Gemma Open Models from Google

Of Late we are getting many queries around what are Google’s open models beyond Google clouds Foundation models which require huge infrastructure to host it. The answer is Gemma 2B and 7B. Let’s try to decode Gemma.

In this blog we will start with Gemma much talked about open models from Google.It is one of the most sought open models in Google cloud.

Gemma derives its name from the latin word gemma meaning “precious stone”. It is the family of lightweight, state of the art (SOTA) open models powered by the same technology powering one of the most popular Google cloud Gemini models.

Gemma models are available in different sizes and capabilities.

• Gemma 2B ~ 2B parameter size — modality support with text use cases.
• Gemma 7B ~ 7B parameter size — modality support with text use cases.

Each of the above models is available in tuned and untuned versions.

• Pretained ~ This model is not trained on any specific tasks. Will require some tuning to have effective results.
• Instruction tuned ~ This model was trained with human language. Quite effective.

Benefits of Gemma ~ Its lower parameters sizes mean lower resource requirements and can be run even in small environments like laptops and workstations.

Gemma versus Gemini ~ Gemini is available in different variants i.e. 1, 1.5 etc very ideal for use cases for enterprise needs which requires…