Commit bddbb54
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | | - | |
| 1 | + | |
Submodule Megatron-LM updated 22 files
- .claude/skills/split-pr/SKILL.md+62
- .github/workflows/_build_test_publish_wheel.yml+3-1
- megatron/core/distributed/fsdp/mcore_fsdp_adapter.py+10
- megatron/core/distributed/fsdp/src/megatron_fsdp/fully_shard.py+20
- megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py+19-9
- megatron/core/distributed/fsdp/src/megatron_fsdp/utils.py+14-14
- megatron/core/fp8_utils.py+1-1
- megatron/core/inference/text_generation_controllers/text_generation_controller.py+10
- megatron/core/models/mamba/mamba_model.py+25
- megatron/core/optimizer/optimizer.py+17-12
- megatron/core/package_info.py+20-1
- megatron/core/parallel_state.py+86-40
- megatron/core/process_groups_config.py+8
- megatron/core/transformer/moe/moe_utils.py+8-3
- megatron/core/transformer/transformer_config.py+1-1
- megatron/training/arguments.py+2-8
- megatron/training/initialize.py+26-12
- megatron/training/training.py+20-1
- pyproject.toml+1-1
- tests/functional_tests/test_cases/nemotron/nemotron3_super_release_g200/model_config.yaml+2-2
- tests/unit_tests/test_parallel_state.py+75-19
- uv.lock+86-27
0 commit comments