This repository provides PyTorch implementations and experimental code for replicating results from a wide range of influential AI/ML research papers, covering classic and state-of-the-art models. It serves researchers and practitioners looking to understand, reproduce, and build upon foundational and cutting-edge AI architectures.

How It Works

The project is structured into self-contained subfolders, each dedicated to a specific paper, model, or technique (e.g., BERT, GPT, ViT, CycleGANs, Llama, Whisper). This modular approach allows users to easily navigate and utilize code for individual replications, with specific instructions typically found within each subfolder's README.

Quick Start & Requirements

• Installation: Clone the repository and navigate to the specific subfolder of interest.
• Prerequisites: PyTorch is the primary framework. Specific model requirements (e.g., CUDA, specific Python versions, large datasets, API keys) will vary per subfolder and should be checked in their respective READMEs.
• Links: HuggingFace Account: https://huggingface.co/YuvrajSingh

Highlighted Details

• Comprehensive coverage of diverse AI domains: NLP (BERT, GPT, Llama), Vision (ViT, GANs), Speech (Whisper), and multimodal models (CLIP, Llava).
• Includes implementations of parameter-efficient fine-tuning (PEFT) methods like LoRA and DPO.
• Features replications of recent SOTA models such as DeepSeekV3, Gemma, and Mixtral.
• Offers examples of attention mechanisms, RNN variants (GRU, LSTM), and sequence-to-sequence models.

Maintenance & Community

The repository is maintained by @YuvrajSingh. Contributions are welcomed via issues and pull requests.

Licensing & Compatibility

The repository is stated to be for educational and research purposes. Users are advised to check individual folders for additional license or citation requirements, implying potential variations in licensing across sub-projects.

Limitations & Caveats

The project's scope is broad, and specific replication accuracy or performance benchmarks are not centrally detailed. Users must consult individual subfolder documentation for precise setup, dependencies, and usage instructions, as these can vary significantly between different paper replications.