January 09, 2023
PyTorch Trace Analysis for the Masses
We are excited to announce the public release of Holistic Trace Analysis (HTA), an open source performance analysis and visualization Python library for PyTorch users. HTA takes as input Kineto traces collected by the PyTorch profiler, which are complex and challenging to interpret, and up-levels the performance information contained in these traces. It was initially developed internally at Meta to understand and debug performance problems for large-scale distributed training jobs on GPUs. Th...
December 31, 2022
Compromised PyTorch-nightly dependency chain between December 25th and December 30th, 2022.
If you installed PyTorch-nightly on Linux via pip between December 25, 2022 and December 30, 2022, please uninstall it and torchtriton immediately, and use the latest nightly binaries (newer than Dec 30th 2022).
December 28, 2022
Torchserve Performance Tuning, Animated Drawings Case-Study
In this post we discuss performance tuning of Torchserve for serving your models in production. One of the biggest challenges in the life cycle of a ML project is deploying models in production. This requires a reliable serving solution along with solutions that address the MLOps needs. A robust serving solution needs to provide support for multi model serving, model versioning, metric logging, monitoring and scaling to serve the peak traffic. In this post, we will have an overview of Torchs...
December 22, 2022
Scaling Vision Model Training Platforms with PyTorch
TL;DR: We demonstrate the use of PyTorch with FairScale’s FullyShardedDataParallel (FSDP) API in writing large vision transformer models. We discuss our techniques for scaling and optimizing these models on a GPU cluster. The goal of this platform scaling effort is to enable research at scale. This blog does not discuss model accuracy, new model architectures, or new training recipes.
December 16, 2022
Efficient Large-Scale Training with Pytorch FSDP and AWS
Cutting-edge AI models are becoming extremely large. The cost and overhead of training these models is increasing rapidly, and involves large amounts of engineering and guesswork to find the right training regime. FSDP reduces these costs significantly by enabling you to train much larger models with the same amount of resources. FSDP lowers the memory footprint on your GPUs, and is usable via a lightweight configuration that requires substantially less effort, typically with just a few lines...
December 15, 2022
Scaling PyTorch FSDP for Training Foundation Models on IBM Cloud
Large model training using a cloud native approach is of growing interest for many enterprises given the emergence and success of foundation models. Some AI practitioners may assume that the only way they can achieve high GPU utilization for distributed training jobs is to run them on HPC systems, such as those inter-connected with Infiniband and may not consider Ethernet connected systems. We demonstrate how the latest distributed training technique, Fully Sharded Data Parallel (FSDP) from P...
December 02, 2022
Get Started with PyTorch 2.0 Summary and Overview
Introducing PyTorch 2.0, our first steps toward the next generation 2-series release of PyTorch. Over the last few years we have innovated and iterated from PyTorch 1.0 to the most recent 1.13 and moved to the newly formed PyTorch Foundation, part of the Linux Foundation.