●Change8

Migrating to Transformers v4.51.0

Version v4.51.0 introduces 1 breaking change. This guide details how to update your code.

Released: 4/5/2025

1
Breaking Changes
5
Migration Steps
9
Affected Symbols

⚠️ Check Your Code

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

Llama4ForConditionalGenerationAutoProcessorPhi4-multimodal-instructDeepSeek-V3Qwen3Qwen3MoEGemma3Chameleonmllama

Breaking Changes

●Issue #1

The introduction of Llama 4 models, which are natively multimodal, requires updating the installation to include necessary dependencies for handling image inputs.

✓Solution

Ensure you have transformers v4.51.0 or later installed with the optional dependency for extended tensor operations: 'pip install -U transformers[hf_xet]'.

Migration Steps

  1. 1
    Upgrade your 'transformers' library to version 4.51.0 or newer by running: pip install -U transformers[hf_xet].
  2. 2
    If migrating from older Llama models, note that Llama 4 models (like Maverick and Scout) support native multimodality (text and image inputs).
  3. 3
    When loading Llama 4 models, ensure your environment meets the hardware requirements (e.g., Llama 4 Maverick may require multi-GPU setups for efficient tensor parallelism).
  4. 4
    Review the specific model cards for Llama 4 Maverick and Llama 4 Scout for details on the Llama 4 Community License Agreement before deployment.
  5. 5
    If using Phi-4-multimodal-instruct, be aware of its specific language support across text, vision (English only), and audio modalities.

Release Summary

This release introduces support for Llama 4, Phi4-Multimodal, DeepSeek-v3, and Qwen3 architectures, alongside a major documentation overhaul and modularization of speech models.

Need More Details?

View the full release notes and all changes for Transformers v4.51.0.

View Full Changelog