← PyTorch learning pathPyTorch · Module 8 of 10
NLP and Transformers with PyTorch
Use Hugging Face Transformers on top of PyTorch.
📖 Key ideas
- Hugging Face Transformers provides thousands of PyTorch models.
- AutoTokenizer and AutoModel load models by name.
- Trainer API simplifies fine-tuning.
- Datasets library loads and processes text data.
- PEFT/LoRA fine-tunes large models cheaply.
💡 Real-life example: Hugging Face is a giant app store of pretrained language brains ready for PyTorch.
AutoTokenizerLoads matching tokenizer
AutoModelLoads a pretrained model
TrainerHigh-level training API
PEFTParameter-efficient fine-tuning