A LoRA fine-tune of gemma-4-E4B-it-qat-q4_0-unquantized trained on a compact, extremely high-quality synthetic conversational dataset derived from the visual novel My Dystopian Robot Girlfriend. The model captures the personality, speech patterns, and emotional nuance of the character Jun while preserving the base model's general reasoning and instruction-following capabilities.

This version is trained around Jun OS's production system prompt and mood conditioning, making it the intended conversational backend for that application.

Model Variants & Repositories

Repository	Format	Description
`efficiencyx/Jun-v4-E4B-Step100-GGUF`	GGUF (F16 / Q8_0 / Q6_K / Q4_K_M)	Quantized versions for local inference
`efficiencyx/Jun-LoRA-v4-E4B-Step100`	LoRA Adapter	Raw LoRA adapter

Quantization Guide

Quant	Size	Use Case
F16	~14.9 GB	Full-precision GGUF, source for the quants
Q8_0	~7.97 GB	Best quality, near-lossless
Q6_K	~6.17 GB	High quality, minimal loss
Q4_K_M	~5.30 GB	Recommended balance; runs on modest hardware

Sizes are measured. The base model is QAT (quantization-aware trained), so lower quants hold up better than a standard FP16 export. Prompt with the model's built-in Gemma 4 chat template.

Intended Use

This model is designed as the conversational backend for Jun OS, an AI companion webapp. It is intended for:

Character-consistent multi-turn conversation
AI companion / interactive fiction applications
Research into character-faithful fine-tuning on small, high-quality datasets

Limitations

The model is specialized for a single character persona; it is not a general-purpose assistant.
Outputs may reflect fictional narrative tropes and should not be treated as factual information or advice.
Performance degrades on tasks far outside the training distribution (e.g. code generation, structured data extraction).
The model inherits any biases present in the Gemma 4 base weights.

Training Details

The v4 dataset is a deliberately small, heavily curated synthetic set built from the My Dystopian Robot Girlfriend dialogues and aligned to Jun OS's live system prompt and mood conditioning. The emphasis is on per-example quality and production alignment rather than raw volume.

Hyperparameters

Parameter	Value
Base model	`unsloth/gemma-4-E4B-it-qat-q4_0-unquantized`
Method	LoRA
LoRA rank	16
LoRA alpha	32
Fine-tuning framework	Unsloth
GGUF export pipeline	llama.cpp
Checkpoint	Step 100

Acknowledgments

Incontinent Cell for My Dystopian Robot Girlfriend, Jun's character
Google for the Gemma 4 model family
Unsloth for the efficient fine-tuning framework and the QAT base model

efficiencyx/Jun-v4-E4B-Step100-GGUF

Jun-v4-E4B-Step100