(docs): fix mcore adapter convert.

文档更新将 Megatron 模型的相关说明更改为 MCoreAdapter 模型，并明确指出转换工具适用于由 `megatron_strategy` 训练的模型。
Link: https://code.alibaba-inc.com/openlm/ScaleAligner/codereview/23069812
* (docs): fix mcore adapter convert.

(cherry picked from commit 44d354363baafc9c8637f6c963696196b1dba95f)

Author: PanAndy

docs_roll/docs/English/UserGuide/megatron_convert_2_hf.md

@@ -1,21 +1,21 @@
-# Converting Megatron Models to Hugging Face Format
+# Converting MCoreAdapter Models to Hugging Face Format
 
-MCoreAdapter provides tools for converting between Megatron and Hugging Face model formats. This document will guide you on how to convert a trained Megatron model to Hugging Face format for use in other projects.
+MCoreAdapter provides tools for converting between Megatron(McoreAdapter) and Hugging Face model formats. This document will guide you on how to convert a trained Megatron model to Hugging Face format for use in other projects.
 
 ## Conversion Tool
 
-MCoreAdapter includes a conversion tool `tools/convert.py` that can convert Megatron models to Hugging Face format.
+MCoreAdapter includes a conversion tool `tools/convert.py` that can convert Megatron(McoreAdapter) models to Hugging Face format.
 
 ## Conversion Command
 
-To convert a Megatron model to a Hugging Face model, use the following command:
+To convert a model trained with megatron_strategy to a Hugging Face model, use the following command:
 
 ```bash
 python tools/convert.py --checkpoint_path path_to_megatron_model --output_path path_to_output_hf_model
 ```
 
 Where:
-- `path_to_megatron_model` is the path to the Megatron model you want to convert
+- `path_to_megatron_model` is the path to the McoreAdapter model you want to convert
 - `path_to_output_hf_model` is the output path for the converted Hugging Face model
 
 ## Notes
@@ -26,4 +26,4 @@ Where:
 
 ## Direct Use of Hugging Face Models
 
-It's worth noting that MCoreAdapter can also directly load Hugging Face models, so in some cases you can skip the step of converting models to Megatron format. This is particularly useful when you want to use Hugging Face models in Roll framework's reinforcement learning workflows.
+It's worth noting that MCoreAdapter can directly load Hugging Face models without explicitly converting models to Megatron(McoreAdapter) format. This is particularly useful when you want to use Hugging Face models for reinforcement learning in the Roll framework.

docs_roll/docs/简体中文/使用指南/megatron_convert_2_hf.md

@@ -1,21 +1,21 @@
-# Megatron 模型转换为 Hugging Face 格式
+# MCoreAdapter 模型转换为 Hugging Face 格式
 
-MCoreAdapter 提供了在 Megatron 和 Hugging Face 模型格式之间进行转换的工具。本文档将指导您如何将训练好的 Megatron 模型转换为 Hugging Face 格式，以便在其他项目中使用。
+MCoreAdapter 提供了在 Megatron(McoreAdapter) 和 Hugging Face 模型格式之间进行转换的工具。本文档将指导您如何将训练好的 Megatron(McoreAdapter) 模型转换为 Hugging Face 格式，以便在其他项目中使用。
 
 ## 转换工具
 
-MCoreAdapter 包含一个转换工具 `tools/convert.py`，可以将 Megatron 模型转换为 Hugging Face 格式。
+MCoreAdapter 包含一个转换工具 `tools/convert.py`，可以将 Megatron(McoreAdapter) 模型转换为 Hugging Face 格式。
 
 ## 转换命令
 
-要将 Megatron 模型转换为 Hugging Face 模型，请使用以下命令：
+由megatron_strategy训练出来的模型，要Hugging Face 模型，请使用以下命令：
 
 ```bash
 python tools/convert.py --checkpoint_path path_to_megatron_model --output_path path_to_output_hf_model
 ```
 
 其中：
-- `path_to_megatron_model` 是您要转换的 Megatron 模型的路径
+- `path_to_megatron_model` 是您要转换的 McoreAdapter 模型的路径
 - `path_to_output_hf_model` 是转换后 Hugging Face 模型的输出路径
 
 ## 注意事项
@@ -26,4 +26,4 @@ python tools/convert.py --checkpoint_path path_to_megatron_model --output_path p
 
 ## 直接使用 Hugging Face 模型
 
-值得注意的是，MCoreAdapter 直接加载 Hugging Face 模型，不需要显式将模型转换为 Megatron 格式的步骤。这在您想将 Hugging Face 模型用于 Roll 框架的强化学习时特别有用。
+值得注意的是，MCoreAdapter 直接加载 Hugging Face 模型，不需要显式将模型转换为 Megatron(McoreAdapter) 格式的步骤。这在您想将 Hugging Face 模型用于 Roll 框架的强化学习时特别有用。