Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

use server embedder (inception v3 etc) models offline #173

Closed
NUSTemple opened this issue Dec 23, 2019 · 5 comments
Closed

use server embedder (inception v3 etc) models offline #173

NUSTemple opened this issue Dec 23, 2019 · 5 comments

Comments

@NUSTemple
Copy link

Image Analytics version

0.4.1

Orange version

3.23.1

Expected behavior

Server models like inception v3, VGG-19 can be used offline

Actual behavior

My computer is configured with proxy and cannot access api.garaza.io:443. Hence I cannot dowload required models.

Steps to reproduce the behavior
Additional info (worksheets, data, screenshots, ...)

image

@NUSTemple NUSTemple changed the title offline use inception v3 etc models use server embedder (inception v3 etc) models offline Dec 23, 2019
@ajdapretnar
Copy link
Collaborator

I must admit, I don't understand the issue fully.
First, server model's can't be used offline. They're server-based, it's in their nature.
Second, you must set your proxy settings correctly. You can do that in Options - Settings - Networks.
Alternatively, please better explain the issue.

@PrimozGodec
Copy link
Collaborator

Other embedders cannot be local since they are too slow. At the server, they run in parallel. Since we wanted to provide an option for users for not sending images to the server there is SqueezeNet implementation that runs locally (and performs well).

We are also aware that there is a problem with accessing our server through the proxy (even proxy server is correctly set up with Orange). It happens due to buggy implementation in a Hyper library which is used to communicate with the server. Due to this and a few other issues with Hyper (it is not maintained anymore) we are now implementing a client with other library HTTPX (#162) we hope that we can solve the proxy problem with the new library.

@NUSTemple
Copy link
Author

Thanks both for the reply. Appreciate that help. I am now using orange behind the proxy and currently api.garaza.io is blocked. Instead, we have a standalone GPU cluster which can be used to train the model itself. Thinking whether we can make this url as a configuration input and we can put our own server name there. Thanks.

@PrimozGodec
Copy link
Collaborator

It is actually implemented already here: #114 It is not merged since we didn't have an opportunity to test it properly. If you want to use it pull this pull request. The server needs to be configured such that it responds to the request (image) with a vector of embeddings.

@PrimozGodec
Copy link
Collaborator

We will merge this PR with the custom server settings in master when we finish the transition to httpx.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants