Upload README.md

README.md

@@ -20,20 +20,28 @@ datasets:
 ---
 
 <div align="center">
-  <img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/logo_qwen.png" width="200"/>
-  <br/>
-  <h1>Qwen3-4B-Cybersecurity</h1>
-  <p><strong>Qwen3-4B fine-tuned on 1.28M cybersecurity samples</strong></p>
-
-  [![Model](https://img.shields.io/badge/🤗_Model-Qwen3--4B--Cybersecurity-blue)](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity)
-  [![Heretic Version](https://img.shields.io/badge/🔪_Abliterated-Heretic_Version-red)](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic)
-  [![GGUF](https://img.shields.io/badge/📦_GGUF-Q8_%2F_Q4-green)](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-GGUF)
-  [![Base Model](https://img.shields.io/badge/Base-Qwen3--4B--Instruct--2507-orange)](https://huggingface.co/unsloth/Qwen3-4B-Instruct-2507)
+
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="https://huggingface.co/datasets/Qwen/documentation-images/resolve/main/Qwen-logo_White.png">
+  <source media="(prefers-color-scheme: light)" srcset="https://huggingface.co/datasets/Qwen/documentation-images/resolve/main/Qwen-logo_Black.png">
+  <img alt="Qwen Logo" src="https://huggingface.co/datasets/Qwen/documentation-images/resolve/main/Qwen-logo_Black.png" width="180"/>
+</picture>
+
+# Qwen3-4B-Cybersecurity
+
+**Qwen3-4B fine-tuned on 1.28M cybersecurity samples**
+
+[![Base Model](https://img.shields.io/badge/Base-Qwen3--4B--Instruct--2507-orange?style=flat-square)](https://huggingface.co/unsloth/Qwen3-4B-Instruct-2507)
+[![Heretic Version](https://img.shields.io/badge/🔪_Abliterated-Heretic_Version-red?style=flat-square)](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit)
+[![GGUF](https://img.shields.io/badge/📦_GGUF-Q8_/_Q4-brightgreen?style=flat-square)](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-GGUF)
+[![Dataset](https://img.shields.io/badge/📊_Dataset-cyber__heretic-blue?style=flat-square)](https://huggingface.co/datasets/DexopT/cyber_heretic)
+[![License](https://img.shields.io/badge/License-Apache_2.0-yellow?style=flat-square)](https://apache.org/licenses/LICENSE-2.0)
+
 </div>
 
 ---
 
-> 🔪 **Abliterated (refusal-removed) version available:** [DexopT/Qwen3-4B-Cybersecurity-Heretic](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic)
+> 🔪 **Abliterated (refusal-removed) version:** [DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit)
 
 ## Model Description
 
@@ -55,7 +63,7 @@ Trained with [Unsloth](https://github.com/unslothai/unsloth) SFT on a curated da
 | Model | Description | Link |
 |-------|-------------|------|
 | **Qwen3-4B-Cybersecurity** | Base fine-tuned model (this repo) | 📍 You are here |
-| Qwen3-4B-Cybersecurity-Heretic | Abliterated — refusal directions removed | [→](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic) |
+| Qwen3-4B-Cybersecurity-Heretic-16bit | Abliterated — refusal directions removed | [→](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit) |
 | Qwen3-4B-Cybersecurity-GGUF | Q8_0 + Q4_K_M quantized for llama.cpp | [→](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-GGUF) |
 
 ---
@@ -123,10 +131,16 @@ print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_t
 
 ```bash
 pip install mlx-lm
-mlx_lm.chat --model DexopT/Qwen3-4B-Cybersecurity
+
+mlx_lm.convert \
+  --hf-path DexopT/Qwen3-4B-Cybersecurity \
+  --mlx-path ~/models/qwen3-cyber-mlx \
+  --quantize --q-bits 8
+
+mlx_lm.chat --model ~/models/qwen3-cyber-mlx
 ```
 
-### LM Studio / Ollama
+### LM Studio / llama.cpp / Ollama
 
 Use the GGUF version: [DexopT/Qwen3-4B-Cybersecurity-GGUF](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-GGUF)
 
@@ -134,9 +148,9 @@ Use the GGUF version: [DexopT/Qwen3-4B-Cybersecurity-GGUF](https://huggingface.c
 
 ## Refusal Behavior
 
-This model retains some of its original refusal behavior from the base Qwen3 model. For a version with refusal directions removed, see the Heretic version:
+This model retains some of the original refusal behavior from the base Qwen3 model. For a version with refusal directions removed via Heretic abliteration, see:
 
-> 🔪 [DexopT/Qwen3-4B-Cybersecurity-Heretic](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic)
+> 🔪 [DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit)
 
 ---
 
@@ -148,8 +162,10 @@ This model is intended for **educational and research purposes only**. Use respo
 
 ## Links
 
-- 🤗 [Heretic Version](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic)
-- 📦 [GGUF Versions](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-GGUF)
-- 📊 [Training Dataset](https://huggingface.co/datasets/DexopT/cyber_heretic)
-- 🔧 [Base Model](https://huggingface.co/unsloth/Qwen3-4B-Instruct-2507)
-- 🏠 [Qwen3 Collection](https://huggingface.co/collections/Qwen/qwen3)
+| | |
+|---|---|
+| 🔪 Heretic Version | [DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit) |
+| 📦 GGUF (Q8 + Q4) | [DexopT/Qwen3-4B-Cybersecurity-GGUF](https://huggingface.co/DexopT/Qwen3-4B-Cybersecurity-GGUF) |
+| 📊 Training Dataset | [DexopT/cyber_heretic](https://huggingface.co/datasets/DexopT/cyber_heretic) |
+| 🔧 Base Model | [unsloth/Qwen3-4B-Instruct-2507](https://huggingface.co/unsloth/Qwen3-4B-Instruct-2507) |
+| 🏠 Qwen3 Collection | [Qwen3 on HuggingFace](https://huggingface.co/collections/Qwen/qwen3) |