OvermaxSA/asl-translator-gemma-3-1b-it

Gemma 3 1B IT (int4, LiteRT-LM) for ASL Translator

This repository hosts Google's Gemma 3 1B instruction-tuned model, int4 quantized, in LiteRT-LM format (.litertlm). It is the optional on-device model that the Android app ASL Translator: Sign Language by Overmax downloads (only if the user chooses to, and only over Wi-Fi) to turn recognized American Sign Language signs and fingerspelled words into a sentence. Everything runs on the phone; nothing is sent to a server.

File	Description
gemma3-1b-it-int4.litertlm	The model in LiteRT-LM format (see the modification notice below)
NOTICE	Gemma notice required by the Gemma Terms of Use
GEMMA_TERMS_OF_USE.txt	Copy of the Gemma Terms of Use
GEMMA_PROHIBITED_USE_POLICY.txt	Copy of the Gemma Prohibited Use Policy
build-recipe.toml, LlmMetadataProto.pbtext	How the .litertlm file was assembled

• Source: Kaggle, google/gemma-3/tfLite/gemma3-1b-it-int4, version 1 (gemma3-1B-it-int4.task).
• SHA-256 of gemma3-1b-it-int4.litertlm: 5a4a58a31425d4158748eed885b580f288bceeb143cf08265dc9e66bf628bd19

Modification notice

Overmax changed the file container, not the model: the .task file published by Google was unpacked and its TF_LITE_PREFILL_DECODE model and TOKENIZER_MODEL were repackaged, byte for byte, into a .litertlm file with litert-lm-builder 0.19.0, using the Gemma 3 chat template metadata from Google's own Gemma 3 .litertlm files. No weights, tokenizer data or model behavior were changed.

License and use restrictions

Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms

By downloading or using this model you agree to the Gemma Terms of Use. The use restrictions in the Gemma Prohibited Use Policy apply to you and to anyone you share the model with. Copies of both documents are included in this repository.