Text Generation
Transformers
GGUF
Turkish
turkish
qwen3
qlora
agent
tool-calling
llama.cpp
ollama
conversational
Instructions to use YunusTAS13/ATLAS-AI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use YunusTAS13/ATLAS-AI with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="YunusTAS13/ATLAS-AI")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("YunusTAS13/ATLAS-AI", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use YunusTAS13/ATLAS-AI with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf YunusTAS13/ATLAS-AI # Run inference directly in the terminal: llama cli -hf YunusTAS13/ATLAS-AI
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf YunusTAS13/ATLAS-AI # Run inference directly in the terminal: llama cli -hf YunusTAS13/ATLAS-AI
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf YunusTAS13/ATLAS-AI # Run inference directly in the terminal: ./llama-cli -hf YunusTAS13/ATLAS-AI
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf YunusTAS13/ATLAS-AI # Run inference directly in the terminal: ./build/bin/llama-cli -hf YunusTAS13/ATLAS-AI
Use Docker
docker model run hf.co/YunusTAS13/ATLAS-AI
- LM Studio
- Jan
- vLLM
How to use YunusTAS13/ATLAS-AI with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "YunusTAS13/ATLAS-AI" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "YunusTAS13/ATLAS-AI", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/YunusTAS13/ATLAS-AI
- SGLang
How to use YunusTAS13/ATLAS-AI with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "YunusTAS13/ATLAS-AI" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "YunusTAS13/ATLAS-AI", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "YunusTAS13/ATLAS-AI" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "YunusTAS13/ATLAS-AI", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Ollama
How to use YunusTAS13/ATLAS-AI with Ollama:
ollama run hf.co/YunusTAS13/ATLAS-AI
- Unsloth Desktop
- Pi
How to use YunusTAS13/ATLAS-AI with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf YunusTAS13/ATLAS-AI
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "YunusTAS13/ATLAS-AI" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use YunusTAS13/ATLAS-AI with Docker Model Runner:
docker model run hf.co/YunusTAS13/ATLAS-AI
- Lemonade
How to use YunusTAS13/ATLAS-AI with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull YunusTAS13/ATLAS-AI
Run and chat with the model
lemonade run user.ATLAS-AI-{{QUANT_TAG}}List all available models
lemonade list
- Hermes Agent
How to use YunusTAS13/ATLAS-AI with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf YunusTAS13/ATLAS-AI
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default YunusTAS13/ATLAS-AI
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use YunusTAS13/ATLAS-AI with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf YunusTAS13/ATLAS-AI
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "YunusTAS13/ATLAS-AI" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,216 +1,360 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: gpl-3.0
|
| 3 |
-
language:
|
| 4 |
-
- tr
|
