langvision 0.0.1__py3-none-any.whl → 0.0.2__py3-none-any.whl

langvision-0.0.2.dist-info/METADATA

@@ -0,0 +1,372 @@
+Metadata-Version: 2.4
+Name: langvision
+Version: 0.0.2
+Summary: A package for finetuning vision models.
+Author-email: Pritesh Raj <priteshraj10@gmail.com>
+License: MIT License
+        
+        Copyright (c) 2025 Plim
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE. 
+Project-URL: Homepage, https://github.com/langtrain-ai/langtrain
+Project-URL: Documentation, https://github.com/langtrain-ai/langtrain/tree/main/docs
+Project-URL: Source, https://github.com/langtrain-ai/langtrain
+Project-URL: Tracker, https://github.com/langtrain-ai/langtrain/issues
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: torch>=1.10
+Requires-Dist: numpy
+Requires-Dist: tqdm
+Requires-Dist: pyyaml
+Requires-Dist: scipy
+Requires-Dist: matplotlib
+Requires-Dist: pillow
+Dynamic: license-file
+
+# Langvision: Vision LLMs with Efficient LoRA Fine-Tuning
+
+<hr/>
+<p align="center">
+  <picture>
+    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/langtrain-ai/langtrain/main/static/langvision-use-dark.png">
+    <img alt="Langvision Logo" src="https://raw.githubusercontent.com/langtrain-ai/langtrain/main/static/langvision-white.png" width="full" />
+  </picture>
+</p>
+
+<!-- Badges -->
+<p align="center">
+  <a href="https://pypi.org/project/langvision/"><img src="https://img.shields.io/pypi/v/langvision.svg" alt="PyPI version"></a>
+  <a href="https://pepy.tech/project/langvision"><img src="https://pepy.tech/badge/langvision" alt="Downloads"></a>
+  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License"></a>
+  <a href="https://img.shields.io/badge/coverage-90%25-brightgreen" alt="Coverage"> <img src="https://img.shields.io/badge/coverage-90%25-brightgreen"/></a>
+  <a href="https://img.shields.io/badge/python-3.8%2B-blue" alt="Python Version"> <img src="https://img.shields.io/badge/python-3.8%2B-blue"/></a>
+  <a href="https://github.com/psf/black"><img src="https://img.shields.io/badge/code%20style-black-000000.svg" alt="Code style: black"></a>
+</p>
+
+<p align="center">
+  <b>Langvision provides modular components for vision models and LoRA-based fine-tuning.</b><br/>
+  <span style="font-size:1.1em"><i>Adapt and fine-tune vision models for a range of tasks.</i></span>
+</p>
+<hr/>
+
+## Quick Links
+- [Documentation](docs/index.md)
+- [Tutorials](docs/tutorials/index.md)
+- [Changelog](CHANGELOG.md)
+- [Contributing Guide](CONTRIBUTING.md)
+- [Roadmap](ROADMAP.md)
+
+---
+
+## Table of Contents
+- [Features](#features)
+- [Showcase](#showcase)
+- [Getting Started](#getting-started)
+- [Supported Python Versions](#supported-python-versions)
+- [Why langvision?](#why-langvision)
+- [Architecture Overview](#architecture-overview)
+- [Core Modules](#core-modules)
+- [Performance & Efficiency](#performance--efficiency)
+- [Advanced Configuration](#advanced-configuration)
+- [Documentation & Resources](#documentation--resources)
+- [Testing & Quality](#testing--quality)
+- [Examples & Use Cases](#examples--use-cases)
+- [Extending the Framework](#extending-the-framework)
+- [Contributing](#contributing)
+- [FAQ](#faq)
+- [Citation](#citation)
+- [Acknowledgements](#acknowledgements)
+- [License](#license)
+
+---
+
+## Features
+- LoRA adapters for parameter-efficient fine-tuning
+- Modular Vision Transformer (ViT) backbone
+- Model zoo for open-source visual 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
+
+Langvision is a framework for building and fine-tuning vision models with LoRA support. It is suitable for tasks such as image classification, visual question answering, and custom vision applications.
