Testing Data Problem

Question

Testing Data Problem

Nhuc Dang - Feeney Brothers 1

Any ideas what this could be? Trying to upload a test data to test the model.

YutongTie-MSFT 53,966 Reputation points Moderator

2022-03-17T20:51:01.683+00:00

Hello @Nhuc Dang - Feeney Brothers

Thanks for reaching out to us for this issue. What does your data set look like? I have seen similar issue for pd.dataframe for your reference, please share more details so that we can look deeper.

https://stackoverflow.com/questions/46771235/valueerror-dataframe-constructor-not-properly-called-with-pandas

Regards,
Yutong
Nhuc Dang - Feeney Brothers 1 Reputation point

2022-03-17T21:02:05.863+00:00

Hi @YutongTie-MSFT ,

Dataset includes 787 columns x 2000 rows, consisting of image/decimal data. Column 787 is the label with either 1 or 0, the rest are all features consisting of column 785-786 are int numbers and Column from 1 to 784 are decimal.

The training is fine but when I upload a test set which consist of 786 (without the label column), the Azure ML throws that error.

Thanks
YutongTie-MSFT 53,966 Reputation points Moderator

2022-03-17T21:41:44.35+00:00

@Nhuc Dang - Feeney Brothers

Thank you for your information, since it's very difficult to debug this error with the data, I would like to recommend you to raise a support ticket for this internal error. Please let me know if you have no support plan, I would like to enable you a free one to assign a support engineer to look deeper.

If you have a support plan, please share the ticket number once you create it, I will track and help also.

Regards,
Yutong
Nhuc Dang - Feeney Brothers 1 Reputation point

2022-03-17T21:45:57.223+00:00

Hi @YutongTie-MSFT
Yes, I have a ticket reference number - 2203170040006725, thanks!
YutongTie-MSFT 53,966 Reputation points Moderator

2022-03-17T22:03:03.85+00:00

@Nhuc Dang - Feeney Brothers

Thank you for the information! I will try my best to see what I can do.

Regards,
Yutong

1 answer

Your answer

YutongTie-MSFT 53,966 Reputation points Moderator

2022-03-17T20:51:01.683+00:00

Hello @Nhuc Dang - Feeney Brothers

Thanks for reaching out to us for this issue. What does your data set look like? I have seen similar issue for pd.dataframe for your reference, please share more details so that we can look deeper.

https://stackoverflow.com/questions/46771235/valueerror-dataframe-constructor-not-properly-called-with-pandas

Regards,
Yutong
Nhuc Dang - Feeney Brothers 1 Reputation point

2022-03-17T21:02:05.863+00:00

Hi @YutongTie-MSFT ,

Dataset includes 787 columns x 2000 rows, consisting of image/decimal data. Column 787 is the label with either 1 or 0, the rest are all features consisting of column 785-786 are int numbers and Column from 1 to 784 are decimal.

The training is fine but when I upload a test set which consist of 786 (without the label column), the Azure ML throws that error.

Thanks
YutongTie-MSFT 53,966 Reputation points Moderator

2022-03-17T21:41:44.35+00:00

@Nhuc Dang - Feeney Brothers

Thank you for your information, since it's very difficult to debug this error with the data, I would like to recommend you to raise a support ticket for this internal error. Please let me know if you have no support plan, I would like to enable you a free one to assign a support engineer to look deeper.

If you have a support plan, please share the ticket number once you create it, I will track and help also.

Regards,
Yutong
Nhuc Dang - Feeney Brothers 1 Reputation point

2022-03-17T21:45:57.223+00:00

Hi @YutongTie-MSFT
Yes, I have a ticket reference number - 2203170040006725, thanks!
YutongTie-MSFT 53,966 Reputation points Moderator

2022-03-17T22:03:03.85+00:00

@Nhuc Dang - Feeney Brothers

Thank you for the information! I will try my best to see what I can do.

Regards,
Yutong

Answer 1

YutongTie-MSFT 53,966 Moderator

Update about this case:

During the support session, support team confirmed that PG has identified a bug where the target column in test data, which ideally should be optional, is currently required. PG will roll out the fix in the next python SDK release.

Sorry for the experience.

Regards,
Yutong

-Please kindly accept the answer if you feel helpful, thanks a lot.

Share via

Testing Data Problem

1 answer

Your answer