On-premise version with self-hosted ML models

Help AI Computer Vision
, ,
1

Hi,

Any idea if you know that they ai cloud solution can be hosted in-house for enterprises?
Sending out certain screens and info towards a uipath server is Often not allowed.

Kind regards,

Tom

2

Hey Tom, yes a self-hosted (on-prem) version is available.
It requires a machine with a decent GPU

4

Hi Cosin,

Are there additional license costs for this on-prem solution?

Kind regards,

Tom

1 Like
5

Hi Cosin,

what about the training data sets for the Deep Learning algorithms? Will the on-premise be useable without them?

Regards
Matthias

1 Like
6

For the moment there’s no additional cost, although we highly recommend you purchase a Google cloud license (1.5$ /1000 OCRs), it’s much better than the default one.

1 Like
7

The whole feature, including the on-prem version is delivered fully functional, no additional training required

1 Like
8

How will CV work if client has no internet access on server or virtual machine? is it possible to train it locally?

1 Like
9

There is an on-prem version available on request. Please contact me if you have such a usecase

5 Likes
10

I was just curious… thanks for clearing it out. :grinning:

11

@Cosin is this available with the stable version release or its just for testing purpose?

12

@megharajky - Sharing @Cosin’s reply from above. Let us know if this doesn’t answer your question. Thanks!

1 Like
13

ha, thx, phone replies…

14

This is really exciting! As many customers will have security issues if potential sensitive data might be sent out to the cloud!
@Cosin: Could you share the actual hardware requirements for teh on prem deployment?

16

Assuming that you’ll continiously improve the recogntion model is there a plan on how the recogntion model will be updated on on-premise installations?

17

The plan is to update it like most of our products, by getting the latest version and upgrading (in this case using an upgrade script).

18

In terms of hardware compatibility, we haven’t tested various models extensively, so we recommend servers on Intel CPUs with NVIDIA V100 GPUs. The ML team is working on testing other configurations, albeit all with at least Pascal-series GPUs.

If you’re interested in deploying it on-prem you can contact us so we can discuss more

1 Like
19

That’s what I assumed. Thanks for the anwer!

20

Hi Everyone! If you’re interested in deploying the on-prem version, please email insider-preview@uipath.com for more information.

4 Likes
21

@Ryan_Rush Thanks! I’ve send a mail.