Modelpublic

coders1/quantstudio-gemma4-e4b-quant-dsl-v0-1-lora-gguf

sourceHugging Facegemmaupdated 5mo agoView on Hugging Face

0likes13downloads

Model Card

Quant Studio Gemma 4 E4B Quant DSL v0.1 LoRA GGUF

This repository contains a runtime-loadable llama.cpp LoRA adapter GGUF for Quant Studio Quant DSL v0.1 experiments.

—Repository: coders1/quantstudio-gemma4-e4b-quant-dsl-v0-1-lora-gguf
—Artifact ID: quantdslv01gemma4_gemma4quantdslv01e4b_lora
—Dataset ID: quantdslv01gemma4
—DSL version: Quant DSL v0.1
—Adapter file: quantdslv01gemma4e4bfinetuned_lora.gguf
—Adapter size: 532.28 MiB
—SHA256: f80f71149950e7b2ab606f775454f0bd9508829754c4ff8e6c06f7ea97b687ec

What This Is

This is an adapter-only artifact. It is not a merged model and it is not a standalone checkpoint.

You must load it together with a compatible Gemma 4 E4B base GGUF in llama.cpp.

—Base preset: gemma4-e4b-it-q4km
—Base display name: Gemma 4 E4B Instruct Q4KM
—Tested local base path: D:/Models/gemma-4-e4b-it-Q4KM.gguf
—Adapter scale: 1.0

LoRA Recipe

—Rank: 64
—Alpha: 128
—Dropout: 0.05
—Target modules: qproj, kproj, vproj, oproj, gateproj, upproj, down_proj

llama.cpp Usage

bash

llama-cli -m /path/to/gemma-4-e4b-it-Q4_K_M.gguf --lora /path/to/quant_dsl_v0_1_gemma4_e4b_finetuned_lora.gguf

You can also use the adapter in Quant Studio with:

—Base model downloaded from Hugging Face
—Adapter GGUF downloaded from the Quant Studio artifact server

Notes

—This artifact was exported as a GGUF LoRA adapter for runtime attachment.
—The tested training model family is google/gemma-4-E4B-it.
—A compatible Gemma base model is still required at inference time.
—Gemma terms apply to Gemma derivatives and downstream usage.