dblalock
Research scientist at MosaicML. MIT PhD. I build fast machine learning algorithms.
MosaicMLSan Francisco, CA
Pinned Repositories
bolt
10x faster matrix and vector operations
scons-example
An example C/C++ project showing how to build a shared library and tests in nested subdirectories using Scons.
sprintz
Sprintz time series compression algorithm
swig-numpy-example
Example project showing how to interface C/C++ arrays and classes with Numpy using SWIG
composer
Supercharge Your Model Training
examples
Fast and flexible reference benchmarks
llm-foundry
LLM training code for Databricks foundation models
streaming
A Data Streaming Library for Efficient Neural Network Training
dblalock's Repositories
dblalock/bolt
10x faster matrix and vector operations
dblalock/sprintz
Sprintz time series compression algorithm
dblalock/scons-example
An example C/C++ project showing how to build a shared library and tests in nested subdirectories using Scons.
dblalock/swig-numpy-example
Example project showing how to interface C/C++ arrays and classes with Numpy using SWIG
dblalock/compress
Moved to https://github.com/dblalock/sprintz
dblalock/llm-foundry
LLM training code for MosaicML foundation models
dblalock/lzbench
fork of lzbench specialized for benchmarking time series compression
dblalock/benchmarking-keras-pytorch
Reproducible Benchmarking of Keras and PyTorch Models
dblalock/composer
Composing methods for ML training efficiency
dblalock/cuda_ext_example
Simple example project that builds a PyTorch CUDA extension
dblalock/fastdist
Fast Dynamic Time Warping and other distances
dblalock/AdrenaliNN
dblalock/arr
dblalock/bolt-paper
dblalock/cifar-vgg
VGG16 models for CIFAR-10 and CIFAR-100 using Keras
dblalock/conv
Fast CUDA kernels
dblalock/dblalock.github.io
:triangular_ruler: A flexible two-column Jekyll theme perfect for building personal sites, blogs, and portfolios.
dblalock/docker-data-sharing-example
Example of getting data into and out of a container. Uses --mount and doesn't need to run as root.
dblalock/dollars
dblalock/examples
Fast and flexible reference benchmarks
dblalock/fft-conv-pytorch
Implementation of 1D, 2D, and 3D FFT convolutions in PyTorch. Much faster than direct convolutions for large kernel sizes.
dblalock/gitbook
dblalock/hipsterplot
Command line plotting for python
dblalock/hyperspacebar
eye of sauron
dblalock/imagenet_validation
dblalock/MakePytorchPlusPlus
How and why you want to make your pytorch CUDA/CPP extension with a Makefile
dblalock/megablocks
dblalock/old-dblalock.github.io
Blog
dblalock/paper-threader
I can't be bothered to write twitter threads
dblalock/training_results_v1.0