更新0.12

160bf237 · wangxj · b01809dd · 160bf237 · 160bf237 · 160bf237
Commit 160bf237 authored Mar 05, 2025 by wangxj
20 changed files
--- a/megatron/core/optimizer/clip_grads.py
+++ b/megatron/core/optimizer/clip_grads.py
--- a/megatron/core/optimizer/cpu_offloading/README.md
+++ b/megatron/core/optimizer/cpu_offloading/README.md
--- a/megatron/core/optimizer/cpu_offloading/__init__.py
+++ b/megatron/core/optimizer/cpu_offloading/__init__.py
--- a/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py
+++ b/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py
--- a/megatron/core/optimizer/distrib_optimizer.py
+++ b/megatron/core/optimizer/distrib_optimizer.py
--- a/megatron/core/optimizer/grad_scaler.py
+++ b/megatron/core/optimizer/grad_scaler.py
--- a/megatron/core/optimizer/optimizer.py
+++ b/megatron/core/optimizer/optimizer.py
--- a/megatron/core/optimizer/optimizer_config.py
+++ b/megatron/core/optimizer/optimizer_config.py
--- a/megatron/core/optimizer_param_scheduler.py
+++ b/megatron/core/optimizer_param_scheduler.py
--- a/megatron/core/package_info.py
+++ b/megatron/core/package_info.py
--- a/megatron/core/packed_seq_params.py
+++ b/megatron/core/packed_seq_params.py
--- a/megatron/core/parallel_state.py
+++ b/megatron/core/parallel_state.py
--- a/megatron/core/pipeline_parallel/__init__.py
+++ b/megatron/core/pipeline_parallel/__init__.py
--- a/megatron/core/pipeline_parallel/p2p_communication.py
+++ b/megatron/core/pipeline_parallel/p2p_communication.py
--- a/megatron/core/pipeline_parallel/schedules.py
+++ b/megatron/core/pipeline_parallel/schedules.py
--- a/megatron/core/post_training/__init__.py
+++ b/megatron/core/post_training/__init__.py
--- a/megatron/core/post_training/modelopt/__init__.py
+++ b/megatron/core/post_training/modelopt/__init__.py
--- a/megatron/inference/algos/__init__.py
+++ b/megatron/inference/algos/__init__.py
--- a/megatron/core/post_training/modelopt/gpt/model_specs.py
+++ b/megatron/core/post_training/modelopt/gpt/model_specs.py
--- a/megatron/core/post_training/modelopt/gpt/state_dict_hooks.py
+++ b/megatron/core/post_training/modelopt/gpt/state_dict_hooks.py