Sync huggingface-transformers with the updated tutorial

[realpython-bot]

Maintenance sync for Hugging Face Transformers: Leverage Open-Source AI in Python. The article's update draft re-pins its dependencies to Transformers 5.

Changes

• requirements.txt: transformers 4.42.4 → 5.19.0, torch 2.3.1 → 2.14.1, accelerate 0.32.1 → 1.15.0, polars 1.2.0 → 2.0.0. Added tqdm==4.70.1 because the notebook imports it directly, and pillow==12.3.0 because the image pipeline needs it.
• pyproject.toml: bumped the same constraints, plus pillow ^12.3.0, notebook ^7.6.3 and ipywidgets ^8.1.9. The Python floor goes from ^3.12 to ^3.10, the real minimum: every pinned package needs >=3.10, and vermin reports 3.6 for the code.
• poetry.lock: regenerated with Poetry 2.5.1.
• README.md: pinned the pip install line to the tested versions.
• running_pipelines.py: changed "HuggingFace" to "Hugging Face" in the first example sentence. The script used different text than the article, so it printed a different score.

Verification (Python 3.14, CPU torch, fresh venv from requirements.txt alone)

• auto_classes.py and running_pipelines.py run cleanly. Every printed value now matches the updated article, including the first sentiment score and the llama prediction.
• I ran the notebook's non-GPU cells (CSV load with Polars 2, time_text_classifier() on a CPU pipeline). I couldn't run the GPU cells because there's no GPU here.
• ruff@0.14.1 format --check and ruff check pass.

Links

• Article draft: https://realpython.com/pbjt/blog/post/2514/change/
• Maintenance card: https://trello.com/c/oRrUFWQU

🤖 Generated with Claude Code