ML Packages Offline Installation

Getting started

Depending on the models you want to use, you need the following:

• For models 2022.10 and newer:
  • Download the needed Document Understanding bundle. Here are the links for all the available bundles. The du bundle contains information about all models included into a specific version. For example, the dusemistructured-2023.4.0.tar.gz contains information about all out-of-the-box pre-trained ML Packages included in the 2023.4.0 version.
• For models 2022.4 and older (python37duv3 and python37duv4):
  • All ML Packages are provided as a .zip file which is uploaded directly as a Custom Package in AI Center. To download the models, contact your Account Manager, CSM, or Support to receive a download link per package.
  • Download the needed Document Understanding bundle. Here are the links for all the available bundles.

Install the Offline Bundle

Offline installations are requiring that the downloaded du bundle to be renamed in the command line into du-ondemand.tar.gz. For instance, if you downloaded the du bundle named dusemistructured-2023.4.0.tar.gz, at installation time you need to rename it as du-ondemand.tar.gz.

1. For Windows machines, directly download through the bundle link and rename the file to du-ondemand.tar.gz

2. For Linux machines, from the machine having access to the internet, download the needed bundle following the below command:

   wget -O ~/<bundle-name.tar.gz> 'bundle-link'

   Here's an example of how to download the needed bundle for Linux:

   wget -O ~/du-ondemand.tar.gz 'https://download.uipath.com/automation-suite/2023.4.0/dusemistructured-2023.4.0.tar.gz'

3. Copy the following bundle to the /uipath/tmp folder on the main machine of the cluster (where the install took place):

   scp ~/<bundle-name.tar.gz> <username>@<node dns>:/uipath/tmp/

4. Connect to this main machine and load the bundle:

   ./configureUiPathAS.sh registry upload --optional-offline-bundle "/uipath/tmp/du.tar.gz" --offline-tmp-folder "/uipath/tmp"

Upload the Model to AI Center

After downloading and installing the models, follow the steps described here to upload them to AI Center.

Upload a DU Bundle to an External Docker Registry

Follow these steps to upload a DU bundle to your external docker registry:

• Pull the desired DU image from the UiPath registry hosted on registry.uipath.com

• Rename the image host as per your docker registry name.

• Push the image to external docker registry

Pull the desired DU images

Pull the images from UiPath registry by running these commands:

docker pull <uipath_registry_server>/<image_name>

Here's an example of how to pull the image of the UiPathDocumentOCR bundle from a registry called registry.uipath.com:

docker pull registry.uipath.com/aicenter/du-doc-ocr:v23.4.0-rc70

Rename the image host

Rename the image host by running the following command:

docker tag <uipath_registry_server>/<image_name> <your_registry_server>/<image_name>

Here's an example of how to rename the image of the UiPathDocumentOCR bundle from a registry called registry.uipath.com to a registry called registory.mycompany.com:

docker tag registry.uipath.com/aicenter/du-doc-ocr:v23.4.0-rc70 registory.mycompany.com/aicenter/du-doc-ocr:v23.4.0-rc70

Push the image to external docker registry

Push the image to your external docker registry by running the following command:

docker push <your_registry_server>/<image_name>

Here's an example of how to push the image of the UiPathDocumentOCR bundle to an external docker registry:

docker push registory.mycompany.com/aicenter/du-doc-ocr:v23.4.0-rc70

Images for each DU bundle

DU Bundle	Image
UiPathDocumentOCR	aicenter/du-doc-ocr:v23.4.0-rc70
UiPathDocumentOCR_CPU	aicenter/du-doc-ocr-cpu:v23.4.0-rc70
OCR for Chinese, Japanese, Korean	du/du-cjk-ocr-proxy:v23.4.0-rc70
DocumentClassifier	aicenter/du-ml-document-type-text-classifier:v23.4.0-rc70
Out-of-the-box Pre-trained ML Packages	aicenter/du-semistructured:v23.4.0-rc70
Dulv4	aicenter/python37duv4:23.4.0
Dulv3	aicenter/python37duv3:23.4.0
Dulv2	aicenter/python37duv2:23.4.0