+
+---
+
+## Getting Started
+
+Install with pip:
+
+```bash
+pip install langvision
+```
+
+Minimal example:
+
+```python
+import torch
+from langvision.models.vision_transformer import VisionTransformer
+from langvision.utils.config import default_config
+
+x = torch.randn(2, 3, 224, 224)
+model = VisionTransformer(
+    img_size=default_config['img_size'],
+    patch_size=default_config['patch_size'],
+    in_chans=default_config['in_chans'],
+    num_classes=default_config['num_classes'],
+    embed_dim=default_config['embed_dim'],
+    depth=default_config['depth'],
+    num_heads=default_config['num_heads'],
+    mlp_ratio=default_config['mlp_ratio'],
+    lora_config=default_config['lora'],
+)
+
+with torch.no_grad():
+    out = model(x)
+    print('Output shape:', out.shape)
+```
+
+For more details, see the [Documentation](docs/index.md) and [src/langvision/cli/finetune.py](src/langvision/cli/finetune.py).
+
+---
+
+## Supported Python Versions
+- Python 3.8+
+
+---
+
+## Why langvision?
+
+- Parameter-efficient fine-tuning with LoRA adapters
+- Modular ViT backbone for flexible model design
+- Unified interface for open-source vision models
+- Designed for both research and production
+- Efficient memory usage for large models
+
+---
+
+## Architecture Overview
+
+Langvision uses a modular Vision Transformer backbone with LoRA adapters in attention and MLP layers. This allows adaptation of pre-trained models with fewer trainable parameters.
+
+### Model Data Flow
+
+```mermaid
+---
+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 Image"]) --> B(["Patch Embedding"])
+    B --> C(["CLS Token & 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 Class 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 |
+|--------|-------------|--------------|
+| PatchEmbedding | Image-to-patch conversion and embedding | Configurable patch sizes, 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 transforms, custom loaders |
+
+---
+
+## Performance & Efficiency
+
+| Metric | Full Fine-tuning | LoRA Fine-tuning | Improvement |
+|--------|------------------|------------------|-------------|
+| Trainable Parameters | 86M | 2.4M | 97% reduction |
+| Memory Usage | 12GB | 4GB | 67% reduction |
+| Training Time | 4h | 1.5h | 62% faster |
+| Storage per Task | 344MB | 9.6MB | 97% smaller |
+
+*Benchmarks: ViT-Base, CIFAR-100, RTX 3090*
+
+Supported model sizes: ViT-Tiny, ViT-Small, ViT-Base, ViT-Large
+
+---
+
+## Advanced Configuration
+
+Example LoRA config:
+
+```python
+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:
+
+```yaml
+model:
+  name: "vit_base"
+  img_size: 224
+  patch_size: 16
+  num_classes: 1000
+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](docs/api/index.md)
+- [Tutorials and Examples](docs/tutorials/index.md)
+- [Research Papers](#research-papers)
+- [Best Practices Guide](docs/best_practices.md)
+- [Troubleshooting](docs/troubleshooting.md)
+
+### Research Papers
+- [LoRA: Low-Rank Adaptation of Large Language Models](https://arxiv.org/abs/2106.09685)
+- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929)
+- [Vision Transformer for Fine-Grained Image Classification](https://arxiv.org/abs/2103.07579)
+
+---
+
+## Testing & Quality
+
+Run tests:
+
+```bash
+pytest tests/
+```
+
+Code quality tools:
+
+```bash
+flake8 src/
+black src/ --check
+mypy src/
+bandit -r src/
+```
+
+---
+
+## Examples & Use Cases
+
+Image classification:
+
+```python
+from langvision import VisionTransformer
+from langvision.datasets import CIFAR10Dataset
+
+model = VisionTransformer.from_pretrained("vit_base_patch16_224")
+dataset = CIFAR10Dataset(train=True, transform=model.default_transform)
+model.finetune(dataset, epochs=10, lora_rank=16)
+```
+
+Custom dataset:
+
+```python
+from langvision.datasets import ImageFolderDataset
+
+dataset = ImageFolderDataset(
+    root="/path/to/dataset",
+    split="train",
+    transform=model.default_transform
+)
+model.finetune(dataset, config_path="configs/custom_config.yaml")
+```
+
+---
+
+## Extending the Framework
+- Add datasets in `src/langvision/data/datasets.py`
+- Add callbacks in `src/langvision/callbacks/`
+- Add models in `src/langvision/models/`
+- Add CLI tools in `src/langvision/cli/`
+
+## Documentation
+- See code comments and docstrings for details.
