Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
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
2 changes: 1 addition & 1 deletion .main.commit
Original file line number Diff line number Diff line change
@@ -1 +1 @@
3758b54b2bd9d00f0adbe8bf343a0e6bbcd75d1d
97e36aaba37245e2fa23d4fda0e956ece21e6c9e
2 changes: 1 addition & 1 deletion 3rdparty/Megatron-LM
Submodule Megatron-LM updated 67 files
+3 −9 .github/actions/action.yml
+1 −1 .github/copy-pr-bot.yaml
+11 −13 .github/workflows/_build_test_publish_wheel.yml
+7 −4 .github/workflows/release-freeze.yml
+1 −0 docker/Dockerfile.ci.dev
+3 −3 docs/get-started/install.md
+8 −13 docs/get-started/overview.md
+4 −4 docs/get-started/quickstart.md
+4 −2 megatron/core/dist_checkpointing/serialization.py
+3 −0 megatron/core/dist_checkpointing/strategies/async_utils.py
+26 −6 megatron/core/dist_checkpointing/strategies/fully_parallel.py
+215 −70 megatron/core/dist_checkpointing/strategies/torch.py
+48 −0 megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py
+7 −0 megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py
+7 −3 megatron/core/extensions/transformer_engine.py
+8 −0 megatron/core/inference/contexts/dynamic_context.py
+59 −5 megatron/core/inference/text_generation_controllers/text_generation_controller.py
+9 −0 megatron/core/model_parallel_config.py
+36 −0 megatron/core/models/common/language_module/language_module.py
+12 −0 megatron/core/models/gpt/fine_grained_callables.py
+0 −43 megatron/core/models/gpt/gpt_model.py
+18 −1 megatron/core/models/mamba/mamba_layer_specs.py
+0 −44 megatron/core/models/mamba/mamba_model.py
+1 −1 megatron/core/package_info.py
+8 −3 megatron/core/ssm/mamba_mixer.py
+86 −4 megatron/core/transformer/moe/moe_layer.py
+13 −0 megatron/core/transformer/transformer_config.py
+2 −2 megatron/training/arguments.py
+48 −17 megatron/training/async_utils.py
+9 −3 megatron/training/checkpointing.py
+10 −0 megatron/training/config/training_config.py
+35 −0 megatron/training/global_vars.py
+4 −2 megatron/training/inprocess_restart.py
+3 −3 pyproject.toml
+61 −61 tests/functional_tests/test_cases/gpt/gpt3_7b_tp4_pp1_memory_speed/golden_values_dev_dgx_h100.json
+0 −1 ...s/functional_tests/test_cases/gpt/gpt3_mcore_te_tp1_pp2_resume_torch_dist_rope_embeddings/model_config.yaml
+0 −1 tests/functional_tests/test_cases/gpt/gpt3_mcore_te_tp2_pp1_fsdp2_resume_torch_dist/model_config.yaml
+1 −0 ...t/gpt3_mcore_te_tp4_pp1_resume_torch_dist_dist_optimizer_overlap_grad_reduce_param_gather/model_config.yaml
+0 −1 tests/functional_tests/test_cases/gpt/gpt3_mcore_tp1_pp1_fsdp2_resume_torch_dist_te/model_config.yaml
+1 −0 tests/functional_tests/test_cases/hybrid/hybrid_mr_mcore_te_tp1_pp4_cp1_dgx_a100_1N8G/model_config.yaml
+1 −0 tests/functional_tests/test_cases/moe/gpt3_moe_mcore_te_ep8_resume_torch_dist_dist_optimizer/model_config.yaml
+220 −0 tests/unit_tests/a2a_overlap/test_delay_wgrad_compute.py
+0 −2 tests/unit_tests/a2a_overlap/test_schedule_chunk_1f1b.py
+0 −2 tests/unit_tests/a2a_overlap/test_schedule_layer_1f1b.py
+28 −7 tests/unit_tests/a2a_overlap/utils.py
+30 −30 tests/unit_tests/dist_checkpointing/test_async_save.py
+1 −1 tests/unit_tests/dist_checkpointing/test_fully_parallel.py
+1 −0 tests/unit_tests/dist_checkpointing/test_pipeline_parallel_layout.py
+1 −1 tests/unit_tests/dist_checkpointing/test_replication.py
+26 −15 tests/unit_tests/inference/contexts/test_dynamic_context.py
+51 −38 tests/unit_tests/inference/contexts/test_dynamic_prefix_caching.py
+76 −18 tests/unit_tests/inference/engines/test_dynamic_engine.py
+192 −2 tests/unit_tests/inference/text_generation_controllers/test_text_generation_controller.py
+283 −0 tests/unit_tests/models/test_mamba_model_expert_parallel_inference.py
+1 −0 tests/unit_tests/models/test_mamba_moe_model.py
+2 −1 tests/unit_tests/pipeline_parallel/test_pipeline_layout.py
+1 −0 tests/unit_tests/test_checkpointing.py
+0 −1 tests/unit_tests/transformer/test_submodule_callables.py
+8 −16 tools/checkpoint/checkpoint_inspector.py
+0 −1 tools/checkpoint/loader_base.py
+0 −1 tools/checkpoint/loader_legacy.py
+0 −1 tools/checkpoint/loader_llama_mistral.py
+0 −1 tools/checkpoint/loader_mixtral_hf.py
+10 −0 tools/checkpoint/saver_base.py
+35 −0 tools/upgrade_dependencies.md
+26 −1 tools/upgrade_dependencies.sh
+2,106 −2,507 uv.lock
Loading
Loading