Add automatic configuration generation for the draft model.#167
Add automatic configuration generation for the draft model.#167sleepcoo merged 9 commits intosgl-project:mainfrom
Conversation
ZhengHSI
requested review from
FlamingoPg,
shuaills and
sleepcoo
as code owners
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Summary of Changes
Hello @ZhengHSI, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request significantly streamlines the setup process for training Eagle3 models by introducing automatic draft model configuration generation. Previously, users were required to manually provide a draft model configuration file. Now, if this file is not specified, the system intelligently creates one by aligning parameters from the target model, thereby reducing manual overhead and potential configuration errors. This enhancement is reflected in both offline and online training scripts and is supported by new utility functions for configuration management.
Highlights
- Automatic Draft Model Configuration: The primary change introduces an automatic configuration generation mechanism for draft models. If the
--draft-model-configargument is not explicitly provided during training, the system will now automatically derive and create a suitable draft model configuration based on the target model's configuration. - Optional
--draft-model-configArgument: The--draft-model-configargument in bothtrain_eagle3_offline.pyandtrain_eagle3_online.pyscripts has been updated to be optional. This enhances usability by reducing the need for manual configuration file creation. - New Configuration Utility Functions: New utility functions (
generate_draft_model_config,save_draft_model_config,create_draft_config_from_target) have been added tospecforge/utils.pyto encapsulate the logic for creating, saving, and managing these auto-generated configurations. - Documentation Update: The
README.mddocumentation has been updated to reflect the new behavior, guiding users on how to leverage the automatic configuration generation feature.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
This pull request adds a convenient feature for automatically generating draft model configurations. The implementation is mostly solid, but there are opportunities to improve robustness and maintainability. Specifically, I've suggested changes to make file path handling more reliable by using __file__ instead of sys.argv[0], simplifying some logic, and refactoring duplicated code into a shared utility function. These changes will make the new feature more robust and the codebase easier to maintain.
| if template_config_path is None: | ||
| # Use the script execution directory as base | ||
| import sys | ||
|
|
||
| script_dir = os.path.dirname(os.path.abspath(sys.argv[0])) | ||
| project_root = os.path.dirname(script_dir) # Go up one level from scripts/ | ||
| template_config_path = os.path.join( | ||
| project_root, "configs", "llama3-8B-eagle3.json" | ||
| ) |
There was a problem hiding this comment.
Using sys.argv[0] to determine the project root can be unreliable as it depends on how and from where the script is executed. A more robust approach is to use __file__ to get a path relative to the current module's location, which is independent of the working directory.
if template_config_path is None:
# Use the path of this file to robustly locate the project root
# Assuming this file is in specforge/, two levels up is the project root.
project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
template_config_path = os.path.join(
project_root, "configs", "llama3-8B-eagle3.json"
)| if output_dir is None: | ||
| # Use the script execution directory as base | ||
| import sys | ||
|
|
||
| script_dir = os.path.dirname(os.path.abspath(sys.argv[0])) | ||
| project_root = os.path.dirname(script_dir) # Go up one level from scripts/ | ||
| output_dir = os.path.join(project_root, "configs") |
There was a problem hiding this comment.
Using sys.argv[0] to determine the project root can be unreliable as it depends on how and from where the script is executed. A more robust approach is to use __file__ to get a path relative to the current module's location, which is independent of the working directory.
if output_dir is None:
# Use the path of this file to robustly locate the project root
# Assuming this file is in specforge/, two levels up is the project root.
project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
output_dir = os.path.join(project_root, "configs")
scripts/train_eagle3_offline.py
Outdated
| if args.draft_model_config is None: | ||
| print_with_rank( | ||
| "No draft model config provided, auto-generating from target model..." | ||
| ) | ||
| # Auto-generate and save config file | ||
| auto_config_path = create_draft_config_from_target( | ||
| target_model_path=args.target_model_path, cache_dir=args.cache_dir | ||
| ) | ||
| draft_model_config = AutoDraftModelConfig.from_file(auto_config_path) | ||
| print_with_rank( | ||
| f"Auto-generated draft model config saved to: {auto_config_path}" | ||
| ) | ||
| else: | ||
| # Use provided config file | ||
| draft_model_config = AutoDraftModelConfig.from_file(args.draft_model_config) | ||
| print_with_rank(f"Using provided draft model config: {args.draft_model_config}") |
There was a problem hiding this comment.
This logic for handling the draft model configuration is duplicated in scripts/train_eagle3_online.py (lines 170-185). To improve maintainability and follow the DRY (Don't Repeat Yourself) principle, consider refactoring this block into a shared utility function.
For example, you could create a function load_or_create_draft_config(args) in a utility module that encapsulates this logic and returns the draft_model_config.
specforge/utils.py
Outdated
| if target_param == "torch_dtype": | ||
| if hasattr(value, "__name__") or "torch" in str(type(value)): | ||
| value_str = str(value) | ||
| if "torch." in value_str: | ||
| value = value_str.split("torch.")[ | ||
| -1 | ||
| ] # Convert torch.float16 to 'float16' | ||
| else: | ||
| value = value_str.split(".")[-1] |
There was a problem hiding this comment.
The current logic for converting a torch.dtype to its string representation is complex and can be simplified for better readability and maintainability. Using isinstance to check the type and str.replace for conversion is more direct.
if target_param == "torch_dtype" and isinstance(value, torch.dtype):
value = str(value).replace("torch.", "")| output_dir = os.path.join(project_root, "configs") | ||
|
|
||
| # Extract model name from model path | ||
| model_name = target_model_path.split("/")[-1].lower() |
There was a problem hiding this comment.
Using split('/')[-1] to extract the model name is not robust for all path formats, especially local paths or paths with trailing slashes. os.path.basename combined with os.path.normpath provides a more reliable way to handle different path structures.
model_name = os.path.basename(os.path.normpath(target_model_path)).lower()
3 participants
Motivation
Add automatic configuration generation for the draft model.
Modifications
When the
--draft-model-configargument is not provided, the script will automatically generate the draft model configuration based on the target model’s configuration.Related Issues
Accuracy Test
Benchmark & Profiling
Checklist
Co-authored-by: lukec 118525388+sleepcoo@users.noreply.github.com