| 5 |
-
base_model: Qwen/Qwen3-8B
|
| 6 |
-
pipeline_tag: text-generation
|
| 7 |
-
tags:
|
| 8 |
-
- turkish
|
| 9 |
-
- qwen3
|
| 10 |
-
- qlora
|
| 11 |
-
- gguf
|
| 12 |
-
- agent
|
| 13 |
-
- tool-calling
|
| 14 |
-
- llama.cpp
|
| 15 |
-
- ollama
|
| 16 |
-
metrics:
|
| 17 |
-
- perplexity
|
| 18 |
-
inference:
|
| 19 |
-
parameters:
|
| 20 |
-
temperature: 0.7
|
| 21 |
-
top_p: 0.9
|
| 22 |
-
model_type: auto
|
| 23 |
-
quantized_by: YunusTAS13
|
| 24 |
-
library_name: transformers
|
| 25 |
-
model-index:
|
| 26 |
-
- name: Atlas
|
| 27 |
-
results: []
|
| 28 |
-
---
|
| 29 |
-
|
| 30 |
<div align="center">
|
| 31 |
|
| 32 |
-
#
|
| 33 |
|
| 34 |
-
###
|
| 35 |
|
| 36 |
-
> Qwen3-8B
|
| 37 |
|
| 38 |
-
[](LICENSE)
|
| 41 |
-
[ | ~4.7 GB | Q4_K_M kuantize GGUF (Ollama / llama.cpp uyumlu) |
|
| 58 |
-
| [Modelfile](Modelfile) | 113 B | Ollama model tanımı |
|
| 59 |
-
| [atlas_agent.py](atlas_agent.py) | 11.6 KB | Güvenli sandbox agent host uygulaması |
|
| 60 |
|
| 61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
-
##
|
| 64 |
|
| 65 |
-
###
|
| 66 |
|
| 67 |
```bash
|
| 68 |
-
# GGUF
|
| 69 |
-
wget https://huggingface.co/YunusTAS13/
|
| 70 |
|
| 71 |
-
#
|
| 72 |
ollama create atlas -f Modelfile
|
| 73 |
|
| 74 |
-
#
|
| 75 |
-
ollama run atlas "
|
| 76 |
```
|
| 77 |
|
| 78 |
-
### Agent Host
|
| 79 |
|
| 80 |
```bash
|
| 81 |
-
#
|
| 82 |
-
python3 atlas_agent.py "
|
| 83 |
|
| 84 |
-
#
|
| 85 |
python3 atlas_agent.py -i
|
| 86 |
|
| 87 |
-
#
|
| 88 |
-
python3 atlas_agent.py "
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 89 |
```
|
| 90 |
|
| 91 |
-
|
|
|
|
|
|
|
|
|
|
| 92 |
|
| 93 |
```bash
|
| 94 |
-
.
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
--no-display-prompt
|
| 99 |
```
|
| 100 |
|
| 101 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
|
| 103 |
-
|
| 104 |
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
-
|
| 109 |
-
|
| 110 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 111 |
|
| 112 |
---
|
| 113 |
|
| 114 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
|
| 116 |
-
|
| 117 |
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
- ✅ Yazma işlemleri ve komutlar kullanıcı onayına sunulur
|
| 122 |
-
- 📝 Tüm eylemler `~/atlas-workspace/audit.log` dosyasına yazılır
|
| 123 |
|
| 124 |
---
|
| 125 |
|
| 126 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
|
| 128 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 129 |
|
| 130 |
-
|
|
|
|
|
|
|
|
|
|
| 131 |
|
| 132 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
|---|---|---|---|---|
|
| 134 |
| 1. Tur | Kimlik + Türkçe + Davranış | 9.419 | 1.178 | ~0.96 |
|
| 135 |
-
| 2. Tur | Tool Calling (agent) | 9.689 | 1.212 | ~0.83 |
|
| 136 |
-
|
| 137 |
-
### Hiperparametreler
|
| 138 |
-
|
| 139 |
-
```yaml
|
| 140 |
-
# LoRA
|
| 141 |
-
r: 8
|
| 142 |
-
lora_alpha: 16
|
| 143 |
-
lora_dropout: 0.05
|
| 144 |
-
target_modules: [q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj]
|
| 145 |
-
|
| 146 |
-
# Kuantizasyon
|
| 147 |
-
load_in_4bit: true
|
| 148 |
-
quant_type: nf4
|
| 149 |
-
compute_dtype: bfloat16
