my personal website

personal rl environment / framework for games. For self teaching purposes

repo for my personal llm experiments in using different optimizers, attention mechs, data preprocessing strategies, quantization schemes, and other stuff. Optimized for my nvidia rtx 4080 w 12gb of…

notebook for personal project progress and other stuff

medium-sized language model based on deepseek (in development)

custom transformer-based image segmentation model

pure pytorch implementation of LeNet-5, a handwritten digit classifier; detailed documentation and comments included. model specifications derived from original paper.

company name generator based on a bigram and trigram

small language model using the megabyte architecture inspired by the transformer

implementation of gpt's tokenizer (byte pair encoding) for personal learning purposes

implementing a custom small transformer like gpt-2 in pytorch

An AI image generator that stylizes a photo given 2 inputs: the image whose style to mimic, the picture to apply it on

testing gpt-fast inferencing using triton on wsl

running a mistral-7b instruct locally and experimenting with it

changing the style transfer model's backbone classifier from VGG into something that can handle variable input sizes

pytorch implementation of FCN-8 described in the attached paper. trained on more difficult-to-segment cityscape dataset

test training an image segmentation model using swin-v2-t as a backbone and a skip connections with fusion (fpn).

implementing the xor classifier neural net in c++ to learn the language

a multi-layered (2 -> 2 ->1) perceptron made without pytorch or other deep learning frameworks; classifies the XOR samples.

script to automate downloading the first search result as an mp3 using: yt_download 'search'