Downloads · 30 days
0
Gemma4-E4B-8B-Heretic-Ultra-QLora-Reverse-Engineering
dopaemon/Gemma4-E4B-8B-Heretic-Ultra-QLora-Reverse-Engineering
Gemma4-E4B-8B-Heretic-Ultra-QLora-Reverse-Engineering is a machine learning model from dopaemon. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx.
QLoRA adapter trained with mlx-lm on top of dopaemon/Gemma4-E4B-8B-Heretic-Ultra-MLX-4Bit, for the task of recovering the original function name from stripped, decompiled C-like code.
Access
Public
Updated Sep 23, 2026
Repo size
27.7 MB
Likes
1
Public
Repo makeup
Click a slice to open those files.
.safetensors27.7 MB · 100%
At a glance
- Library
- mlx
- Access
- Public
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026
- SHA
- 6d36a00d
Base models
Model card
From the Hugging Face model README
Gemma4 E4B 8B Heretic Ultra — RE QLoRA adapter
QLoRA adapter trained with mlx-lm on top of dopaemon/Gemma4-E4B-8B-Heretic-Ultra-MLX-4Bit, for the task of recovering the original function name from stripped, decompiled C-like code.
Training data
- ~2000 rows of decompiled function/name pairs at
-O0(unoptimized x86), then - ~2000 rows at
-O2(optimized x86, more inlining), resumed from the-O0adapter.
Training config
| Parameter | Value |
|---|---|
| fine_tune_type | lora |
| rank | 8 |
| scale | 20.0 |
| num_layers | 16 |
| iters | 1800 + 1800 (chained) |
| learning_rate | 1e-05 |
| max_seq_length | 1024 |
Usage
mlx_lm.generate \
--model dopaemon/Gemma4-E4B-8B-Heretic-Ultra-MLX-4Bit \
--adapter-path <this repo> \
--prompt "..."
or with mlx_lm.server --adapter-path <this repo>.
- Library
- mlx
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026