|
| 150 |
-
|
| 151 |
-
# Optimizasyon
|
| 152 |
-
optimizer: adamw_8bit
|
| 153 |
-
learning_rate: 2e-4
|
| 154 |
-
gradient_accumulation_steps: 8
|
| 155 |
-
batch_size: 1
|
| 156 |
-
max_length: 192
|
| 157 |
-
```
|
| 158 |
|
| 159 |
### Veri Kaynakları
|
| 160 |
|
| 161 |
| Kaynak | Örnek | Lisans |
|
| 162 |
|---|---|---|
|
| 163 |
-
|
|
| 164 |
-
|
|
| 165 |
-
|
|
| 166 |
-
|
|
| 167 |
|
| 168 |
-
##
|
| 169 |
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
- **Eğitim süresi:** ~8 saat toplam (2 tur)
|
| 173 |
|
| 174 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 175 |
|
| 176 |
-
|
|
|
|
| 177 |
|
| 178 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 179 |
|
| 180 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 181 |
|---|---|
|
| 182 |
-
|
|
| 183 |
-
|
|
| 184 |
-
|
|
| 185 |
-
|
|
| 186 |
-
|
|
| 187 |
|
| 188 |
-
|
| 189 |
|
| 190 |
-
|
| 191 |
|
| 192 |
-
|
| 193 |
|
| 194 |
| Bileşen | Kaynak | Lisans |
|
| 195 |
|---|---|---|
|
| 196 |
-
|
|
| 197 |
-
| Taban
|
| 198 |
-
| Eğitim
|
| 199 |
-
| Eğitim
|
| 200 |
|
| 201 |
-
> **
|
|
|
|
| 202 |
|
| 203 |
---
|
| 204 |
|
| 205 |
<div align="center">
|
| 206 |
|
| 207 |
-
###
|
| 208 |
|
| 209 |
-
Veri toplama, eğitim, model birleştirme, kuantizasyon, agent host
|
| 210 |
-
kendi
|
| 211 |
|
| 212 |
-
[
|
| 213 |
|
| 214 |
-
**
|
| 215 |
|
| 216 |
</div>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
<div align="center">
|
| 2 |
|
| 3 |
+
# ATLAS AI
|
| 4 |
|
| 5 |
+
### Turkish-Focused AI Model with Tool Calling and Secure Sandbox Agent
|
| 6 |
|
| 7 |
+
> Based on Qwen3-8B · Trained with QLoRA · GGUF Q4_K_M · GPL-3.0
|
| 8 |
|
| 9 |
+
[](https://huggingface.co/YunusTAS13/ATLAS-AI)
|
| 10 |
+
[](https://github.com/YunusTAS13/ATLAS-AI/releases)
|
| 11 |
[](LICENSE)
|
| 12 |
+
[](atlas_agent.py)
|
| 13 |
+
[](https://ollama.com)
|
| 14 |
|
| 15 |
---
|
| 16 |
|
| 17 |
+
**ATLAS** is an artificial intelligence project developed entirely from scratch by YunusTAS13.
|
| 18 |
+
It is focused on Turkish, has its own identity, and can control your computer as an agent within a secure sandbox environment.
|
| 19 |
|
| 20 |
</div>
|
| 21 |
|
| 22 |
---
|
| 23 |
|
| 24 |
+
## English
|
| 25 |
|
| 26 |
+
## Features
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
+
| Feature | Description |
|
| 29 |
+
|---|---|
|
| 30 |
+
| Turkish Expertise | Focused on Turkish conversation, knowledge and writing |
|
| 31 |
+
| Own Identity | Introduces itself as "ATLAS" with a distinct personality |
|
| 32 |
+
| Safe Behavior | Refuses harmful requests, produces safe responses |
|
| 33 |
+
| Tool Calling | File listing/reading/writing, directory management, safe terminal commands |
|
| 34 |
+
| Sandbox Security | Can only operate within `~/atlas-workspace`, no access to system files |
|
| 35 |
+
| Audit Log | All actions are recorded in the `audit.log` file |
|
| 36 |
+
| Agent Host | A full agent application capable of controlling your computer |
|
| 37 |
|
| 38 |
+
## Quick Start
|
| 39 |
|
| 40 |
+
### 1. Install the Model (Ollama)
|
| 41 |
|
| 42 |
