About
Hi, I am Ubaada. I am a Software Engineer based in New Zealand.
Projects
Word Map:
Maps how machine learning models associate word with different cultures and countries in vector
space.
VBE tool:
A simple tool to convert Variable Byte Encoding to and from decimal numbers.
Cleaned BookSum
dataset:
A cleaned version of the BookSum dataset published on HuggingFace. The dataset is a collection of book chapters, whole books,
and their summaries.
BookSum dataset is used for training and evaluating summarization machine learning models.
Summarization LLM Models:
Some Efficient Attention Transformer models fine-tuned on the BookSum dataset above for summarization.
Efficient Attention, as opposed to the regular attention mechanism used in Transformer language models like ChatGPT,
allows us to process longer sequences of text more efficiently with less hardware resources.
Search Engine
An information retriever (search engine) written in C for parsing and searching the
WSJ collection using an inverted index.
GitHub Stats
Recent Posts
September 2024Running Llama 3.1 70B on a Single Consumer-Grade GPU (RTX 4090 24GB) at 60 Tokens/s
July 2024Which Countries Seed the Most? Mapping Seeder IPs on Movie Torrents.
June 2024Comparing Different Colab GPU Offerings
March 2024Which features are considered most attractive in celebrities? Exploring CelebA dataset.
April 2023What if AI redrew the world map into new regions? Clustering ML text embeddings.