Welcome Gemma - Google’s new open LLM

Hugging Face Blog · 2 years ago

Google released Gemma, a family of four open-source language models in 2B and 7B parameter sizes, each available in base and instruction-tuned variants. The 7B model achieves performance comparable to Mistral 7B on standard benchmarks, while the 2B version is designed for deployment on consumer hardware and mobile devices. The models are integrated across Hugging Face, Google Cloud, and can be fine-tuned on consumer GPUs using techniques like QLoRA, with a single A10G GPU requiring approximately 9 hours to fine-tune on standard datasets.