Skip to content

Clarify NotImplementedError for fast_inference with full_finetuning#3768

Merged
danielhanchen merged 3 commits into
unslothai:mainfrom
Fizza-Mukhtar:improve-fast-inference-error
Dec 26, 2025
Merged

Clarify NotImplementedError for fast_inference with full_finetuning#3768
danielhanchen merged 3 commits into
unslothai:mainfrom
Fizza-Mukhtar:improve-fast-inference-error

Conversation

@Fizza-Mukhtar

Copy link
Copy Markdown
Contributor

This PR improves the error message raised when
fast_inference=True is used together with full_finetuning=True.

The updated message explains why the combination is unsupported
and suggests practical workarounds (disabling fast_inference or using LoRA).

No behavior changes are introduced.

Closes #3577

@gemini-code-assist

Copy link
Copy Markdown
Contributor

Summary of Changes

Hello @Fizza-Mukhtar, 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 focuses on improving the user experience by providing a more informative and actionable error message within the unsloth library. It addresses a specific scenario where fast_inference and full_finetuning are used together, guiding users on why this configuration is not supported and offering clear alternatives, without introducing any functional changes.

Highlights

  • Improved Error Message: The pull request enhances the error message displayed when fast_inference=True is used in conjunction with full_finetuning=True.
  • Clarity and Workarounds: The updated message now clearly explains why this combination is unsupported and provides practical workarounds, such as disabling fast_inference or utilizing LoRA (with a suggested rank).
  • No Behavior Change: This change is purely informational and does not alter any existing behavior or functionality within the system.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

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 by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

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 pull request 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

  1. 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.

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown
Contributor

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 significantly improves the error message for an unsupported configuration, making it much clearer for users by explaining the reason and providing actionable workarounds. The change is a great enhancement to user experience. I've included one suggestion to further refine the string formatting for better readability and to remove an unnecessary blank line.

Comment thread unsloth/models/vision.py
Comment on lines 720 to +727
raise NotImplementedError(
f"Unsloth: `fast_inference = True` does not yet support `full_finetuning = True`.\n"
f"Use LoRA rank `r = {max_lora_rank}` as the closest replacement for full finetuning with Unsloth for RL."
f"`fast_inference=True` cannot be used together with `full_finetuning=True`.\n"
f"Reason: fast_inference is optimized for inference-only workflows and "
f"does not currently support full fine-tuning.\n"
f"Workaround: disable fast_inference, or use parameter-efficient fine-tuning "
f"(e.g. LoRA with rank r={max_lora_rank})."
)

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The improved error message is great for user experience. For better readability and to adhere to standard Python conventions, you can simplify the string formatting. Using implicit string concatenation and removing f-string prefixes from lines without variables makes the code cleaner. This change also removes the extra blank line after the raise statement.

Suggested change
raise NotImplementedError(
f"Unsloth: `fast_inference = True` does not yet support `full_finetuning = True`.\n"
f"Use LoRA rank `r = {max_lora_rank}` as the closest replacement for full finetuning with Unsloth for RL."
f"`fast_inference=True` cannot be used together with `full_finetuning=True`.\n"
f"Reason: fast_inference is optimized for inference-only workflows and "
f"does not currently support full fine-tuning.\n"
f"Workaround: disable fast_inference, or use parameter-efficient fine-tuning "
f"(e.g. LoRA with rank r={max_lora_rank})."
)
raise NotImplementedError(
"`fast_inference=True` cannot be used together with `full_finetuning=True`.\n"
"Reason: fast_inference is optimized for inference-only workflows and "
"does not currently support full fine-tuning.\n"
"Workaround: disable fast_inference, or use parameter-efficient fine-tuning "
f"(e.g. LoRA with rank r={max_lora_rank})."
)

Comment thread unsloth/models/vision.py Outdated
@danielhanchen

Copy link
Copy Markdown
Member

This is a fair change - thanks for this!

@danielhanchen danielhanchen merged commit d83fbf6 into unslothai:main Dec 26, 2025
1 check passed
@Fizza-Mukhtar Fizza-Mukhtar deleted the improve-fast-inference-error branch December 26, 2025 06:03
abiswas-realadvice pushed a commit to abiswas-realadvice/unsloth that referenced this pull request May 14, 2026
…nslothai#3768)

* Improve error message for fast_inference and full_finetuning

* Refine error message string formatting

* Update unsloth/models/vision.py

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Feature] NotImplementedError: Unsloth: fast_inference = True does not yet support full_finetuning = True

2 participants