Can you clarify if you mean TextPrediction or TextClassification? TextClassification is a deprecated task and TextPrediction should be used instead (as of 0.0.13).

I mean TextPrediction.

@sxjscience Could you answer?

@redhood95 There are two scenarios for resuming a trained model: 1) After you have accidentally killed the job, we may resume the training. 2) We load a trained model and continue the training process with a new sets of hparams, or on a new dataset. Which one is your use case? For now, we cannot resume training but I can add the functionality.

i need to resume training on previous weights for the same dataset, please let me know if its possible. and the two scenarios @sxjscience mentioned how can i resume training when i killed it accidentally.

hey, if it's okay can I work on adding this feature, for all tasks that are present currently?

@redhood95 We are more than welcome for any contributions to AutoGluon. Adding the second functionality (load a trained model and call .fit() again) in TextPrediction is not very difficult. We will need to extend the loaded predictor

@sxjscience where can I join the slack channel?

@redhood95 I removed the previous slack invitation. For those who are willing to obtain the slack invitation, just send emails to me or other developers in AutoGluon.