+- For advanced usage, see `src/langvision/cli/finetune.py`.
+
+## Contributing
+We welcome contributions. See the [Contributing Guide](CONTRIBUTING.md) for details.
+
+## License & Citation
+
+This project is licensed under the MIT License. See [LICENSE](LICENSE) for details.
+
+If you use langvision in your research, please cite:
+
+```bibtex
+@software{langtrain2025,
+  author = {Pritesh Raj},
+  title = {langtrain: Vision LLMs with Efficient LoRA Fine-Tuning},
+  url = {https://github.com/langtrain-ai/langvision},
+  year = {2025},
+  version = {1.0.0}
+}
+```
+
+## Acknowledgements
+
+We thank the following projects and communities:
+- [PyTorch](https://pytorch.org/)
+- [HuggingFace](https://huggingface.co/)
+- [timm](https://github.com/rwightman/pytorch-image-models)
+- [PEFT](https://github.com/huggingface/peft)
+
+<p align="center">
+  <b>Made in India 🇮🇳 with ❤️ by the langtrain team</b><br/>
+  <i>Star ⭐ this repo if you find it useful!</i>
+</p>

{langvision-0.0.1.dist-info → langvision-0.0.2.dist-info}/RECORD

@@ -32,9 +32,9 @@ langvision/utils/config.py,sha256=0T5Vl8S0hVhCq1NTpiQnXFmsRxOidiYPsXGv51ruEBM,27
 langvision/utils/cuda.py,sha256=rpp0j-tvwGK6uWVe7fFTOS1BdZcsfRkw2AS5B08L-R0,1372
 langvision/utils/data.py,sha256=95U8Z_R2slwpaBVFrqyvcBA6kYINcEuzp1r5djvwQYg,277
 langvision/utils/device.py,sha256=TLVwXHFcildDyKh1Q7zczIyU702UJj7OHpsGhjRwLjk,552
-langvision-0.0.1.dist-info/licenses/LICENSE,sha256=S5-0GNdCWFpkatz1c_DXeMYJM1uiWBWx-LNGRcFNP70,1061
-langvision-0.0.1.dist-info/METADATA,sha256=5X3tjtJy8GuK7RVAm_DVJR7FL7quSK87WUNZk5B72WM,16076
-langvision-0.0.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-langvision-0.0.1.dist-info/entry_points.txt,sha256=9kxJHpRvjLKPMf88vdPS_Aed9nXr3HmoqwN7eEb_Nv4,69
-langvision-0.0.1.dist-info/top_level.txt,sha256=l3g51kgU-5cacP4nhASSfYADOCAwLC4UzlgVwuoJCgc,11
-langvision-0.0.1.dist-info/RECORD,,
+langvision-0.0.2.dist-info/licenses/LICENSE,sha256=S5-0GNdCWFpkatz1c_DXeMYJM1uiWBWx-LNGRcFNP70,1061
+langvision-0.0.2.dist-info/METADATA,sha256=nGqZ-YWp_swWrBUj6mWRMioae0SRWnVbTQA81XI6YNE,11285
+langvision-0.0.2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+langvision-0.0.2.dist-info/entry_points.txt,sha256=9kxJHpRvjLKPMf88vdPS_Aed9nXr3HmoqwN7eEb_Nv4,69
+langvision-0.0.2.dist-info/top_level.txt,sha256=l3g51kgU-5cacP4nhASSfYADOCAwLC4UzlgVwuoJCgc,11
+langvision-0.0.2.dist-info/RECORD,,

langvision-0.0.1.dist-info/METADATA

@@ -1,463 +0,0 @@
-Metadata-Version: 2.4
-Name: langvision
-Version: 0.0.1
-Summary: Vision LLMs with Efficient LoRA Fine-Tuning
-Author-email: Pritesh Raj <priteshraj10@gmail.com>
-License: MIT License
-        
-        Copyright (c) 2025 Plim
-        
-        Permission is hereby granted, free of charge, to any person obtaining a copy
-        of this software and associated documentation files (the "Software"), to deal
-        in the Software without restriction, including without limitation the rights
-        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-        copies of the Software, and to permit persons to whom the Software is
-        furnished to do so, subject to the following conditions:
-        
-        The above copyright notice and this permission notice shall be included in all
-        copies or substantial portions of the Software.
