langtune 0.0.2

A package for finetuning text models.

Langtune: Large Language Models (LLMs) with Efficient LoRA Fine-Tuning for Text

---

Langtune provides modular components for text models and LoRA-based fine-tuning.
Adapt and fine-tune language models for a range of NLP tasks.

---

Quick Links

• Documentation
• Tutorials
• Changelog
• Contributing Guide
• Roadmap

---

Table of Contents

• Features
• Showcase
• Getting Started
• Supported Python Versions
• Why langtune?
• Architecture Overview
• Core Modules
• Performance & Efficiency
• Advanced Configuration
• Documentation & Resources
• Testing & Quality
• Examples & Use Cases
• Extending the Framework
• Contributing
• FAQ
• Citation
• Acknowledgements
• License

---

Features

• LoRA adapters for parameter-efficient fine-tuning of LLMs
• Modular transformer backbone
• Model zoo for open-source language models
• Configurable and extensible codebase
• Checkpointing and resume support
• Mixed precision and distributed training
• Built-in metrics and visualization tools
• CLI for fine-tuning and evaluation
• Extensible callbacks (early stopping, logging, etc.)

---

Showcase

Langtune is a framework for building and fine-tuning large language models with LoRA support. It is suitable for tasks such as text classification, summarization, question answering, and other NLP applications.

---

Getting Started

Install with pip:

pip install langtune

Minimal example:

import torch
from langtune.models.llm import LanguageModel
from langtune.utils.config import default_config

input_ids = torch.randint(0, 1000, (2, 128))
model = LanguageModel(
    vocab_size=default_config['vocab_size'],
    embed_dim=default_config['embed_dim'],
    num_layers=default_config['num_layers'],
    num_heads=default_config['num_heads'],
    mlp_ratio=default_config['mlp_ratio'],
    lora_config=default_config['lora'],
)

with torch.no_grad():
    out = model(input_ids)
    print('Output shape:', out.shape)

For more details, see the Documentation and src/langtune/cli/finetune.py.

---

Supported Python Versions

• Python 3.8+

---

Why langtune?

• Parameter-efficient fine-tuning with LoRA adapters
• Modular transformer backbone for flexible model design
• Unified interface for open-source language models
• Designed for both research and production
• Efficient memory usage for large models

---

Architecture Overview

Langtune uses a modular transformer backbone with LoRA adapters in attention and MLP layers. This allows adaptation of pre-trained models with fewer trainable parameters.

Model Data Flow

---
config:
  layout: dagre
---
flowchart TD
 subgraph LoRA_Adapters["LoRA Adapters in Attention and MLP"]
        LA1(["LoRA Adapter 1"])
        LA2(["LoRA Adapter 2"])
        LA3(["LoRA Adapter N"])
  end
    A(["Input Tokens"]) --> B(["Embedding Layer"])
    B --> C(["Positional Encoding"])
    C --> D1(["Encoder Layer 1"])
    D1 --> D2(["Encoder Layer 2"])
    D2 --> D3(["Encoder Layer N"])
    D3 --> E(["LayerNorm"])
    E --> F(["MLP Head"])
    F --> G(["Output Logits"])
    LA1 -.-> D1
    LA2 -.-> D2
    LA3 -.-> D3
     LA1:::loraStyle
     LA2:::loraStyle
     LA3:::loraStyle
    classDef loraStyle fill:#e1f5fe,stroke:#0277bd,stroke-width:2px

---

Core Modules

Module	Description	Key Features
Embedding	Token embedding and positional encoding	Configurable vocab size, position embeddings
TransformerEncoder	Multi-layer transformer backbone	Self-attention, LoRA integration, checkpointing
LoRALinear	Low-rank adaptation layers	Configurable rank, memory-efficient updates
MLPHead	Output projection layer	Classification, regression, dropout
Config System	Centralized configuration	YAML/JSON config, CLI overrides
Data Utils	Preprocessing and augmentation	Built-in tokenization, custom loaders

---

Performance & Efficiency

Metric	Full Fine-tuning	LoRA Fine-tuning	Improvement
Trainable Parameters	125M	3.2M	97% reduction
Memory Usage	16GB	5GB	69% reduction
Training Time	6h	2h	67% faster
Storage per Task	500MB	12MB	98% smaller

Benchmarks: Transformer-Base, WikiText-103, RTX 3090

Supported model sizes: Transformer-Tiny, Transformer-Small, Transformer-Base, Transformer-Large

---

Advanced Configuration

Example LoRA config:

lora_config = {
    "rank": 16,
    "alpha": 32,
    "dropout": 0.1,
    "target_modules": ["attention.qkv", "attention.proj", "mlp.fc1", "mlp.fc2"],
    "merge_weights": False
}

Example training config:

model:
  name: "transformer_base"
  vocab_size: 50257
  embed_dim: 768
  num_layers: 12
  num_heads: 12
training:
  epochs: 10
  batch_size: 32
  learning_rate: 1e-4
  weight_decay: 0.01
  warmup_steps: 1000
lora:
  rank: 16
  alpha: 32
  dropout: 0.1

---

Documentation & Resources

• API Reference
• Tutorials and Examples
• Research Papers
• Best Practices Guide
• Troubleshooting

Research Papers

• LoRA: Low-Rank Adaptation of Large Language Models
• Attention Is All You Need
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

---

Testing & Quality

Run tests:

pytest tests/

Code quality tools:

flake8 src/
black src/ --check
mypy src/
bandit -r src/

---

Examples & Use Cases

Text classification:

from langtune import LanguageModel
from langtune.datasets import TextClassificationDataset

model = LanguageModel.from_pretrained("transformer_base")
dataset = TextClassificationDataset(train=True, tokenizer=model.tokenizer)
model.finetune(dataset, epochs=10, lora_rank=16)

Custom dataset:

from langtune.datasets import CustomTextDataset

dataset = CustomTextDataset(
    file_path="/path/to/dataset.txt",
    split="train",
    tokenizer=model.tokenizer
)
model.finetune(dataset, config_path="configs/custom_config.yaml")

---

Extending the Framework

• Add datasets in src/langtune/data/datasets.py
• Add callbacks in src/langtune/callbacks/
• Add models in src/langtune/models/
• Add CLI tools in src/langtune/cli/

Documentation

• See code comments and docstrings for details.
• For advanced usage, see src/langtune/cli/finetune.py.

Contributing

We welcome contributions. See the Contributing Guide for details.

License & Citation

This project is licensed under the MIT License. See LICENSE for details.

If you use langtune in your research, please cite:

@software{langtune2025,
  author = {Pritesh Raj},
  title = {langtune: LLMs with Efficient LoRA Fine-Tuning},
  url = {https://github.com/langtrain-ai/langtune},
  year = {2025},
  version = {0.1.0}
}

Acknowledgements

We thank the following projects and communities:

• PyTorch
• HuggingFace
• PEFT

Made in India 🇮🇳 with ❤️ by the langtune team
Star ⭐ this repo if you find it useful!