edit

GPUs and Performance

GPUs

PipelineAI supports GPUs natively throughout the entire platform.

Nvidia GPU

Here are some publically-available resources including Docker images, source code, Nvidia driver and toolkit configuration, videos, slides, and workshop materials that demonstrate PipelineAI's support for GPUs.

GitHub Repo

fluxcapacitor/pipeline/gpu.ml

fluxcapacitor/pipeline/package.ml

Docker Images

AWS (GPU) + TensorFlow + Spark + HDFS + Docker

Google Cloud (GPU) + TensorFlow + Spark + HDFS + Docker

CPU Versions are Below.

Performance

PipelineAI maintains a collection of Docker Images with many optimizations already enabled including OpenBLAS, AVX, AVX2, FMA, etc.

GitHub Repo

fluxcapacitor/pipeline/gpu.ml

fluxcapacitor/pipeline/package.ml

Docker Images

AWS (CPU) + TensorFlow + Spark + HDFS + Docker

Google Cloud (CPU) + TensorFlow + Spark + HDFS + Docker

ML/AI Model Performance Optimizations

Click HERE for examples of optimizing model prediction performance with PipelineAI.

TensorFlow Optimization Example

Before Optimization

This model is very slow at prediction time.

Unoptimized TensorFlow Model

After Optimization

This model is much faster at prediction time!

Optimized TensorFlow Model

Recent Events

Videos and Slides


Try our Free Community Edition!

Click HERE to get started with our basic Community Edition

Register for PipelineAI Advanced Edition


Take a Quick, Fun Survey!


Help Us Improve

Option 1: Edit the documentation directly.

Option 2: Create a GitHub issue.