You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After introducing free translation via "traditional" Google Translator API, there's also a possibility to leverage on-device translation capability: by MLKit https://pub.dev/packages/google_mlkit_translation
(This points towards a long term "wholesome" off-line capability. I was able to run Gemma 2b 16FP 4bit it with MLKit on device, however there's a long road ahead for these models to be as capable regarding function calling, structured output and other features of larger closed SOTA models liek Gemini. RecurrentGemma may help somewhat. On top of that we MMPA is multi modal, so we'd need at least PaliGemma, but we still haven't covered a re-ranking model. Gemma might be able to act as a reranker to some extent.)
The text was updated successfully, but these errors were encountered:
After introducing free translation via "traditional" Google Translator API, there's also a possibility to leverage on-device translation capability: by MLKit
https://pub.dev/packages/google_mlkit_translation
(This points towards a long term "wholesome" off-line capability. I was able to run Gemma 2b 16FP 4bit it with MLKit on device, however there's a long road ahead for these models to be as capable regarding function calling, structured output and other features of larger closed SOTA models liek Gemini. RecurrentGemma may help somewhat. On top of that we MMPA is multi modal, so we'd need at least PaliGemma, but we still haven't covered a re-ranking model. Gemma might be able to act as a reranker to some extent.)
The text was updated successfully, but these errors were encountered: