Sebastian Veile comments

Results 13 comments of


                                            Sebastian Veile

visible gpu setting and bert model download

This is only the case if you have more than 1 gpu. If you only use 1 gpu, try with gpus = 0. Then it will download the model and...

Debugging hint for Index Error encountered after supplying the network with Japanese data

Are you using a different bert model? I got an error like this when I used a smaller bert model to train on a dataset that was preprocessed using a...

Debugging hint for Index Error encountered after supplying the network with Japanese data

I believe your issue might be that you have preprocessed the data using the bert_multilingual, but you are trying to train on bert_base-uncased - It all depends on if you...

FileNotFoundError: [Errno 2] No such file or directory: '../bert_data/cnndm.test.0.bert.pt'

You should re-frame the bert_data path as follow "../bert_data/cnndm"

CUDA runtime error on Google Colab

Late reply. But here is the solution -visible_gpus = 0

Advice on prediction and replacing the bert model

Check this pull request for a Japanese model - https://github.com/nlpyang/PreSumm/pull/118

Pretrained Models

Model path is where you save each checkpoint. Currently you save a checkpoint at every 2000 step

https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-pytorch_model.bin

This is not an error. It is just informing you that it is downloading the model to "/tmp/tmpq67imxb7"

Training On other Dataset Doesnot Seems Satisfactory

What was your training command and how many gpus did you train on?

Training On other Dataset Doesnot Seems Satisfactory

The author describes here https://github.com/nlpyang/PreSumm/issues/44 that you need to adjust the accum_count when only training on one GPU