Gabriel Mongaras

Diffusion_models_from_scratch. Creating a diffusion model from scratch in PyTorch to learn exactly how they work.

AI_Girlfriend. Creating a waifu

Stable-Diffusion-3-From-Scratch. A repo that attempts to train stable diffusion 3 from scratch

YOLOX_From_Scratch. Attempting to build YOLOX from scratch

Latent_Diffusion_Model_Imagenet2012. A latent flow-based diffusion model trained on the 2012 ImageNet dataset from scratch.

Cottention_Transformer. Code for the paper "Cottention: Linear Transformers With Cosine Attention"

Llama-2_Huggingface_4Bit_QLoRA. A working example of a 4bit QLoRA Falcon model using huggingface

AI_Girlfriend_Medium. Code that goes with my AI Girlfriend medium article.

PyTorchLite_Demo. Demo for PyTorch Lite on Android for an article I wrote on Medium

Wizard_QLoRA_Finetuning. Finetuning Some Wizard Models With QLoRA

On-the-Expressiveness-of-Softmax-Attention-A-Recurrent-Neural-Network-Perspective. Jupyter Notebook

AI_Girlfriend_Reduced. Reduced version of AI Girlfriend for faster downloading

Consistency_Models. Jupyter Notebook

Learnable_Rotary_Embeddings. Just some tests I ran

Anime_StyleGAN_Website. Website that uses StyleGAN2 to produce fake anime pictures.

GRPO-DAPO-Tests. Playing around with GRPO and DAPO

MetaU_Capstone. Get a fortune a day that can either be generated using an AI I built from scratch or by picking from a list of real fortunes. Then, add some friends and share your fortunes! This app was created for MetaU

Reinforcement_Learning_Stuff. I need to learn some RL

PPO_CartPole. Using PPO, I am attempting to solve the cartpole environment

Protogen_Code_Public. Public code for my protogen

Visualizing_Gradient_Descent_For_BCE_Loss. Using only numpy in Python, a neural network with a forward and backward method is used to classify given points (x1, x2) to a color of red or blue.

Dino_Game_AI_V2. AI that learns to play the Google dino game using an improved version of NEAT

GAN_TextGen. Attempting to use a GAN to generate random sentences

Dino_Game_AI. AI that learns to play the Google dino game

Token_Merging_Tests. Jupyter Notebook

Diffusion-TTS. Diffusion TTS - Synthesized voice using stable diffusion to map between noise and stylized speech