-        
-        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-        SOFTWARE. 
-Project-URL: Homepage, https://github.com/langtrain-ai/langtrain
-Project-URL: Documentation, https://github.com/langtrain-ai/langtrain/tree/main/docs
-Project-URL: Source, https://github.com/langtrain-ai/langtrain
-Project-URL: Tracker, https://github.com/langtrain-ai/langtrain/issues
-Requires-Python: >=3.8
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: torch>=1.10
-Requires-Dist: numpy
-Requires-Dist: tqdm
-Requires-Dist: pyyaml
-Requires-Dist: scipy
-Requires-Dist: matplotlib
-Requires-Dist: pillow
-Dynamic: license-file
-
-# langtrain: Vision LLMs (Large Language Models for Vision) with Efficient LoRA Fine-Tuning
-
-<hr/>
-<p align="center">
-  <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/langtrain-ai/langtrain/main/static/langvision-use-dark.png">
-    <img alt="Langvision Logo" src="https://raw.githubusercontent.com/langtrain-ai/langtrain/main/static/langvision-white.png" width="full" />
-  </picture>
-</p>
-
-<!-- Badges -->
-<p align="center">
-  <a href="https://pypi.org/project/langvision/"><img src="https://img.shields.io/pypi/v/langvision.svg" alt="PyPI version"></a>
-  <a href="https://pepy.tech/project/langvision"><img src="https://pepy.tech/badge/langvision" alt="Downloads"></a>
-  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License"></a>
-  <a href="https://img.shields.io/badge/coverage-90%25-brightgreen" alt="Coverage"> <img src="https://img.shields.io/badge/coverage-90%25-brightgreen"/></a>
-  <a href="https://img.shields.io/badge/python-3.8%2B-blue" alt="Python Version"> <img src="https://img.shields.io/badge/python-3.8%2B-blue"/></a>
-  <a href="https://github.com/psf/black"><img src="https://img.shields.io/badge/code%20style-black-000000.svg" alt="Code style: black"></a>
-</p>
-
-<p align="center">
-  <b>Modular Vision LLMs (Large Language Models for Vision) with Efficient LoRA Fine-Tuning</b><br/>
-  <span style="font-size:1.1em"><i>Build, adapt, and fine-tune vision models with ease and efficiency.</i></span>
-</p>
-<hr/>
-
-## 🚀 Quick Links
-- [Documentation](docs/index.md)
-- [Tutorials](docs/tutorials/index.md)
-- [Changelog](CHANGELOG.md)
-- [Contributing Guide](CONTRIBUTING.md)
-- [Roadmap](ROADMAP.md)
-
----
-
-## 📚 Table of Contents
-- [Features](#-features)
-- [Showcase](#-showcase)
-- [Getting Started](#-getting-started)
-- [Supported Python Versions](#-supported-python-versions)
-- [Why langvision?](#-why-langvision)
-- [Architecture Overview](#-architecture-overview)
-- [Core Modules](#-core-modules)
-- [Performance & Efficiency](#-performance--efficiency)
-- [Advanced Configuration](#-advanced-configuration)
-- [Documentation & Resources](#-documentation--resources)
-- [Testing & Quality](#-testing--quality)
-- [Examples & Use Cases](#-examples--use-cases)
-- [Extending the Framework](#-extending-the-framework)
-- [Contributing](#-contributing)
-- [FAQ](#-faq)
-- [Citation](#-citation)
-- [Acknowledgements](#-acknowledgements)
-- [License](#-license)
-
----
-
-## ✨ Features
-- 🔧 **Plug-and-play LoRA adapters** for parameter-efficient fine-tuning
-- 🏗️ **Modular Vision Transformer (ViT) backbone** with customizable components
-- 🎯 **Unified model zoo** for open-source visual models
-- ⚙️ **Easy configuration** and extensible codebase
-- 🚀 **Production ready** with comprehensive testing and documentation
-- 💾 **Memory efficient** training with gradient checkpointing support
-- 📊 **Built-in metrics** and visualization tools
-- 🧩 **Modular training loop** with LoRA support
-- 🎯 **Unified CLI** for fine-tuning and evaluation
-- 🔌 **Extensible callbacks** (early stopping, logging, etc.)
