can the loss or some information during training be shown?

[823858275]

i use model.fit() to train the model, but it doesn't show the detail of training, such as loss. I don't find some parameter to show those information.can the loss or some information during training be shown?

[nreimers]

You can pass an evaluator to the fit method, as shown here:
https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/nli/training_nli.py

The evaluator is called every evaluation_steps and computes whatever scores you like. Usually I use it to compute scores (like accuracy, correlation) on some held-out dev set.

[iamlockelightning]

Hi, @nreimers . Can it just return the training loss values during fit? (On progress bar, or logging to file/std, or Tensorboard support, ...)

I don't have a held-out dev set. Just want to monitor the training process.

[Rj7]

@nreimers This would really useful for Unsupervised learning.