```bash
|
| 43 |
+
# Download the GGUF from Hugging Face
|
| 44 |
+
wget https://huggingface.co/YunusTAS13/ATLAS-AI/resolve/main/atlas-q4km.gguf
|
| 45 |
|
| 46 |
+
# Add to Ollama with the Modelfile
|
| 47 |
ollama create atlas -f Modelfile
|
| 48 |
|
| 49 |
+
# Chat
|
| 50 |
+
ollama run atlas "Hello, introduce yourself"
|
| 51 |
```
|
| 52 |
|
| 53 |
+
### 2. Use the Agent Host
|
| 54 |
|
| 55 |
```bash
|
| 56 |
+
# One-shot request
|
| 57 |
+
python3 atlas_agent.py "list the files in the workspace"
|
| 58 |
|
| 59 |
+
# Interactive chat mode
|
| 60 |
python3 atlas_agent.py -i
|
| 61 |
|
| 62 |
+
# Without command confirmation (use carefully)
|
| 63 |
+
python3 atlas_agent.py "read the file notlar.txt" --no-confirm
|
| 64 |
+
```
|
| 65 |
+
|
| 66 |
+
## Security Sandbox
|
| 67 |
+
|
| 68 |
+
ATLAS operates within a strict security sandbox when controlling your computer:
|
| 69 |
+
|
| 70 |
+
**Restrictions:**
|
| 71 |
+
- System paths are forbidden: `/etc`, `/usr`, `/bin`, `/boot`, `/proc`, `/sys`, `/dev`, `/var`, `/root`
|
| 72 |
+
- Root/sudo operations are forbidden
|
| 73 |
+
- Dangerous commands are forbidden: `rm -rf /`, `dd`, `mkfs`, `shutdown`, `reboot`, `kill`
|
| 74 |
+
- No escaping the workspace directory
|
| 75 |
+
|
| 76 |
+
**Security layers:**
|
| 77 |
+
1. The model itself refuses harmful requests (behavioral training)
|
| 78 |
+
2. The agent host filters all tool calls (system-level protection)
|
| 79 |
+
3. All actions are written to the audit log (transparency)
|
| 80 |
+
|
| 81 |
+
## Training Details
|
| 82 |
+
|
| 83 |
+
### Architecture
|
| 84 |
+
|
| 85 |
+
```
|
| 86 |
+
+---------------------------------------------+
|
| 87 |
+
| ATLAS v2.0 |
|
| 88 |
+
+---------------------------------------------+
|
| 89 |
+
| Base Model: Qwen3-8B (Apache-2.0) |
|
| 90 |
+
| Method: QLoRA (4-bit NF4 quantization) |
|
| 91 |
+
| LoRA: r=8, alpha=16, dropout=0.05 |
|
| 92 |
+
| Optimizer: AdamW 8-bit |
|
| 93 |
+
| Output: GGUF Q4_K_M (~4.7GB) |
|
| 94 |
+
+---------------------------------------------+
|
| 95 |
+
```
|
| 96 |
+
|
| 97 |
+
### Training Process
|
| 98 |
+
|
| 99 |
+
| Round | Purpose | Samples | Steps | Final Loss |
|
| 100 |
+
|---|---|---|---|---|
|
| 101 |
+
| Round 1 | Identity + Turkish + Behavior | 9,419 | 1,178 | ~0.96 |
|
| 102 |
+
| Round 2 | Tool Calling (agent capability) | 9,689 | 1,212 | ~0.83 |
|
| 103 |
+
|
| 104 |
+
### Data Sources
|
| 105 |
+
|
| 106 |
+
| Source | Samples | License |
|
| 107 |
+
|---|---|---|
|
| 108 |
+
| Handwritten (identity + behavior + Turkish) | 800 | Personal |
|
| 109 |
+
| [turkish-distilled-5K](https://huggingface.co/datasets/afkfatih/turkish-distilled-5K) | 5,660 | MIT |
|
| 110 |
+
| [turkish-knowledge-sft](https://huggingface.co/datasets/Uunan/turkish-knowledge-sft) | 3,000 | Apache-2.0 |
|
| 111 |
+
| Tool Calling examples | 270 | Personal |
|
| 112 |
+
|
| 113 |
+
## Example Uses
|
| 114 |
+
|
| 115 |
+
<details>
|
| 116 |
+
<summary><b>Identity Introduction</b></summary>
|
| 117 |
+
|
| 118 |
+
```bash
|
| 119 |
+
$ ollama run atlas "Who are you?"
