Saathi reply-style LoRA (round 3)
A reply-style LoRA for the Saathi journaling companion, replies only. It changes how Saathi's replies read (warm, short, like a close friend texting); it is not used for extraction, memory cards, safety screens or any other task.
- Base model: gemma-4-E2B-it, GGUF Q4_K_M (
unsloth/gemma-4-E2B-it-GGUF,gemma-4-E2B-it-Q4_K_M.gguf) - File:
saathi-lora-r3-f16.gguf(GGUF LoRA adapter, f16), 50,709,632 bytes - SHA-256:
17fc0714c3662e1e664ce21b76f5a4c30d05f153f9513c1fea0e7c329a08fbee - Runtime: llama.cpp (tested with b11185 on a laptop and b11192 through llama.rn on Android), scale 1.0
Use
llama-server -m gemma-4-E2B-it-Q4_K_M.gguf --lora saathi-lora-r3-f16.gguf
The Saathi app downloads this file after install and verifies the SHA-256 before loading it.
Terms
This adapter is a derivative of Gemma and is distributed under the Gemma Terms of Use. Use is subject to the Gemma Prohibited Use Policy.
Saathi is an AI companion, not a therapist, medical or crisis service.
- Downloads last month
- -
Hardware compatibility
Log In to add your hardware
16-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support