- Release Notes
- Requirements
- Hardware and software requirements
- AI Fabric architecture
- Installation
- Getting Started
- Projects
- Datasets
- ML Packages
- Pipelines
- ML Skills
- ML Logs
- Document Understanding in AI Fabric
- Basic Troubleshooting Guide
AI Center
Hardware and software requirements
The table below makes some recommendations averaged across generic models (small and large).
CPU | RAM (GB) | OS/Boot Disk (GB) | External Data Disk (GB) | Models Served | Concurrent Models Trained |
---|---|---|---|---|---|
8 | 52 | 200 | 500 | 3 | 1-2 |
12 | 64 | 200 | 1000 | 3-4 | 2 |
There is no universal value for how much resources one ML Skill/Pipeline job will consume as it depends of the model. However, here are minimum resources used by an ML Skill/Pipeline job along with resources used by a UiPath Document Understanding Model as a baseline.
Use | CPU | RAM (GB) |
---|---|---|
Core Services | 4 | 10 |
Minimum for Serving (ML SKill) | 0.5 | 2 |
Minimum for Training (Pipeline) | 1 | 4 |
DU model Serving | 1 | 4 |
DU model Training (500 images) | 2 | 24 |
disk
not partition
. See step 1. Provision a
Machine.
Only NVIDIA GPUs are currently supported. Most scenarios will not require training on a GPU, as the majority of model architectures can execute with both GPU and CPU. If you have constrains on model training time, it is recommended you add a GPU with at least 8 GB of Video RAM. You are responsible for installing GPU drivers before you can use GPU in AI Fabric. For more information on this, see the Prerequisite section.
Trainable Document Understanding ML Packages provided by UiPath will work on both CPU or on GPU for datasets up to 500 images in size. GPU is strongly recommended to achieve faster training times and better model performance. Validation Station retraining loop is not supported on deployments without a GPU, due to the fact that dataset sizes increase too fast and may hit a compute wall with CPU very quickly. If you run UiPath OCR (non-edge version) on AI Center to process more than 2 million pages of documents a year, GPU is strongly recommended for better product experience.
The following table lists the operating system(s) officially supported for the AI Fabric on-premises installation.
OS | Version |
---|---|
Ubuntu | 18.04 LTS |
RHEL | 7.4, 7.5, 7.6, 7.7, 7.8, 7.9 |
CentOS | 7.4, 7.5, 7.6, 7.7, 7.8, 7.9 |
- Machine should have lvm2 installed.
- OverlayFS storage drivers for docker should be usable, you can check prerequisites on docker documentation. No need to install anything just fulfill the prerequistes.
Before starting the installation, the following prerequisites are needed:
- Orchestrator 20.4.3 (or
higher)
See the guide here for various ways to install Orchestrator.
- SQL Server 2014 (or
higher)
It is highly recommended that you use the same SQL Server as was used when installing Orchestrator as detailed here. For the installation, you will require the hostname, admin username, and password of this SQL Server.
Important: Make sure that SQL Server Authentication mode is enabled.Note: AI Fabric uses SQL solely for metadata storage. This means that the amount of data store is very small. There is no need to provision a lot of storage capacity for these tables. - GPU Prequirements
For AI Fabric installation, it is a pre-requirement that the node have NVIDIA driver version 450.51.06 installed, as well as nvidia-container-runtime.