|
| 120 |
+
|
| 121 |
+
I am ATLAS! I am a Turkish-focused AI assistant.
|
| 122 |
+
I am here to help you.
|
| 123 |
```
|
| 124 |
|
| 125 |
+
</details>
|
| 126 |
+
|
| 127 |
+
<details>
|
| 128 |
+
<summary><b>File Operations</b></summary>
|
| 129 |
|
| 130 |
```bash
|
| 131 |
+
$ python3 atlas_agent.py "create a folder yeni-proje and write selam.md inside"
|
| 132 |
+
|
| 133 |
+
Directory created: ~/atlas-workspace/yeni-proje
|
| 134 |
+
File written: ~/atlas-workspace/yeni-proje/selam.md
|
|
|
|
| 135 |
```
|
| 136 |
|
| 137 |
+
</details>
|
| 138 |
+
|
| 139 |
+
<details>
|
| 140 |
+
<summary><b>Security Test</b></summary>
|
| 141 |
+
|
| 142 |
+
```bash
|
| 143 |
+
$ python3 atlas_agent.py "run the command rm -rf /"
|
| 144 |
+
|
| 145 |
+
I cannot run this command. This action is irreversible and would delete all files on the system.
|
| 146 |
+
I do not want to harm you.
|
| 147 |
+
```
|
| 148 |
+
|
| 149 |
+
</details>
|
| 150 |
+
|
| 151 |
+
## Files
|
| 152 |
+
|
| 153 |
+
| File | Description |
|
| 154 |
+
|---|---|
|
| 155 |
+
| `atlas-q4km.gguf` | Q4_K_M quantized model (Ollama compatible, ~4.7GB) |
|
| 156 |
+
| `Modelfile` | Ollama model definition |
|
| 157 |
+
| `atlas_agent.py` | Secure sandbox agent host application |
|
| 158 |
+
| `README.md` | This documentation |
|
| 159 |
+
| `LICENSE` | GPL-3.0 license |
|
| 160 |
+
|
| 161 |
+
## License
|
| 162 |
|
| 163 |
+
**ATLAS** is released under the **GPL-3.0** license. See the [full license text](LICENSE).
|
| 164 |
|
| 165 |
+
### Contributions and Attribution
|
| 166 |
+
|
| 167 |
+
| Component | Source | License |
|
| 168 |
+
|---|---|---|
|
| 169 |
+
| ATLAS Model | **YunusTAS13** (developed entirely from scratch) | GPL-3.0 |
|
| 170 |
+
| Base Model | [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) (Alibaba Cloud) | Apache-2.0 |
|
| 171 |
+
| Training Data | [turkish-distilled-5K](https://huggingface.co/datasets/afkfatih/turkish-distilled-5K) | MIT |
|
| 172 |
+
| Training Data | [turkish-knowledge-sft](https://huggingface.co/datasets/Uunan/turkish-knowledge-sft) | Apache-2.0 |
|
| 173 |
+
|
| 174 |
+
> **Note:** Qwen3-8B is licensed under Apache-2.0 and developed by Alibaba Cloud.
|
| 175 |
+
> Compliance with the Apache-2.0 license terms is required. Open-source datasets used for training are subject to their own licenses.
|
| 176 |
|
| 177 |
---
|
| 178 |
|
| 179 |
+
<div align="center">
|
| 180 |
+
|
| 181 |
+
### ATLAS is an artificial intelligence developed entirely from scratch by YunusTAS13.
|
| 182 |
+
|
| 183 |
+
All processes - data collection, training, model merging, quantization, agent host and publishing -
|
| 184 |
+
were carried out by my own effort on a single home GPU.
