Description

A recent commit introduced checks for different file extensions (.exe and non-.exe) to support both Linux and Windows environments. However, this change results in a RuntimeError when the required files (llama-quantize, llama-quantize.exe, quantize, or quantize.exe) are not found in the expected directories, causing the llama.cpp installation to fail in Google Colab.

Commit Details

• Commit: 222b1e7
• Author: Gennadii Manzhos
• Date: 2025-02-16
• Changes:
  • Fixed llama-quantize error on Windows WSL by modifying save.py (fixes GGUF saving issue).
  • Added extension checks for .exe and non-.exe files for Linux and Windows.

Pull Request Details

• PR: Gemma 3 Bug Fixes
• Author: Daniel Han
• Merged by: Daniel Han
• Files Changed: 7
• Additions: 224
• Deletions: 115
• State: Merged

Error Message

RuntimeError: Unsloth: The file ('llama.cpp/llama-quantize' or 'llama.cpp/llama-quantize.exe' if you are on Windows WSL) or 'llama.cpp/quantize' does not exist. But we expect this file to exist! Maybe the llama.cpp developers changed the name or check extension of the llama-quantize file.

Steps to Reproduce

1. Install llama.cpp in a Google Colab environment.
2. Ensure that llama-quantize or quantize files are missing or renamed.
3. The installation process fails with a RuntimeError.

Expected Behavior

The installation process should complete successfully when the required files are correctly identified and present.

Actual Behavior

The installation fails with a RuntimeError due to missing or misidentified llama-quantize or quantize files in Google Colab.

Additional Information

For reference, please see:

• Commit 222b1e7
• PR 2005

Suggested Fix

Consider modifying the file detection logic to accommodate different execution environments, particularly for Google Colab. A fallback mechanism or a clear warning message with resolution steps would improve usability.