examples/README.md

OktoEngine Examples

Complete working examples demonstrating OktoEngine capabilities.

---

Table of Contents

1. Basic Training
2. LoRA Fine-tuning
3. Chatbot Training
4. Multi-format Export

---

Basic Training

Location: basic-training/

Minimal working example for training a simple model.

Files:

• scripts/train.okt - Training configuration
• dataset/train.jsonl - Sample training data
• dataset/val.jsonl - Sample validation data

Usage:

cd basic-training
okto validate
okto train

---

LoRA Fine-tuning

Location: lora-training/

Example of efficient LoRA fine-tuning for large models.

Files:

• scripts/train.okt - LoRA configuration
• dataset/train.jsonl - Training data

Usage:

cd lora-training
okto validate
okto train

---

Chatbot Training

Location: chatbot/

Complete example for training a conversational AI model.

Files:

• scripts/train.okt - Chatbot configuration
• dataset/train.jsonl - Conversation data
• dataset/val.jsonl - Validation conversations

Usage:

cd chatbot
okto validate
okto train
okto eval

---

Multi-format Export

Location: multi-export/

Example showing how to export models to multiple formats.

Files:

• scripts/train.okt - Configuration with multiple export formats

Usage:

cd multi-export
okto train
okto export --format okm,onnx,gguf

---

Running Examples

1. Navigate to example directory:

   cd examples/basic-training
   

2. Validate configuration:

   okto validate
   

3. Train the model:

   okto train
   

4. Check results:

   ls runs/
   ls export/
   

---

Customizing Examples

All examples can be customized:

1. Edit scripts/train.okt - Modify training parameters
2. Replace dataset/*.jsonl - Use your own data
3. Adjust MODEL.base - Use different base models
4. Modify EXPORT.format - Change export formats

---

Example Output

Training output:

🐙 OktoEngine v0.1
📄 Reading: "scripts/train.okt"

📊 Environment Check:
  ✔ Runtime: Python 3.14.0
  ✔ GPU: NVIDIA GeForce RTX 4070
  ✔ RAM: 63GB (40GB available)

🚀 Starting training pipeline...

Epoch 1/5: 100%|████████████| 500/500 [02:15<00:00, 3.70it/s]
  Loss: 2.345 → 1.892

✅ Training completed successfully!
📁 Output: runs/MyModel/

---

Need help? Check the Getting Started Guide or FAQ.