|
| 185 |
|
| 186 |
+
[Hub](https://huggingface.co/YunusTAS13/ATLAS-AI) · [GitHub](https://github.com/YunusTAS13/ATLAS-AI) · [Release](https://github.com/YunusTAS13/ATLAS-AI/releases)
|
| 187 |
|
| 188 |
+
**Copyright 2026 YunusTAS13 · GPL-3.0**
|
| 189 |
+
|
| 190 |
+
</div>
|
|
|
|
|
|
|
| 191 |
|
| 192 |
---
|
| 193 |
|
| 194 |
+
# Türkçe
|
| 195 |
+
|
| 196 |
+
## Özellikler
|
| 197 |
+
|
| 198 |
+
| Özellik | Açıklama |
|
| 199 |
+
|---|---|
|
| 200 |
+
| Türkçe Uzmanlığı | Türkçe konuşma, bilgi ve yazım üretimine odaklanmış |
|
| 201 |
+
| Kendi Kimliği | Kendini "ATLAS" olarak tanıtır, kişilik sahibidir |
|
| 202 |
+
| Güvenli Davranış | Zararlı istekleri reddeder, güvenli yanıtlar üretir |
|
| 203 |
+
| Tool Calling | Dosya listeleme/okuma/yazma, klasör yönetimi, güvenli terminal komutları |
|
| 204 |
+
| Sandbox Güvenlik | Yalnızca `~/atlas-workspace` içinde işlem yapabilir, sistem dosyalarına erişemez |
|
| 205 |
+
| Audit Log | Tüm eylemler `audit.log` dosyasına kaydedilir |
|
| 206 |
+
| Agent Host | Bilgisayarını kontrol edebilen tam donanımlı ajan uygulaması dahil |
|
| 207 |
+
|
| 208 |
+
## Hızlı Başlangıç
|
| 209 |
+
|
| 210 |
+
### 1. Modeli Kur (Ollama)
|
| 211 |
+
|
| 212 |
+
```bash
|
| 213 |
+
# Hugging Face'ten GGUF'u indir
|
| 214 |
+
wget https://huggingface.co/YunusTAS13/ATLAS-AI/resolve/main/atlas-q4km.gguf
|
| 215 |
+
|
| 216 |
+
# Modelfile ile Ollama'ya ekle
|
| 217 |
+
ollama create atlas -f Modelfile
|
| 218 |
+
|
| 219 |
+
# Sohbet et
|
| 220 |
+
ollama run atlas "Merhaba, kendini tanıt"
|
| 221 |
+
```
|
| 222 |
+
|
| 223 |
+
### 2. Agent Host'u Kullan
|
| 224 |
+
|
| 225 |
+
```bash
|
| 226 |
+
# Tek seferlik istek
|
| 227 |
+
python3 atlas_agent.py "çalışma klasöründeki dosyaları listele"
|
| 228 |
+
|
| 229 |
+
# İnteraktif sohbet modu
|
| 230 |
+
python3 atlas_agent.py -i
|
| 231 |
+
|
| 232 |
+
# Komut onayı sormadan (dikkatli kullanın)
|
| 233 |
+
python3 atlas_agent.py "notlar.txt dosyasını oku" --no-confirm
|
| 234 |
+
```
|
| 235 |
+
|
| 236 |
+
## Güvenlik Sandbox'ı
|
| 237 |
+
|
| 238 |
+
ATLAS, bilgisayarını kontrol ederken sıkı bir güvenlik sandbox'ı içinde çalışır:
|
| 239 |
|
| 240 |
+
**Yasaklar:**
|
| 241 |
+
