Demo: Using Gemma with the Hugging Face ecosystem
summary
Gemma models run in the Hugging Face ecosystem with three lines of code for tokenizer, model load, and generate; Transformers adds performance gains (Flash Attention 2, Torch Compile) making Gemma up to three to four times faster. Training uses Datasets plus PEFT and TRL so you can fine-tune 7B in 16GB by combining 4/8-bit quantization and LoRA, and TRL supplies trainers for SFT, reward modeling, PPO. Models push to the Hub and deploy via inference API, AWS, or Vertex AI; a notebook fine-tunes Gemma 2B in Colab.
AI-written summary · about
chapters (5)
- 0:00Introduction to ecosystem
- 1:46Model inference techniques
- 4:11Data loading strategies
- 5:13Training and fine-tuning
- 7:40Deployment and conclusion
related talks
09mUsing and fine-tuning Gemma 3
39mFlutter Project with Gemma: A State-of-the-Art Open Model
07mWelcome to the Gemmaverse
44mBuilding amazing Front-End Project with Gemma: A State-of-the-Art Open Model
40mBuilding amazing Front-End Project with Gemma: A State-of-the-Art Open Model
10mAI for everyone with Gemma
12mWelcome to Gemma 3
14mGemma on mobile and web. Best and worst practices