🤗

Hugging Face Transformers Learning Roadmap

Pretrained transformer models — fine-tuning, deployment, and the broader HF ecosystem.

Estimated time:1–2 months

1🌱

Phase 1 — Fundamentals

Tokenizers & pretrained models

pipeline() for quick inference

Model Hub navigation

AutoModel/AutoTokenizer classes

2🏋️

Trainer API for fine-tuning

Datasets library for data loading

PEFT / LoRA for efficient fine-tuning

Evaluation metrics (evaluate library)

3🧩

Accelerate for multi-GPU training

Quantization (bitsandbytes)

Text Generation Inference (TGI) server

4🚀

important

Inference Endpoints / self-hosted serving

Batching & latency optimization

Model versioning on the Hub

Cost/latency tradeoffs for LLM serving

Begin with interview questions to assess your current level.