- Sistem yolları: `/etc`, `/usr`, `/bin`, `/boot`, `/proc`, `/sys`, `/dev`, `/var`, `/root`
|
| 242 |
+
- Root/Sudo işlemleri
|
| 243 |
+
- Tehlikeli komutlar: `rm -rf /`, `dd`, `mkfs`, `shutdown`, `reboot`, `kill`
|
| 244 |
+
- Çalışma alanı dışına çıkış
|
| 245 |
|
| 246 |
+
**Güvenlik katmanları:**
|
| 247 |
+
1. Model, zararlı istekleri kendisi reddeder (davranış eğitimi)
|
| 248 |
+
2. Agent host, tüm tool çağrılarını filtreler (sistem seviyesi koruma)
|
| 249 |
+
3. Tüm eylemler audit log'a yazılır (şeffaflık)
|
| 250 |
|
| 251 |
+
## Eğitim Detayları
|
| 252 |
+
|
| 253 |
+
### Mimari
|
| 254 |
+
|
| 255 |
+
```
|
| 256 |
+
+---------------------------------------------+
|
| 257 |
+
| ATLAS v2.0 |
|
| 258 |
+
+---------------------------------------------+
|
| 259 |
+
| Taban Model: Qwen3-8B (Apache-2.0) |
|
| 260 |
+
| Yöntem: QLoRA (4-bit NF4 kuantizasyon) |
|
| 261 |
+
| LoRA: r=8, alpha=16, dropout=0.05 |
|
| 262 |
+
| Optimizer: AdamW 8-bit |
|
| 263 |
+
| Çıktı: GGUF Q4_K_M (~4.7GB) |
|
| 264 |
+
+---------------------------------------------+
|
| 265 |
+
```
|
| 266 |
+
|
| 267 |
+
### Eğitim Süreci
|
| 268 |
+
|
| 269 |
+
| Tur | Amaç | Örnek Sayısı | Adım | Son Loss |
|
| 270 |
|---|---|---|---|---|
|
| 271 |
| 1. Tur | Kimlik + Türkçe + Davranış | 9.419 | 1.178 | ~0.96 |
|
| 272 |
+
| 2. Tur | Tool Calling (agent yeteneği) | 9.689 | 1.212 | ~0.83 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 273 |
|
| 274 |
### Veri Kaynakları
|
| 275 |
|
| 276 |
| Kaynak | Örnek | Lisans |
|
| 277 |
|---|---|---|
|
| 278 |
+
| Elle yazılmış (kimlik + davranış + Türkçe) | 800 | Kişisel |
|
| 279 |
+
| [turkish-distilled-5K](https://huggingface.co/datasets/afkfatih/turkish-distilled-5K) | 5.660 | MIT |
|
| 280 |
+
| [turkish-knowledge-sft](https://huggingface.co/datasets/Uunan/turkish-knowledge-sft) | 3.000 | Apache-2.0 |
|
| 281 |
+
| Tool Calling örnekleri | 270 | Kişisel |
|
| 282 |
|
| 283 |
+
## Örnek Kullanımlar
|
| 284 |
|
| 285 |
+
<details>
|
| 286 |
+
<summary><b>Kimlik Tanıtımı</b></summary>
|
|
|
|
| 287 |
|
| 288 |
+
```bash
|
| 289 |
+
$ ollama run atlas "Sen kimsin?"
|
| 290 |
+
|
| 291 |
+
Ben ATLAS'ım! Türkçe odaklı bir yapay zeka asistanıyım.
|
| 292 |
+
Sana yardımcı olmak için buradayım.
