Task lost on "page refresh"

, ,
1

Hi,

What I have observed is that the task gets lost once the page is refreshed.
How to avoid this?

2

Hi @akshayklr057,

Have you tried setting "force_stream_order" in your recipe? When you set that value to true, the app will repeat the same question until you answer it, even if you "page refresh".

3

Where to set it and how to use it? I don't see any reference to it in documentation.

4

I have set the "force_stream_order":true in prodigy.json, but it is still not working.

5

@akshayklr057 that's strange, can you tell me more about your setup? I have a few questions to help us figure out what's wrong.

What version of prodigy are you using? "force_stream_order" was introduced recently in v1.9, so if you're using an old version it might not work.

Can you take a screenshot or video of the problem happening? Your description isn't quite enough to help figure out what exactly is happening.

Can you start prodigy with verbose logging and show me the logs from your console when it doesn't work? You can do it like:

PRODIGY_LOGGING=verbose prodigy your_args_here

You can set the value in prodigy.json file in the current working directory that you're running. Here's a minimal config that uses the forced stream order option:

{
  "theme": "eighties",
  "force_stream_order": true
}

I applied the eighties theme so it's easy for you to tell that the new configuration has been applied correctly. It's important to know that if you have a prodigy.json file in your user's home directory, it will take precedence over one that's in the current working directory.

6

My version is 1.8, I guess that't why setting force_stream_order doesn't work.
Thank you for the quick responses.

1 Like
7

Is there anyway to fix this issue for version 1.8

8

The force_stream_order setting was only introduced in Prodigy v1.9. If you're on v1.8 and not able to upgrade, you could implement something similar yourself with a custom stream generator that keeps looping over the incoming examples, checks them against the already annotated hashes and keeps sending out examples until they're annotated. See this post for example code: