Stas Bekman

ml-engineering. Machine Learning Engineering Open Book

the-art-of-debugging. The Art of Debugging Open Book

ipyexperiments. Automatic GPU+CPU memory profiling, re-use and memory leaks detection using jupyter/ipython experiment containers

ml-ways. ML/DL Math and Method notes

porting. Helper scripts and notes that were used while porting various nlp models

python-tools. Python tools

git-tools. helper git tools

stas00.

reddit-to-threads. Convert arctic_shift Reddit data dumps into thread-view documents

fastai-transcript. Video Transcripts of fast.ai MOOC courses made into searchable ebooks

transformers. 🤗Transformers: State-of-the-art Natural Language Processing for Pytorch and TensorFlow 2.0.

conda-tools. helper tools for the conda environment

gpt-neo-fine-tuning-example. Fine-Tune EleutherAI GPT-Neo And GPT-J-6B To Generate Netflix Movie Descriptions Using Hugginface And DeepSpeed

google-earth-grid. a very basic Google Earth kml generator

pytest-ipynb. A pytest plugin to run ipython/jupyter notebook cells as unit/integration tests

Sophia. Effortless plugin and play Optimizer to cut model training costs by 50%. New optimizer that is 2x faster than Adam on LLMs.

jpdftweak. Java

pytorch. Tensors and Dynamic neural networks in Python with strong GPU acceleration

backups. an effort to resist Internet entropy - save useful knowledge

jupyter-notebook-tools. jupyter-notebook-tools

bash-tools. Bash tools

ipygpulogger. GPU Logger for jupyter/ipython (memory usage, execution time)

accelerate. A simple way to train and use NLP models with multi-GPU, TPU, mixed-precision

DeepSpeed. DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.

FlexFlow. A distributed deep learning framework that supports flexible parallelization strategies.

datasets. 🤗 Fast, efficient, open-access datasets and evaluation metrics for Natural Language Processing and more in PyTorch, TensorFlow, NumPy and Pandas

peft. 🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.

cookbook. Deep learning for dummies. All the practical details and useful utilities that go into working with real models.

PdfMerger. Merges pdfs and creates handy outline

functorch. functorch is a prototype of JAX-like composable function transforms for PyTorch.

Megatron-LM. Ongoing research training transformer models at scale

pytorch-communication-benchmarks. pytorch code examples for measuring the performance of collective communication calls in AI workloads

psg. Pocket Survival Guide for Sys Admin - http://psg.skinforum.org/ -

TransformerSizing. Python

io-watchdog. Monitor processes and parallel workloads for hangs

unix-tools. Miscellaneous tools

vllm. A high-throughput and memory-efficient inference and serving engine for LLMs

code. Codebase for the project

transformers-bloom-inference. Fast Inference Solutions for BLOOM

Megatron-DeepSpeed. Ongoing research training transformer language models at scale, including: BERT & GPT-2

hpc-toolkit. Cloud HPC Toolkit is an open-source software offered by Google Cloud which makes it easy for customers to deploy HPC environments on Google Cloud.

AutoAWQ. AutoAWQ implements the AWQ algorithm for 4-bit quantization with a 2x speedup during inference. Documentation:

tinypar. TP/PP/DP implementation of llama using apex blocks

make-tools. make tools

tensorboard-tools. tensorboard tools

NeMo. NeMo: a toolkit for conversational AI

k6-docs. The k6 documentation website.

machine-learning-note. my machine learning note

apex. A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch

fairseq. Facebook AI Research Sequence-to-Sequence Toolkit written in Python.

benchmarks. Benchmarks are the key to sustainable software developement

nestedtensor. [Prototype] Tools for the concurrent manipulation of variably sized Tensors.

pytorch_memlab. Profiling and inspecting memory in pytorch

nvidia-system-monitor-qt. Task Manager for Linux for Nvidia graphics cards

model_card.

DeepSpeedExamples. Example models using DeepSpeed