-- 📦 **Checkpointing and resume**
-- 🚀 **Mixed precision training**
-- 🔧 **Easy dataset and model extension**
-- ⚡ **Ready for distributed/multi-GPU training**
-
----
-
-## 🚀 Showcase
-
-**langvision** is a modular, research-friendly framework for building and fine-tuning Vision Large Language Models (LLMs) with efficient Low-Rank Adaptation (LoRA) support. Whether you're working on image classification, visual question answering, or custom vision tasks, langvision provides the tools you need for parameter-efficient model adaptation.
-
----
-
-## 🏁 Getting Started
-
-Here's a minimal example to get you up and running:
-
-```bash
-pip install langvision
-```
-
-```python
-import torch
-from langvision.models.vision_transformer import VisionTransformer
-from langvision.utils.config import default_config
-
-# Create model
-x = torch.randn(2, 3, 224, 224)
-model = VisionTransformer(
-    img_size=default_config['img_size'],
-    patch_size=default_config['patch_size'],
-    in_chans=default_config['in_chans'],
-    num_classes=default_config['num_classes'],
-    embed_dim=default_config['embed_dim'],
-    depth=default_config['depth'],
-    num_heads=default_config['num_heads'],
-    mlp_ratio=default_config['mlp_ratio'],
-    lora_config=default_config['lora'],
-)
-
-# Forward pass
-with torch.no_grad():
-    out = model(x)
-    print('Output shape:', out.shape)
-```
-
-For advanced usage, CLI details, and more, see the [Documentation](docs/index.md) and [src/langvision/cli/finetune.py](src/langvision/cli/finetune.py).
-
----
-
-## 🐍 Supported Python Versions
-- Python 3.8+
-
----
-
-## 🧩 Why langvision?
-
-- **Parameter-efficient fine-tuning**: Plug-and-play LoRA adapters for fast, memory-efficient adaptation with minimal computational overhead
-- **Modular ViT backbone**: Swap or extend components like patch embedding, attention, or MLP heads with ease
-- **Unified model zoo**: Access and experiment with open-source visual models through a consistent interface
-- **Research & production ready**: Clean, extensible codebase with comprehensive configuration options and robust utilities
-- **Memory efficient**: Fine-tune large models on consumer hardware by updating only a small fraction of parameters
-
----
-
-## 🏗️ Architecture Overview
-
-langvision is built around a modular Vision Transformer (ViT) backbone, with LoRA adapters strategically injected into attention and MLP layers for efficient fine-tuning. This approach allows you to adapt large pre-trained models using only a fraction of the original parameters.
