An abliterated (uncensored) version of Google’s Gemma 4 E2B Instruct, converted to LiteRT format (.litertlm) for on-device inference. — Text generation — instruction-tuned chat — Vision — image understanding — Audio — audio input processing — Thinking — extended reasoning — Tool use — function/tool calling Refusal behaviours have been removed via abliteration (see below). The model will respond to requests that the base model would decline. The package contains INT8 weight-quantized TFLite subgraphs (4x compression from original BF16 weights): Drop model.litertlm into your app and load it with the LiteRT-LM LlmInference API: The refusal direction was identified and removed from the model weights using the FailSpy abliterator methodology: 1. Forward passes were run on 15 harmful and 15 harmless prompts through the base model 2. The mean difference in hidden states at layer 17 (the probe layer, mid-model) was computed and normalised to produce the refusal direction 3. That direction was projected out of the oproj and downproj weight matrices in layers 11–22 (the middle third of the model’s 35 layers) using: where r is the unit refusal direction. This removes the model’s ability to…
Модальности:
Генерация текста Мультимодальность
Задача: Генерация текста
Автор: brine7302
Теги: litert, litertlm, gemma, gemma4, abliterated, uncensored, multimodal, image-text-to-text
Лайков: 3 | Загрузок: 0
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
