Update Megatron

[pan-x-c]

Description

1. Update Megatron-LM to the latest source version
2. Update bench tool to support using sglang

Checklist

Please check the following items before code is ready to be reviewed.

• Code has passed all tests
• Docstrings have been added/updated in Google Style
• Documentation has been updated
• Code is ready for review

[pan-x-c]

/unittest-module-trainer

[pan-x-c]

/unittest-module-trainer

[pan-x-c]

/unittest-module-trainer

[github-actions]

Summary

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Other ❓	Flaky 🍂	Duration ⏱️
28	23	2	3	0	0	1h 4m

Failed Tests

Failed Tests ❌	Fail Message
❌ tests/trainer/trainer_test.py::TestTrainerGSM8K_2_megatron::test_trainer	The test failed in the call phase due to an assertion error
❌ tests/trainer/trainer_test.py::TestTrainerCheckpointSave_1_megatron::test_trainer	The test failed in the call phase due to an assertion error

Skipped

Tests	Status
tests/trainer/trainer_test.py::TestTrainerSFTWarmupGSM8K::test_trainer	skipped ⏭️
tests/trainer/trainer_test.py::TestTinkerTrainer::test_trainer	skipped ⏭️
tests/trainer/trainer_test.py::TestTinkerTrainer::test_trainer_class	skipped ⏭️

Tests

Test Name	Status	Flaky	Duration
tests/trainer/trainer_test.py::TestTrainerCountdown_0_fsdp2::test_trainer	✅		5m 32s
tests/trainer/trainer_test.py::TestTrainerCountdown_1_megatron::test_trainer	✅		6m 54s
tests/trainer/trainer_test.py::TestStepAheadAsyncRL::test_trainer	✅		2m 28s
tests/trainer/trainer_test.py::TestTrainerGSM8K_0_megatron::test_trainer	✅		2m 5s
tests/trainer/trainer_test.py::TestTrainerGSM8K_1_fsdp2::test_trainer	✅		1m 39s
tests/trainer/trainer_test.py::TestTrainerGSM8K_2_megatron::test_trainer	❌		4m 13s
tests/trainer/trainer_test.py::TestTrainerGSM8K_3_fsdp2::test_trainer	✅		3m 42s
tests/trainer/trainer_test.py::TestTrainerSFTWarmupGSM8K::test_trainer	⏭️		1ms
tests/trainer/trainer_test.py::TestTrainerDPO::test_trainer	✅		41.5s
tests/trainer/trainer_test.py::TestTrainerSFT::test_trainer	✅		37.9s
tests/trainer/trainer_test.py::TestTrainerToolsSFT::test_trainer_tools	✅		40.5s
tests/trainer/trainer_test.py::TestFullyAsyncMode_0_fsdp::test_fully_async_mode	✅		2m 16s
tests/trainer/trainer_test.py::TestFullyAsyncMode_1_fsdp2::test_fully_async_mode	✅		2m 4s
tests/trainer/trainer_test.py::TestFullyAsyncMode_2_megatron::test_fully_async_mode	✅		2m 49s
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_0_fsdp::test_trainer	✅		3m 25s
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_1_megatron::test_trainer	❌		1m 53s
tests/trainer/trainer_test.py::TestTrainerMIX::test_trainer	✅		3m 8s
tests/trainer/trainer_test.py::TestServeWithTrainer::test_serve_with_trainer	✅		2m 9s
tests/trainer/trainer_test.py::TestMultiModalGRPO::test_trainer	✅		3m 36s
tests/trainer/trainer_test.py::TestMultiModalSFT::test_trainer	✅		1m 40s
tests/trainer/trainer_test.py::TestTrainerLoRA::test_trainer	✅		4m 51s
tests/trainer/trainer_test.py::TestOverRollout::test_trainer	✅		1m 39s
tests/trainer/trainer_test.py::TestTrainerPromptTruncation::test_trainer	✅		1m 39s
tests/trainer/trainer_test.py::TestTinkerTrainer::test_trainer	⏭️		1ms
tests/trainer/trainer_test.py::TestTinkerTrainer::test_trainer_class	⏭️		1ms
tests/trainer/trainer_test.py::AgentScopeTunerTest::test_agentscope_tuner	✅		2m 23s
tests/trainer/trainer_test.py::ColocateModeTest::test_trainer	✅		2m 25s
tests/trainer/verl_utils_test.py::TestToDataProtoRoutedExperts::test_to_data_proto_pads_routed_experts_to_full_sequence	✅		2ms

Github Test Reporter by CTRF 💚

[pan-x-c]

/unittest-pattern-TestTrainerGSM8K_2_megatron

[pan-x-c]

/unittest-pattern-TestTrainerCheckpointSave_1_megatron

[github-actions]

Summary

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Other ❓	Flaky 🍂	Duration ⏱️
1	0	1	0	0	0	5m 4s

Failed Tests

Failed Tests ❌	Fail Message
❌ tests/trainer/trainer_test.py::TestTrainerGSM8K_2_megatron::test_trainer	The test failed in the call phase due to an assertion error

Tests

Test Name	Status	Flaky	Duration
tests/trainer/trainer_test.py::TestTrainerGSM8K_2_megatron::test_trainer	❌		4m 43s

Github Test Reporter by CTRF 💚

[github-actions]

Summary

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Other ❓	Flaky 🍂	Duration ⏱️
1	0	1	0	0	0	2m 54s

Failed Tests

Failed Tests ❌	Fail Message
❌ tests/trainer/trainer_test.py::TestTrainerCheckpointSave_1_megatron::test_trainer	The test failed in the call phase due to an assertion error

Tests

Test Name	Status	Flaky	Duration
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_1_megatron::test_trainer	❌		2m 34s

Github Test Reporter by CTRF 💚