-
-### Model Data Flow
-
-```mermaid
----
-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 Image"]) --> B(["Patch Embedding"])
-    B --> C(["CLS Token & 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 Class Logits"])
-    LA1 -.-> D1
-    LA2 -.-> D2
-    LA3 -.-> D3
-     LA1:::loraStyle
-     LA2:::loraStyle
-     LA3:::loraStyle
-    classDef loraStyle fill:#e1f5fe,stroke:#0277bd,stroke-width:2px
-```
-
-### Architecture Components
-
-**Legend:**
-- **Solid arrows**: Main data flow through the Vision Transformer
-- **Dashed arrows**: LoRA adapter injection points in encoder layers
-- **Blue boxes**: LoRA adapters for parameter-efficient fine-tuning
-
-**Data Flow Steps:**
-1. **Input Image** (224×224×3): Raw image data ready for processing
-2. **Patch Embedding**: Image split into 16×16 patches and projected to embedding dimension
-3. **CLS Token & Positional Encoding**: Classification token prepended with learnable position embeddings
-4. **Transformer Encoder Stack**: Multi-layer transformer with self-attention and MLP blocks
-   - **LoRA Integration**: Low-rank adapters injected into attention and MLP layers
-   - **Efficient Updates**: Only LoRA parameters updated during fine-tuning
-5. **LayerNorm**: Final normalization of encoder outputs
-6. **MLP Head**: Task-specific classification or regression head
-7. **Output**: Final predictions (class probabilities, regression values, etc.)
-
----
-
-## 🧩 Core Modules
-
-| Module | Description | Key Features |
-|--------|-------------|--------------|
-| **PatchEmbedding** | Image-to-patch conversion and embedding | • Configurable patch sizes<br>• Learnable position embeddings<br>• Support for different input resolutions |
-| **TransformerEncoder** | Multi-layer transformer backbone | • Self-attention mechanisms<br>• LoRA adapter integration<br>• Gradient checkpointing support |
-| **LoRALinear** | Low-rank adaptation layers | • Configurable rank and scaling<br>• Memory-efficient updates<br>• Easy enable/disable functionality |
-| **MLPHead** | Output projection layer | • Multi-class classification<br>• Regression support<br>• Dropout regularization |
-| **Config System** | Centralized configuration management | • YAML/JSON config files<br>• Command-line overrides<br>• Validation and defaults |
-| **Data Utils** | Preprocessing and augmentation | • Built-in transforms<br>• Custom dataset loaders<br>• Efficient data pipelines |
-
----
-
-## 📊 Performance & Efficiency
-
-### LoRA Benefits
-
-| Metric | Full Fine-tuning | LoRA Fine-tuning | Improvement |
-|--------|------------------|------------------|-------------|
-| **Trainable Parameters** | 86M | 2.4M | **97% reduction** |
-| **Memory Usage** | 12GB | 4GB | **67% reduction** |
-| **Training Time** | 4 hours | 1.5 hours | **62% faster** |
-| **Storage per Task** | 344MB | 9.6MB | **97% smaller** |
-
-*Benchmarks on ViT-Base with CIFAR-100, RTX 3090*
-
-### Supported Model Sizes
-
-- **ViT-Tiny**: 5.7M parameters, perfect for experimentation
-- **ViT-Small**: 22M parameters, good balance of performance and efficiency  
-- **ViT-Base**: 86M parameters, strong performance across tasks
-- **ViT-Large**: 307M parameters, state-of-the-art results
-
----
-
-## 🔧 Advanced Configuration
-
-### LoRA Configuration
-
-```python
-lora_config = {
-    "rank": 16,                    # Low-rank dimension
-    "alpha": 32,                   # Scaling factor
-    "dropout": 0.1,                # Dropout rate
-    "target_modules": [            # Modules to adapt
-        "attention.qkv",
-        "attention.proj", 
-        "mlp.fc1",
-        "mlp.fc2"
-    ],
-    "merge_weights": False         # Whether to merge during inference
-}
-```
-
-### Training Configuration
-
-```yaml
-# config.yaml
-model:
-  name: "vit_base"
-  img_size: 224
-  patch_size: 16
-  num_classes: 1000
-
-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
-
-- 📖 [Complete API Reference](docs/api/index.md)
-- 🎓 [Tutorials and Examples](docs/tutorials/index.md)
-- 🔬 [Research Papers](#research-papers)
