Overview

Get started with SLM Builder in under 5 minutes.

Table of contents

1. Installation
2. Your First Model
   1. 1. Import and Initialize
   2. 2. Load Data
   3. 3. Train the Model
   4. 4. Evaluate
   5. 5. Use Your Model
   6. 6. Export
3. Complete Example
4. What’s Next?
   1. Learn More About Features
   2. Try More Examples
   3. Configure Your Project
5. Common Issues
   1. Import Error
   2. CUDA Out of Memory
   3. Model Download Fails
6. Need Help?

---

Installation

pip install slm-builder

For specific features:

# Database support
pip install slm-builder[db]

# All features
pip install slm-builder[all]

---

Your First Model

1. Import and Initialize

from slm_builder import SLMBuilder

builder = SLMBuilder(
    project_name="my-first-model",
    base_model="google/flan-t5-small"
)

2. Load Data

From a CSV file:

builder.load_data_from_csv("training_data.csv")

Your CSV should have these columns:

• question or input
• answer or output

3. Train the Model

builder.train(
    epochs=3,
    learning_rate=2e-5,
    method="lora"  # Efficient fine-tuning
)

4. Evaluate

metrics = builder.evaluate()
print(f"Accuracy: {metrics['accuracy']:.2%}")
print(f"F1 Score: {metrics['f1']:.3f}")

5. Use Your Model

response = builder.generate("What is machine learning?")
print(response)

6. Export

# Export to HuggingFace format
builder.export_model("./output", format="huggingface")

# Or export to ONNX
builder.export_model("./output", format="onnx")

---

Complete Example

from slm_builder import SLMBuilder

# Initialize
builder = SLMBuilder(
    project_name="qa-bot",
    base_model="google/flan-t5-small"
)

# Load data
builder.load_data_from_csv("qa_data.csv")

# Train
builder.train(
    epochs=3,
    learning_rate=2e-5,
    method="lora",
    batch_size=8
)

# Evaluate
metrics = builder.evaluate()
print(f"✅ Model trained! Accuracy: {metrics['accuracy']:.2%}")

# Test
response = builder.generate("How do I install Python?")
print(f"Bot: {response}")

# Export
builder.export_model("./qa-bot-model", format="huggingface")

---

What’s Next?

Learn More About Features

• Model Loading - Load from different sources
• Data Sources - Use databases, APIs, and more
• Training Methods - Configure training options
• Evaluation Metrics - Understand metrics

Try More Examples

• Getting Started Examples - More basic examples
• Database Integration - Use databases
• API Integration - Fetch from APIs

Configure Your Project

• Configuration Reference - All options
• Best Practices - Production tips

---

Common Issues

Import Error

ModuleNotFoundError: No module named 'slm_builder'

Solution: Make sure you’ve installed the package:

pip install slm-builder

CUDA Out of Memory

RuntimeError: CUDA out of memory

Solution: Reduce batch size or use CPU:

builder.train(batch_size=4, device="cpu")

Model Download Fails

ConnectionError: Failed to download model

Solution: Check your internet connection or use a local model:

builder = SLMBuilder(base_model="/path/to/local/model")

---

Need Help?

• GitHub Issues
• Discussions
• Full Documentation