basilmh25 commited on
Commit
7421741
·
verified ·
1 Parent(s): 109df69

update readme

Browse files
Files changed (1) hide show
  1. README.md +159 -14
README.md CHANGED
@@ -1,23 +1,168 @@
1
  ---
 
 
 
 
 
 
2
  tags:
3
- - gguf
4
- - llama.cpp
5
- - unsloth
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
 
7
  ---
8
 
9
- # qwen_Cybersecucity : GGUF
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
 
11
- This model was finetuned and converted to GGUF format using [Unsloth](https://github.com/unslothai/unsloth).
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12
 
13
- **Example usage**:
14
- - For text only LLMs: `llama-cli -hf basilmh25/qwen_Cybersecucity --jinja`
15
- - For multimodal models: `llama-mtmd-cli -hf basilmh25/qwen_Cybersecucity --jinja`
16
 
17
- ## Available Model files:
18
- - `qwen3-4b-instruct-2507.Q4_K_M.gguf`
19
 
20
- ## Ollama
21
- An Ollama Modelfile is included for easy deployment.
22
- This was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth)
23
- [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
 
1
  ---
2
+ license: apache-2.0
3
+ base_model: unsloth/Qwen3-4B-Instruct-2507
4
+ library_name: transformers
5
+ pipeline_tag: text-generation
6
+ language:
7
+ - en
8
  tags:
9
+ - qwen3
10
+ - lora
11
+ - unsloth
12
+ - cybersecurity
13
+ - instruction-tuning
14
+ - security
15
+ - llm
16
+ datasets:
17
+ - AlicanKiraz0/Cybersecurity-Dataset-Fenrir-v2.1
18
+ ---
19
+
20
+ # Qwen3-4B-Instruct-2507 Cybersecurity LoRA
21
+
22
+ ## Model Description
23
+
24
+ This repository contains a **LoRA fine-tuned** version of **Qwen3-4B-Instruct-2507**, specialized for cybersecurity-related instruction following and question answering.
25
+
26
+ The model was trained using **LoRA (Low-Rank Adaptation)** with the **Unsloth** framework on the **Cybersecurity Dataset Fenrir v2.1**.
27
+
28
+ The goal of this fine-tuning is to improve the model's ability to answer cybersecurity questions, explain security concepts, assist with defensive security tasks, and provide technically accurate responses while maintaining the general conversational capabilities of the base model.
29
 
30
  ---
31
 
32
+ ## Base Model
33
+
34
+ - **Model:** unsloth/Qwen3-4B-Instruct-2507
35
+ - **Architecture:** Qwen3
36
+ - **Parameters:** 4 Billion
37
+ - **License:** Apache-2.0
38
+
39
+ The model builds upon the official Unsloth implementation of Qwen3-4B-Instruct-2507, which is optimized for efficient fine-tuning and inference. :contentReference[oaicite:0]{index=0}
40
+
41
+ ---
42
+
43
+ ## Training Dataset
44
+
45
+ **Dataset**
46
+
47
+ - AlicanKiraz0/Cybersecurity-Dataset-Fenrir-v2.1
48
+
49
+ The dataset contains cybersecurity-focused instruction-response pairs covering topics such as:
50
+
51
+ - Network Security
52
+ - Malware Analysis
53
+ - Web Security
54
+ - Cryptography
55
+ - Incident Response
56
+ - Digital Forensics
57
+ - Penetration Testing
58
+ - Vulnerability Assessment
59
+ - Threat Intelligence
60
+ - Secure Programming
61
+ - General Cybersecurity Knowledge
62
+
63
+ ---
64
+
65
+ ## Fine-Tuning Method
66
+
67
+ - Method: Supervised Fine-Tuning (SFT)
68
+ - Adapter: LoRA
69
+ - Framework: Unsloth
70
+ - Base Model: Qwen3-4B-Instruct-2507
71
+
72
+ Only LoRA adapter weights were trained while the original model weights remained frozen.
73
+
74
+ ---
75
+
76
+ ## Intended Use
77
+
78
+ This model is intended for:
79
+
80
+ - Cybersecurity education
81
+ - Security awareness
82
+ - Technical question answering
83
+ - Security documentation assistance
84
+ - Capture The Flag (CTF) learning
85
+ - Blue Team support
86
+ - SOC analyst assistance
87
+ - Research
88
+
89
+ ---
90
+
91
+ ## Limitations
92
+
93
+ This model:
94
 
95
+ - is not a replacement for professional security expertise;
96
+ - may generate inaccurate or outdated information;
97
+ - should not be relied upon as the sole source for security-critical decisions;
98
+ - should always be validated against official documentation and current security guidance.
99
+
100
+ ---
101
+
102
+ ## Example
103
+
104
+ ### User
105
+
106
+ ```
107
+ Explain the difference between a vulnerability, an exploit, and a payload.
108
+ ```
109
+
110
+ ### Assistant
111
+
112
+ ```
113
+ A vulnerability is a weakness in software or a system that can be abused.
114
+
115
+ An exploit is the technique or code used to take advantage of that vulnerability.
116
+
117
+ A payload is the code executed after successful exploitation, such as opening a reverse shell, installing malware, or escalating privileges.
118
+ ```
119
+
120
+ ---
121
+
122
+ ## Loading the LoRA Adapter
123
+
124
+ ```python
125
+ from transformers import AutoTokenizer, AutoModelForCausalLM
126
+ from peft import PeftModel
127
+
128
+ base_model = AutoModelForCausalLM.from_pretrained(
129
+ "unsloth/Qwen3-4B-Instruct-2507"
130
+ )
131
+
132
+ tokenizer = AutoTokenizer.from_pretrained(
133
+ "unsloth/Qwen3-4B-Instruct-2507"
134
+ )
135
+
136
+ model = PeftModel.from_pretrained(
137
+ base_model,
138
+ "YOUR_USERNAME/YOUR_MODEL_NAME"
139
+ )
140
+ ```
141
+
142
+ ---
143
+
144
+ ## Citation
145
+
146
+ If you use this model in your work, please cite both:
147
+
148
+ - Qwen3
149
+ - Unsloth
150
+ - Cybersecurity Dataset Fenrir v2.1
151
+
152
+ ---
153
+
154
+ ## Acknowledgements
155
+
156
+ Thanks to:
157
+
158
+ - Alibaba Qwen Team
159
+ - Unsloth
160
+ - AlicanKiraz for the Cybersecurity Dataset Fenrir v2.1
161
+
162
+ ---
163
 
164
+ ## Disclaimer
 
 
165
 
166
+ This model is intended for educational, research, and defensive cybersecurity purposes only.
 
167
 
168
+ Users are responsible for complying with all applicable laws and regulations. The authors assume no responsibility for misuse of this model.