-- 💡 [Best Practices Guide](docs/best_practices.md)
-- 🐛 [Troubleshooting](docs/troubleshooting.md)
-
-### Research Papers
-- [LoRA: Low-Rank Adaptation of Large Language Models](https://arxiv.org/abs/2106.09685)
-- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929)
-- [Vision Transformer for Fine-Grained Image Classification](https://arxiv.org/abs/2103.07579)
-
----
-
-## 🧪 Testing & Quality
-
-Run the comprehensive test suite:
-
-```bash
-# Unit tests
-pytest tests/unit/
-
-# Integration tests  
-pytest tests/integration/
-
-# Performance benchmarks
-pytest tests/benchmarks/
-
-# All tests with coverage
-pytest tests/ --cov=langvision --cov-report=html
-```
-
-### Code Quality Tools
-
-```bash
-# Linting
-flake8 src/
-black src/ --check
-
-# Type checking
-mypy src/
-
-# Security scanning
-bandit -r src/
-```
-
----
-
-## 🚀 Examples & Use Cases
-
-### Image Classification
-```python
-from langvision import VisionTransformer
-from langvision.datasets import CIFAR10Dataset
-
-# Load pre-trained model
-model = VisionTransformer.from_pretrained("vit_base_patch16_224")
-
-# Fine-tune on CIFAR-10
-dataset = CIFAR10Dataset(train=True, transform=model.default_transform)
-model.finetune(dataset, epochs=10, lora_rank=16)
-```
-
-### Custom Dataset
-```python
-from langvision.datasets import ImageFolderDataset
-
-# Your custom dataset
-dataset = ImageFolderDataset(
-    root="/path/to/dataset",
-    split="train",
-    transform=model.default_transform
-)
-
-# Fine-tune with custom configuration
-model.finetune(
-    dataset, 
-    config_path="configs/custom_config.yaml"
-)
-```
-
----
-
-## 🧩 Extending the Framework
-- Add new datasets in `src/langvision/data/datasets.py`
-- Add new callbacks in `src/langvision/callbacks/`
-- Add new models in `src/langvision/models/`
-- Add new CLI tools in `src/langvision/cli/`
-
-## 📖 Documentation
-- See code comments and docstrings for details on each module.
-- For advanced usage, see the `src/langvision/cli/finetune.py` script.
-
-## 🤝 Contributing
-We welcome contributions from the community! Here's how you can get involved:
-
-### Ways to Contribute
-- 🐛 **Report bugs** by opening issues with detailed reproduction steps
-- 💡 **Suggest features** through feature requests and discussions
-- 📝 **Improve documentation** with examples, tutorials, and API docs
-- 🔧 **Submit pull requests** for bug fixes and new features
-- 🧪 **Add tests** to improve code coverage and reliability
-
-### Development Setup
-```bash
-# Clone and setup development environment
-git clone https://github.com/langtrain-ai/langvision.git
-cd langvision
-pip install -e ".[dev]"
-
-# Install pre-commit hooks
-pre-commit install
-
-# Run tests
-pytest tests/
-```
-
-### Community Resources
-- 💬 [GitHub Discussions](https://github.com/langtrain-ai/langvision/discussions) - Ask questions and share ideas
-- 🐛 [Issue Tracker](https://github.com/langtrain-ai/langvision/issues) - Report bugs and request features
-- 📖 [Contributing Guide](CONTRIBUTING.md) - Detailed contribution guidelines
-- 🎯 [Roadmap](ROADMAP.md) - See what's planned for future releases
-
-## 📄 License & Citation
-
-This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
-
-### Citation
-
-If you use langtrain in your research, please cite:
-
-```bibtex
-@software{langtrain2025,
-  author = {Pritesh Raj},
-  title = {langtrain: Vision LLMs with Efficient LoRA Fine-Tuning},
-  url = {https://github.com/langtrain-ai/langvision},
-  year = {2025},
-  version = {1.0.0}
-}
-```
-
-## 🌟 Acknowledgements
-
-We thank the following projects and communities:
-
-- [PyTorch](https://pytorch.org/) - Deep learning framework
-- [HuggingFace](https://huggingface.co/) - Transformers and model hub
-- [timm](https://github.com/rwightman/pytorch-image-models) - Vision model implementations
-- [PEFT](https://github.com/huggingface/peft) - Parameter-efficient fine-tuning methods
-
-<p align="center">
-  <b>Made in India 🇮🇳 with ❤️ by the langtrain team</b><br/>
-  <i>Star ⭐ this repo if you find it useful!</i>
-</p>