读取pandas数据训练loss为nan

[nxgych]

------
# DMatrix transition
xdm_train = xl.DMatrix(xtr, ytr)
xdm_test = xl.DMatrix(xte, yte)

# Training task
fm_model = xl.create_fm()  # Use factorization machine
fm_model.setOnDisk()
fm_model.setNoBin()
fm_model.setTrain(xdm_train)    # Training data
fm_model.setValidate(xdm_test)  # Validation data
------

[ ACTION ] Start to train ...
[------------] Epoch Train log_loss Test log_loss Test AUC Time cost (sec)
[ 5% ] 1 -nan -nan 0.500000 2.51
[ 10% ] 2 -nan -nan 0.500000 2.90
[ 15% ] 3 -nan -nan 0.500000 2.50
[ 20% ] 4 -nan -nan 0.500000 2.89

[kyle-pjk]

一样遇到这个问题。。