feat(genai): Add tuning examples #13360
Conversation
|
Here is the summary of changes. You are about to add 4 region tags.
This comment is generated by snippet-bot.
|
There was a problem hiding this comment.
Hello @yishan-pu, 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!
Summary of Changes
This pull request adds examples for tuning generative AI models using the google-genai library. It includes code snippets for creating, getting, listing, and testing tuning jobs. The examples demonstrate how to interact with the Gemini 2.0 family of models for fine-tuning. The PR also includes necessary test files and updates to the project's dependencies.
Highlights
- Tuning Examples: Adds examples for creating, getting, listing, and testing tuning jobs using the
google-genailibrary. - New Files: Introduces new Python files (
tuning_create.py,tuning_get.py,tuning_list.py,tuning_test.py,test_tuning_examples.py) to showcase the tuning functionality. - Dependencies: Updates
requirements.txtandrequirements-test.txtto include necessary dependencies likegoogle-genai,google-api-core,google-cloud-storage, andpytest.
Changelog
Click here to see the changelog
- genai/tuning/noxfile_config.py
- Added
noxfile_config.pyto configure the test environment, ignoring Python versions 2.7, 3.7, 3.8, 3.10, 3.11 and 3.12, enforcing type hints, and setting the Google Cloud project environment variable.
- Added
- genai/tuning/requirements-test.txt
- Added
requirements-test.txtto specify testing dependencies, includinggoogle-api-core,google-cloud-storage, andpytest.
- Added
- genai/tuning/requirements.txt
- Added
requirements.txtto specify thegoogle-genaidependency.
- Added
- genai/tuning/test_tuning_examples.py
- Added
test_tuning_examples.pyto include tests for the tuning examples, mocking thegoogle.genai.Clientand testing thecreate,get,list, andtestfunctions.
- Added
- genai/tuning/tuning_create.py
- Added
tuning_create.pyto demonstrate how to create a tuning job, including setting the base model, training dataset, and tuning job configuration.
- Added
- genai/tuning/tuning_get.py
- Added
tuning_get.pyto demonstrate how to get a tuning job by name.
- Added
- genai/tuning/tuning_list.py
- Added
tuning_list.pyto demonstrate how to list all tuning jobs.
- Added
- genai/tuning/tuning_test.py
- Added
tuning_test.pyto demonstrate how to test a tuned endpoint.
- Added
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 is currently in preview and 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 to provide feedback.
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.
A model's potential,
Hidden deep, waiting to bloom,
Tuning sets it free.
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
The pull request introduces examples for tuning jobs using the google-genai SDK. The examples cover creating, getting, listing, and testing tuning jobs. The code appears well-structured and includes tests. However, there are a few areas that could be improved for clarity and robustness.
Summary of Findings
- Error Handling: The examples lack explicit error handling. While they print the state of the tuning job, they don't handle potential exceptions or API errors, which could lead to unexpected behavior in real-world scenarios.
- Input Validation: The
tuning_get.py,tuning_list.py, andtuning_test.pyexamples take user input for the tuning job name. There's no validation on this input, which could lead to errors if the user provides an invalid name. - Test Coverage: The tests mock the API responses, which is good for isolating the code. However, the tests could be expanded to cover more edge cases and error scenarios.
Merge Readiness
The pull request introduces valuable examples for tuning jobs. However, the lack of error handling and input validation are concerns that should be addressed before merging. I recommend addressing the review comments to improve the robustness and user-friendliness of the examples. I am unable to directly approve this pull request, and recommend that others review and approve this code before merging.
- Updates tuning file names to fix cicd - word `test` is reserved for test files - Update region tag and file names. - add word `job` to filenames & region tags
- rename methods and function according to prior changes
|
|
||
| tuning_job = client.tunings.tune( | ||
| base_model="gemini-2.0-flash-lite-001", | ||
| training_dataset="gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/text/sft_train_data.jsonl", |
There was a problem hiding this comment.
Avoid using model names in filename.
- OK: Using
gemini - Not OK: Using
gemini-2.0orgemini-2.0-flash-lite-001
| training_dataset="gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/text/sft_train_data.jsonl", | |
| training_dataset="gs://cloud-samples-data/ai-platform/generative_ai/gemini/text/sft_train_data.jsonl", |
There was a problem hiding this comment.
That's a good point. Using a more generic path makes the example more robust against future model name changes.
| training_dataset="gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/text/sft_train_data.jsonl", | |
| training_dataset="gs://cloud-samples-data/ai-platform/generative_ai/gemini/text/sft_train_data.jsonl", |
There was a problem hiding this comment.
Thank you. We would like to align the file paths with the tuning public documentations. We used model names as folders because the other folders under gs://cloud-samples-data/ai-platform/generative_ai/ include other tuning formats, and the model names are necessary to distinguish between the different formats.
Description
Fixes b/416471329
Note: Before submitting a pull request, please open an issue for discussion if you are not associated with Google.
Checklist
nox -s py-3.9(see Test Environment Setup)nox -s lint(see Test Environment Setup)