Closed benfred closed 1 year ago
The GPU ALS model would sometimes return incorrect results with the calculate_training_loss parameter enabled. This happend when the number_of_users * number_of_items was bigger than 2**31 due to an overflow in the loss function calculation.
calculate_training_loss
Fix and add tests that would have caught this bug
Closes https://github.com/benfred/implicit/issues/367 Closes https://github.com/benfred/implicit/issues/441
The GPU ALS model would sometimes return incorrect results with the
calculate_training_lossparameter enabled. This happend when the number_of_users * number_of_items was bigger than 2**31 due to an overflow in the loss function calculation.Fix and add tests that would have caught this bug
Closes https://github.com/benfred/implicit/issues/367 Closes https://github.com/benfred/implicit/issues/441