Hi,
I notice a stark difference between the evaluations created on the 9th and 10th of Apr in terms of the UI in the console. While similar test datasets with human labels are uploaded to these two evals and they both ran successfully, the earlier one has an export button and a filter option. This is gone for the latter. Is there a recent update that causes the change? How do I download the results now?
Besides, even though the export button is available for the earlier eval, there is a server related error.
Appreciate I can get some help quickly.
the problem is a syntax error as u can clearly see u have the same thing more than twice as different results in pic 3
Hello, there was a change, we swapped the frontend to use the new Evals API.
The Pass/Fail filter is not there currently, and there isn’t an export.
To unblock, you can use the API to download your data items (https://platform.openai.com/docs/api-reference/evals/getRunOutputItems)
Sorry about that! We’ll work on bringing the button back. Let me know if you can’t get the API to work for this.
4 Likes
I used the API you suggested to download the results. However it appears that the ‘limit’ parameter has a max value of 100. I have about 800 samples for the eval run. What’s the easiest way to download them all using the API?
