},
{'test_file': 'e2e/short/test_qwen3_0.6B_fsdp_colocated_2xGPU.py', 'num_gpus': 8},
],
'tests': short_tests,
},
'e2e-test-fsdp': {
'label': 'run-ci-fsdp',
'tests': [
{'test_file': 'e2e/fsdp/test_qwen3_4B_fsdp_true_on_policy.py', 'num_gpus': 8},
{'test_file': 'e2e/fsdp/test_qwen3_vl_4B_fsdp.py', 'num_gpus': 8},
{'test_file': 'e2e/fsdp/test_qwen3_0.6B_fsdp_distributed.py', 'num_gpus': 8},
{'test_file': 'e2e/fsdp/test_qwen3_0.6B_megatron_fsdp_align.py', 'num_gpus': 8},
],
'tests': fsdp_tests,
},
'e2e-test-megatron': {
'label': 'run-ci-megatron',
'tests': [
{'test_file': 'e2e/megatron/test_quick_start_glm4_9B.py', 'num_gpus': 8},
{'test_file': 'e2e/megatron/test_qwen3_30B_A3B.py', 'num_gpus': 8, 'use_deepep': '1', 'use_fp8_rollout': '1'},
{'test_file': 'e2e/megatron/test_qwen3_30B_A3B_r3.py', 'num_gpus': 8, 'use_deepep': '1', 'use_fp8_rollout': '1', 'enable_eval': '0'},
{'test_file': 'e2e/megatron/test_qwen3_30B_A3B_r3.py', 'num_gpus': 8, 'enable_eval': '0'},
{'test_file': 'e2e/megatron/test_qwen3_4B_ppo.py', 'num_gpus': 8},
{'test_file': 'e2e/megatron/test_moonlight_16B_A3B.py', 'num_gpus': 8},
{'test_file': 'e2e/megatron/test_moonlight_16B_A3B_r3.py', 'num_gpus': 8, 'enable_eval': '0'},
{'test_file': 'e2e/megatron/test_mimo_7B_mtp_only_grad.py', 'num_gpus': 8},
],
'tests': megatron_tests,
},
'e2e-test-precision': {
'label': 'run-ci-precision',
'tests': [
{'test_file': 'e2e/precision/test_qwen3_0.6B_parallel_check.py', 'num_gpus': 8},
{'test_file': 'e2e/precision/test_qwen3_0.6B_megatron_fsdp_align.py', 'num_gpus': 8},
],
'tests': precision_tests,
},
'e2e-test-ckpt': {
'label': 'run-ci-ckpt',
'tests': [
{'test_file': 'e2e/ckpt/test_qwen3_4B_ckpt.py', 'num_gpus': 8},
{'test_file': 'e2e/ckpt/test_qwen3_4B_ckpt.py --async-save', 'num_gpus': 8},
],
'tests': ckpt_tests,
},
'e2e-test-long': {
'label': 'run-ci-long',
'tests': [
{'test_file': 'e2e/long/test_qwen2.5_0.5B_gsm8k.py', 'num_gpus': 8},
{'test_file': 'e2e/long/test_qwen2.5_0.5B_gsm8k_async.py', 'num_gpus': 8},
],
'tests': long_tests,
},
'e2e-test-image': {
'label': 'run-ci-image',
'tests': [
{'test_file': 'e2e/image/test_qwen2.5_0.5B_gsm8k_async_short.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen2.5_0.5B_gsm8k_short.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_0.6B_fsdp_colocated_2xGPU.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_4B_fsdp_true_on_policy.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_vl_4B_fsdp.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_0.6B_fsdp_distributed.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_quick_start_glm4_9B.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_30B_A3B.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_4B_ppo.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_moonlight_16B_A3B.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_mimo_7B_mtp_only_grad.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_0.6B_parallel_check.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_0.6B_megatron_fsdp_align.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_4B_ckpt.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen3_4B_ckpt.py --async-save', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen2.5_0.5B_gsm8k.py', 'num_gpus': 8},
{'test_file': 'e2e/image/test_qwen2.5_0.5B_gsm8k_async.py', 'num_gpus': 8},
],
'tests': fsdp_tests + megatron_tests + short_tests + precision_tests + ckpt_tests + long_tests,
},
} %>
name: PR Test
Expand Down
147 changes: 0 additions & 147 deletions tests/e2e/image/test_mimo_7B_mtp_only_grad.py
View file
Open in desktop

This file was deleted.

Loading