unslothai/unsloth

[Bug] Tool_Calling does not work properly

Open

#3,092 opened on Aug 5, 2025

View on GitHub
 (3 comments) (0 reactions) (0 assignees)Python (64,271 stars) (5,658 forks)batch import
help wanted

Description

I followed this notebook: https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Qwen2.5_Coder_(1.5B)-Tool_Calling.ipynb

Can I only use qwen for tool calling? When I use llama 3.2 1B, when executing the Function for Generation Constraint, I get NotImplementedError: Tokenizer not supported: PreTrainedTokenizerFast

When using phi-4-mini, I get ValueError: The following model_kwargs are not used by the model: ['num_logits_to_keep'] (note: types in the generate arguments will also be displayed in this list).

How can I make it run smoothly?

Contributor guide

Tech stack
python
Domain
machine learningai
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
needs investigation
Prerequisites
basic Python knowledgefamiliarity with LLM tool calling
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
40
Research direction
The issue reports NotImplementedError and ValueError when using Llama 3.2 and Phi 4 mini with unsloth's tool calling. First, verify the error by running the provided Colab notebook with those models. Then, check unsloth's tokenizer support for these models in the source code, particularly in the file 'unsloth/models/ utils.py' or similar. Look for existing discussions or PRs regarding PreTrainedTokenizerFast or num logits to keep. Also, confirm if the notebook is intended only for Qwen models. Report findings on the issue thread.