-
Notifications
You must be signed in to change notification settings - Fork 235
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
Issue with xgboost export in python: not same values splits #537
Comments
I have a similar issue with the export to C.
returns In cases if the values are around 0.5, the classification result differs using the probability threshold of 0.5. (64 errors in 2000 samples). Any ideas how I could debug that? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi,
Thanks for the work on this package. I'm using m2cgen to convert a XGBoost model into VBA code. But when using the code produce by m2cgen I've got some predictions that are really different from the one get after training my model in Python. Here are some examples:
Capture
And I've looked in the XGBoost booster after training and compare to the output (in Python) form m2cgen. Here is what I have from m2cgen
And this is what I have in the booster:
For me, but I'm not expert, it seems that in the function returned by m2cgen the floats are 32 bits floats in the if/else conditions and not in the booster. So if in my data one sample has the value of the split m2cgen is not giving back the right value. Is there a trick to force floats to 64 bits?
Thanks in advance for your return,
Antoine
The text was updated successfully, but these errors were encountered: