A question regarding the `Could not parse JSON: b...
# squadsm
Mariana Menchero
03/07/2025, 6:01 PMA question regarding the
Could not parse JSON: b'Internal Server Error'Mariana Menchero
03/07/2025, 6:02 PMI previously saw this error when a client ran cross-validation with refit=True and 250 windows—setting it to False resolved it.
However, the client from my first question gets the error occasionally, even when working with a single time series.
Has anyone encountered this in a similar situation?
y
Yibei
03/07/2025, 6:09 PMCould it be that’s a time out error?
m
Mariana Menchero
03/07/2025, 6:10 PMmaybe 🤔
Mariana Menchero
03/07/2025, 6:10 PMbut that's a good suggestion that worked with another client
y
Yibei
03/07/2025, 6:11 PMNikhil answered this question on intercom earlier: https://app.intercom.com/a/inbox/j7y9c2ep/inbox/shared/all/conversation/820?view=List
Maybe the client can also give us their full error and request_id
Yibei
03/07/2025, 6:12 PMcross-validation with 250 windows and refit=True is a lot. If they also use finetune, then the model is refitting in each window.
m
Mariana Menchero
03/07/2025, 6:13 PMYes, for sure, and setting it to False fixed the issue. The only thing is that this is a different client 😅
Mariana Menchero
03/07/2025, 6:14 PMI don't have access to intercom. Can you share Nikhil's response here?
Mariana Menchero
03/07/2025, 6:14 PMor maybe add me to intercom?
y
Yibei
03/07/2025, 6:19 PMOh sorry about that.
This is the question:
ApiError: status_code: 500, body: {‘detail’: ‘Internal server error, please contact us at ops@nixtla.io’, ‘request_id’: ‘QPQ83S3GSK’}
And this is the response:
It seems that this request timed out since it took more than 60 seconds. You can increase this in the client to a higher value if you want, but there is an upper limit to this. If your dataset is large, the better approach would be to use the num_partitions argument to split your data into smaller chunks. In this case, the client will split your data into the number of partitions you mention and make that many (smaller) calls under the hood.
m
Mariana Menchero
03/07/2025, 6:28 PMThank you
c
Cristian Challu
03/07/2025, 7:09 PMAlso important to mention the client is using the self hosted docker