aakhundov
Software Engineer, ML Compilers @ Meta. Working on PyTorch, Triton, AITemplate.
MetaSan Francisco Bay Area
Pinned Repositories
aakhundov.github.io
Live in-browser demo of "Attend, Infer, Repeat" (Eslami et al., 2016, arXiv:1603.08575).
AITemplate-1
AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference.
c-generics
Examples of type-generic code in pure C (with void *). Inspired by Jerry Cain's "Programming Paradigms" lecture (Stanford CS107, 2008).
c-lispy
Simple lisp dialect interpreter in pure C.
c-scheme
Interpeter and compiler for Scheme programming language. Based on a register machine (with garbage collection). Implemented separately in C (done) and C++ (unfinished). Inspired by the "Wizard Book" (SICP).
deep-german
Classification of German nouns by genders using different RNN, CNN, and MLP architectures.
sequence-labeling
Accompanying repository of our NLP paper "Sequence Labeling: A Practical Approach" (arXiv: 1808.03926).
tf-attend-infer-repeat
TensorFlow-based implementation of "Attend, Infer, Repeat" paper (Eslami et al., 2016, arXiv:1603.08575).
tf-example-models
TensorFlow-based implementation of (Gaussian) Mixture Model and some other examples.
tf-rnn-adaptive
Adaptive Computation Time (Graves, 2016, arXiv:1603.08983) wrapper for TensorFlow RNN cells.
aakhundov's Repositories
aakhundov/tf-attend-infer-repeat
TensorFlow-based implementation of "Attend, Infer, Repeat" paper (Eslami et al., 2016, arXiv:1603.08575).
aakhundov/tf-example-models
TensorFlow-based implementation of (Gaussian) Mixture Model and some other examples.
aakhundov/deep-german
Classification of German nouns by genders using different RNN, CNN, and MLP architectures.
aakhundov/tf-rnn-adaptive
Adaptive Computation Time (Graves, 2016, arXiv:1603.08983) wrapper for TensorFlow RNN cells.
aakhundov/sequence-labeling
Accompanying repository of our NLP paper "Sequence Labeling: A Practical Approach" (arXiv: 1808.03926).
aakhundov/c-generics
Examples of type-generic code in pure C (with void *). Inspired by Jerry Cain's "Programming Paradigms" lecture (Stanford CS107, 2008).
aakhundov/c-scheme
Interpeter and compiler for Scheme programming language. Based on a register machine (with garbage collection). Implemented separately in C (done) and C++ (unfinished). Inspired by the "Wizard Book" (SICP).
aakhundov/aakhundov.github.io
Live in-browser demo of "Attend, Infer, Repeat" (Eslami et al., 2016, arXiv:1603.08575).
aakhundov/AITemplate-1
AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference.
aakhundov/c-lispy
Simple lisp dialect interpreter in pure C.
aakhundov/mnist-challenge
MNIST classification challenge held as a part of Machine Learning course at TU Munich.
aakhundov/cutlass
CUDA Templates for Linear Algebra Subroutines
aakhundov/FBGEMM
FB (Facebook) + GEMM (General Matrix-Matrix Multiplication) - https://code.fb.com/ml-applications/fbgemm/
aakhundov/lox
aakhundov/pymdown-extensions
Extensions for Python Markdown
aakhundov/pytorch
Tensors and Dynamic neural networks in Python with strong GPU acceleration
aakhundov/rl
A modular, primitive-first, python-first PyTorch library for Reinforcement Learning.
aakhundov/sphinx-action
Github action that builds docs using sphinx and places errors inline
aakhundov/triton
Development repository for the Triton language and compiler