Transformers

Transformers is Hugging Face's library that made open-weights AI practical: one consistent API (AutoModel, AutoTokenizer, pipeline()) to download and run hundreds of thousands of checkpoints from the Hub — LLMs, vision-language models, audio, and more — across PyTorch, TensorFlow, and JAX. It standardizes tokenization, training loops (Trainer), quantization, and deployment formats, so code written against one model usually works against the next. If you self-host open models, you almost certainly depend on this library directly or transitively.

```bash

pip install transformers

```

```python

from transformers import pipeline

generator = pipeline("text-generation", model="Qwen/Qwen2.5-7B-Instruct")

print(generator("Explain gravity in one sentence.")[0]["generated_text"])

```

Add accelerate for multi-GPU/big-model support. Full docs: huggingface.co/docs/transformers.