●Change8

Migrating to Transformers v4.55.1

Version v4.55.1 introduces 1 breaking change. This guide details how to update your code.

Released: 8/13/2025

1
Breaking Changes
4
Migration Steps
10
Affected Symbols

⚠️ Check Your Code

If you use any of these symbols, you need to read this guide:

Idefics2Idefics3SmolVLMGPTBigCodeGemma3FuyuPerceptionLMIdeficsMXFP4get_placeholder_mask

Breaking Changes

●Issue #1

The dependency on the external `triton_kernels` package has been removed, as these kernels are now included internally.

✓Solution

If your project explicitly depended on `triton_kernels`, you should remove it from your project's dependencies (e.g., `requirements.txt` or `setup.cfg`).

Migration Steps

  1. 1
    Update your library to version 4.55.1 or later.
  2. 2
    Review your project dependencies and remove any explicit dependency on the `triton_kernels` package, as it is now bundled internally.
  3. 3
    If you were using MXFP4 quantization and encountered issues with CPU inference, ensure you are using the latest configuration, as the default behavior now favors dequantization when the CPU is present in `device_map`.
  4. 4
    If you are using Gemma3n models, ensure your code handles potential `None` values correctly for `get_placeholder_mask` if you were relying on previous implicit behavior, although this fix primarily addresses internal defaults.

Release Summary

Patch release 4.55.1 focuses on stabilizing the MXFP4 quantization for GPT-OSS models and resolving device-related bugs across several multimodal models like Idefics and SmolVLM.

Need More Details?

View the full release notes and all changes for Transformers v4.55.1.

View Full Changelog