[docs] update megatron docs#9249
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds PR submission guidelines and updates the documentation index in both Chinese and English. Feedback points out that the referenced 'Megatron-SWIFT/Custom-Model.md' file is missing, which will result in broken links. It is also recommended to update the pre-commit instructions to include 'pre-commit install' for setting up automated git hooks.
| Megatron-SWIFT/GRPO.md | ||
| Megatron-SWIFT/GKD.md | ||
| Megatron-SWIFT/Ascend.md | ||
| Megatron-SWIFT/Custom-Model.md |
There was a problem hiding this comment.
The file Megatron-SWIFT/Custom-Model.md is being added to the table of contents, but this file does not appear to be included in the pull request's changes or existing files. Please ensure that this file exists in the correct path to avoid broken links in the documentation.
| Megatron-SWIFT/GRPO.md | ||
| Megatron-SWIFT/GKD.md | ||
| Megatron-SWIFT/Ascend.md | ||
| Megatron-SWIFT/Custom-Model.md |
There was a problem hiding this comment.
The file Megatron-SWIFT/Custom-Model.md is being added to the table of contents, but this file does not appear to be included in the pull request's changes or existing files. Please ensure that this file exists in the correct path to avoid broken links in the documentation.
|
/gemini review |
There was a problem hiding this comment.
Code Review
This pull request introduces a new documentation guide, "Megatron-SWIFT Custom Model," in both Chinese and English, detailing the process of registering and training custom models using Mcore-Bridge. It also updates existing registration guides with instructions for using pre-commit hooks before submitting pull requests. Review feedback suggests replacing hardcoded local file paths and specific checkpoint directories with generic placeholders in the shell command examples to improve clarity and usability for end-users.
| NPROC_PER_NODE=2 \ | ||
| CUDA_VISIBLE_DEVICES=0,1 \ | ||
| megatron sft \ | ||
| --model /root/.cache/modelscope/models/MiniMax/MiniMax-M2.7 \ |
There was a problem hiding this comment.
It is recommended to use a placeholder like <model_path> instead of a hardcoded local path to make the example more general and easier for users to follow.
| --model /root/.cache/modelscope/models/MiniMax/MiniMax-M2.7 \ | |
| --model <model_path> \ |
| ```shell | ||
| CUDA_VISIBLE_DEVICES=0 \ | ||
| swift infer \ | ||
| --model megatron_output/v3-20260430-143926/checkpoint-310 \ |
There was a problem hiding this comment.
The checkpoint path is hardcoded to a specific timestamped directory. Using a placeholder like <checkpoint_path> would be clearer for users.
| --model megatron_output/v3-20260430-143926/checkpoint-310 \ | |
| --model <checkpoint_path> \ |
| NPROC_PER_NODE=2 \ | ||
| CUDA_VISIBLE_DEVICES=0,1 \ | ||
| megatron sft \ | ||
| --model /root/.cache/modelscope/models/MiniMax/MiniMax-M2.7 \ |
There was a problem hiding this comment.
It is recommended to use a placeholder like <model_path> instead of a hardcoded local path to make the example more general and easier for users to follow.
| --model /root/.cache/modelscope/models/MiniMax/MiniMax-M2.7 \ | |
| --model <model_path> \ |
| ```shell | ||
| CUDA_VISIBLE_DEVICES=0 \ | ||
| swift infer \ | ||
| --model megatron_output/v3-20260430-143926/checkpoint-310 \ |
There was a problem hiding this comment.
The checkpoint path is hardcoded to a specific timestamped directory. Using a placeholder like <checkpoint_path> would be clearer for users.
| --model megatron_output/v3-20260430-143926/checkpoint-310 \ | |
| --model <checkpoint_path> \ |
2 participants
No description provided.