Calculation Consulting 's repositories
WeightWatcher
The WeightWatcher tool for predicting the accuracy of Deep Neural Networks
ImplicitSelfRegularization
Repo to accompany paper "Implicit Self-Regularization in Deep Neural Networks..."
PredictingTestAccuracies
Repo to accompany paper: Heavy-Tailed Universality Predicts Trends in Test Accuracies for Very Large Pre-Trained Deep Neural Networks
ww-trends-2021
Repo to accompanying the paper: "Predicting trends in the quality of state-of-the-art neural networks without access to training or testing data"
RememberingGeneralization
Repo to accompany the paper: Rethinking generalization requires revisiting old ideas: statistical mechanics approaches and complex learning behavior
ww-phys_theory
Notebooks for the ww phys-theory channel
WeightWatcher-Examples
User Examples of the WeightWatcher project
cc-website
Calculation Consulting Website Using Bootstrap Agency theme
DLGeneralization
My project to recreate the results in "Understanding Deep Learning Requires Rethinking Generalization" and do something similar on NLP
LLM-Finetuning-Hub
Repository that contains LLM fine-tuning and deployment scripts along with our research findings.
CKA-Centered-Kernel-Alignment
Reproduce CKA: Similarity of Neural Network Representations Revisited
Deep-Siamese-Networks-with-Attention
This project explorers enhancements to traditional Siamese Architecture to achieve better accuracy by incorporating more semantic information. I have explored 3 enhancements :
fine-tuning-demo
deck and notebook for fine tuning demo
gpt-2-simple
Python package to easily retrain OpenAI's GPT-2 text-generating model on new texts
grokking
PyTorch implementation of "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets"
imgclsmob
Sandbox for training convolutional networks for computer vision
magnitude
A fast, efficient universal vector embedding utility package.
mediawiki
MediaWiki API wrapper in python http://pymediawiki.readthedocs.io/en/latest/
multihead-siamese-nets
Implementation of Siamese Neural Networks built upon multihead attention mechanism for text semantic similarity task.
pyRMT-1
Python library for Random Matrix Theory, cleaning schemes for correlation matrices, and portfolio optimization
pytorch-siamese
Siamese Network implementation using Pytorch
siamese-triplet
Siamese and triplet networks with online pair/triplet mining in PyTorch
White-box-Cartoonization
Official tensorflow implementation for CVPR2020 paper “Learning to Cartoonize Using White-box Cartoon Representations”