Getting Started

Prerequisites

Python 3.10+
CUDA GPU (8GB+ VRAM)
PyTorch 2.0+

Installation

pip install backpropagate[standard]    # Recommended: unsloth + ui

Other install options:

Extra	What you get
`backpropagate`	Core API only — minimal footprint
`[unsloth]`	2x faster training, 50% less VRAM
`[ui]`	Gradio web interface
`[validation]`	Pydantic config validation
`[export]`	GGUF export for Ollama
`[monitoring]`	WandB + system monitoring
`[observability]`	OpenTelemetry tracing
`[logging]`	Structured logging via structlog
`[security]`	JWT auth + token generation
`[standard]`	unsloth + ui (recommended)
`[production]`	unsloth + ui + validation + logging + security
`[full]`	Everything

Train in 3 lines

from backpropagate import Trainer

trainer = Trainer("unsloth/Qwen2.5-7B-Instruct-bnb-4bit")
trainer.train("my_data.jsonl", steps=100)
trainer.export("gguf", quantization="q4_k_m")  # Ready for Ollama

CLI usage

backprop train --data my_data.jsonl --model unsloth/Qwen2.5-7B-Instruct-bnb-4bit --steps 100
backprop export ./output/lora --format gguf --quantization q4_k_m --ollama --ollama-name my-model

Verify your setup

backprop info

This prints your Python version, PyTorch version, CUDA status, GPU name and VRAM, which optional features are installed, and current configuration defaults. Run it before your first training to confirm everything is working.

Web UI

backprop ui --port 7862