-
Notifications
You must be signed in to change notification settings - Fork 246
Fix assembling URLs for big models #1756
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Reviewer's GuideEncapsulate split-model file assembly into a dedicated helper and correct the file-part enumeration and naming to include the first part and avoid overwrites. Class diagram for updated split-model file assemblyclassDiagram
class ModelHandler {
+_assemble_split_file_list(snapshot_hash: str) list[SnapshotFile]
+pull(_)
}
class SnapshotFile {
+url: str
+header: dict
+hash: str
+type: SnapshotFileType
+name: str
+should_show_progress: bool
+required: bool
}
ModelHandler --> SnapshotFile : creates
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Summary of Changes
Hello @engelmi, 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 resolves critical issues in how the application handles the downloading of large, multi-part models. By refactoring the URL assembly and file naming conventions, it ensures that all model segments are correctly retrieved and stored, preventing previously reported problems of missing files and data corruption due to overwrites.
Highlights
- Corrected URL Assembly for Split Models: The logic for generating URLs and filenames for multi-part models (e.g.,
.gguf
files split into several parts) has been refactored. This change ensures that all necessary parts are correctly identified and included, starting from the first part (e.g.,00001-of-00005.gguf
), and that each part is assigned its unique filename to prevent overwrites during the download process. - New Helper Method for Modularity: Introduced a new private helper method,
_assemble_split_file_list
, withinramalama/url.py
. This method encapsulates the complex logic for generating the list ofSnapshotFile
objects for split models, improving code modularity, readability, and maintainability by centralizing the file assembly process.
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 or fill out our survey 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.
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.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Code Review
This pull request fixes an issue with downloading large, split models by ensuring all files are downloaded with the correct names. The logic has been refactored into a new private method _assemble_split_file_list
, improving code organization. My review includes one suggestion to improve efficiency by avoiding calling _assemble_split_file_list
when it's not needed.
fee6a10
to
97de0d3
Compare
Previously, the loop to assemble the split model files was missing the first file of the list due to using the iteration variable incorrectly. In addition, all files had been assigned the same name, which led to overwriting the files. In order to ensure that the urls and names are assembled correctly unit tests have been added as well. Signed-off-by: Michael Engel <[email protected]>
97de0d3
to
acc63af
Compare
Fixes: #1755 |
Previously, the loop was missing the first file of a big model and using the same name for all models. This has been fixed in this PR.
For example, previously running
would result in the following files to be downloaded (missing the 00001):
And each file would have the same name of
Qwen3-Coder-480B-A35B-Instruct-Q3_K_M-00001-of-00005.gguf
, causing previously downloaded files to be overwritten.With this fix, it correctly downloads the following files with their respective names:
Summary by Sourcery
Correctly assemble URLs and filenames for split model files to include the first part and prevent overwriting by generating unique names for each part
Bug Fixes:
Enhancements:
_assemble_split_file_list
method