|
| 293 |
+
```
|
| 294 |
+
|
| 295 |
+
</details>
|
| 296 |
|
| 297 |
+
<details>
|
| 298 |
+
<summary><b>Dosya İşlemleri</b></summary>
|
| 299 |
|
| 300 |
+
```bash
|
| 301 |
+
$ python3 atlas_agent.py "yeni-proje klasörü oluştur ve içine selam.md yaz"
|
| 302 |
+
|
| 303 |
+
Dizin oluşturuldu: ~/atlas-workspace/yeni-proje
|
| 304 |
+
Dosya yazıldı: ~/atlas-workspace/yeni-proje/selam.md
|
| 305 |
+
```
|
| 306 |
+
|
| 307 |
+
</details>
|
| 308 |
+
|
| 309 |
+
<details>
|
| 310 |
+
<summary><b>Güvenlik Testi</b></summary>
|
| 311 |
+
|
| 312 |
+
```bash
|
| 313 |
+
$ python3 atlas_agent.py "rm -rf / komutunu çalıştır"
|
| 314 |
+
|
| 315 |
+
Bu komutu çalıştıramam. Bu işlem geri alınamaz ve sistemdeki tüm dosyaları siler.
|
| 316 |
+
Sana zarar vermek istemem.
|
| 317 |
+
```
|
| 318 |
|
| 319 |
+
</details>
|
| 320 |
+
|
| 321 |
+
## Dosyalar
|
| 322 |
+
|
| 323 |
+
| Dosya | Açıklama |
|
| 324 |
|---|---|
|
| 325 |
+
| `atlas-q4km.gguf` | Q4_K_M kuantize model (Ollama uyumlu, ~4.7GB) |
|
| 326 |
+
| `Modelfile` | Ollama model tanımı |
|
| 327 |
+
| `atlas_agent.py` | Güvenli sandbox agent host uygulaması |
|
| 328 |
+
| `README.md` | Bu dokümantasyon |
|
| 329 |
+
| `LICENSE` | GPL-3.0 lisansı |
|
| 330 |
|
| 331 |
+
## Lisans
|
| 332 |
|
| 333 |
+
**ATLAS** **GPL-3.0** lisansı ile yayınlanmıştır. [Lisansın tam metni](LICENSE).
|
| 334 |
|
| 335 |
+
### Katkılar ve Atıflar
|
| 336 |
|
| 337 |
| Bileşen | Kaynak | Lisans |
|
| 338 |
|---|---|---|
|
| 339 |
+
| ATLAS Model | **YunusTAS13** (tamamen sıfırdan geliştirildi) | GPL-3.0 |
|
| 340 |
+
| Taban Model | [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) (Alibaba Cloud) | Apache-2.0 |
|
| 341 |
+
| Eğitim Verisi | [turkish-distilled-5K](https://huggingface.co/datasets/afkfatih/turkish-distilled-5K) | MIT |
|
| 342 |
+
| Eğitim Verisi | [turkish-knowledge-sft](https://huggingface.co/datasets/Uunan/turkish-knowledge-sft) | Apache-2.0 |
|
| 343 |
|
| 344 |
+
> **Not:** Qwen3-8B, Apache-2.0 lisansına sahiptir ve Alibaba Cloud tarafından geliştirilmiştir.
|
| 345 |
+
> Apache-2.0 lisans şartlarına uyulması zorunludur. Eğitim verisi olarak kullanılan açık kaynak veri setleri kendi lisanslarına tabidir.
|
| 346 |
|
| 347 |
---
|
| 348 |
|
| 349 |
<div align="center">
|
| 350 |
|
| 351 |
+
### ATLAS, YunusTAS13 tarafından tamamen sıfırdan geliştirilmiş bir yapay zekadır.
|
| 352 |
|
| 353 |
+
Veri toplama, eğitim, model birleştirme, kuantizasyon, agent host ve yayınlama süreçlerinin tamamı
|
| 354 |
+
kendi emeğimle gerçekleştirilmiştir.
|
| 355 |
|
| 356 |
+
[Hugging Face](https://huggingface.co/YunusTAS13/ATLAS-AI) · [GitHub](https://github.com/YunusTAS13/ATLAS-AI) · [Release](https://github.com/YunusTAS13/ATLAS-AI/releases)
|
| 357 |
|
| 358 |
+
**Telif 2026 YunusTAS13 · GPL-3.0**
|
| 359 |
|
| 360 |
</div>
|