Fix Deepseek OCR Lora Model Load

[mmathew23]

Most models types have logic to load the model directly from a peft_config (only saved lora adapters and tokenizer). Deepseek OCR since it requires trust_remote_code=True fails in AutoConfig.from_pretrained since trust_remote_code isn't passed. This PR adds a default kwarg for it that can be explicitly set. There is an accompanying PR in unsloth to enable this feature.

Additionally, a common failure mode is missing package, but that error get swallowed up. This will explicitly raise if an ImportError happens.

This notebook: https://colab.research.google.com/drive/1rQcpPJzFpTId-7KGSvU2zDIg2-emUJTd?usp=sharing
shows regular deepseek ocr works with new code.

Thie notebook: https://colab.research.google.com/drive/1bADSkRQuLQ7VkAV1XlCOH_38nqPhMPye?usp=sharing
shows that direct loading of a save lora_model now works with new code.

[gemini-code-assist]

Summary of Changes

Hello @mmathew23, 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 a critical issue preventing the correct loading of models like Deepseek OCR that rely on trust_remote_code=True by ensuring this flag is properly handled during configuration loading. Additionally, it significantly improves the diagnostic capabilities for common model loading failures, offering clearer error messages for missing dependencies or unsupported architectures.

Highlights

• Deepseek OCR Model Loading: Enabled proper loading of Deepseek OCR models and other models requiring trust_remote_code=True by passing this parameter to AutoConfig.from_pretrained.
• Improved Error Handling: Enhanced error reporting for AutoConfig.from_pretrained to provide more specific feedback for missing packages (ImportError) and architecture-related issues, preventing silent failures.

🧠 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]

Code Review

This pull request effectively addresses the loading issue with Deepseek OCR models by correctly passing trust_remote_code=True. The addition of more specific error handling for missing packages and unsupported architectures is a great improvement, making the library more user-friendly. I've added one suggestion to further improve the robustness of the error handling by logging unexpected exceptions instead of swallowing them silently. Overall, this is a solid contribution.

[gemini-code-assist]

While the new exception handling is a great improvement for specific, known errors, it still silently swallows all other exceptions by setting config = None. This can hide underlying issues and make debugging harder, as the failure will likely occur later with a less informative error message (e.g., a TypeError because config is None). It would be more robust to log these unexpected errors to provide visibility into what went wrong.

Suggested change

	if "architecture" in autoconfig_error:
	raise ValueError(
	f"`{base_model_name_or_path}` is not supported yet in `transformers=={transformers_version}`.\n"
	f"Please update transformers via `pip install --upgrade transformers` and try again."
	)
	elif "No module named" in autoconfig_error:
	raise error
	config = None
	if "architecture" in autoconfig_error:
	raise ValueError(
	f"`{base_model_name_or_path}` is not supported yet in `transformers=={transformers_version}`.\n"
	f"Please update transformers via `pip install --upgrade transformers` and try again."
	)
	elif "No module named" in autoconfig_error:
	raise error
	else:
	logger.warning(
	f"Unsloth: Failed to get config for {base_model_name_or_path} "
	f"due to an unknown error: {error}. "
	f"Proceeding without it."
	)
	config = None

[danielhanchen]

Suggested change

	def get_transformers_model_type(config, trust_remote_code=False):
	def get_transformers_model_type(config, trust_remote_code = False):