TypeError: __init__() got an unexpected keyword argument 'batch_size_per_gpu'
If I comment out batch_size_per_gpu, then it will get a TypeError on max_seq_length. If I comment both out, and the __innit__ presumably uses the default values, then the BertDataBunch will construct fine. However, then building the BertLearner object fails, with it failing to recognize model_type as an attribute of a BertDataBunch.
I'm at a loss as to what could cause this- I installed fast-bert by cloning the repo and using pip, though I've done so into a conda environment (never ideal, but fast-bert is the final package installed in building the environment, and the only package I've used pip with, so that should limit potential problems). Any help would be much appreciated.
I'm trying to construct an instance of a BertDataBunch, but the constructor doesn't seem to recognize two of the standard arguments.
Which yields-
TypeError: __init__() got an unexpected keyword argument 'batch_size_per_gpu'
If I comment out batch_size_per_gpu, then it will get a TypeError on max_seq_length. If I comment both out, and the __innit__ presumably uses the default values, then the BertDataBunch will construct fine. However, then building the BertLearner object fails, with it failing to recognize model_type as an attribute of a BertDataBunch.
I'm at a loss as to what could cause this- I installed fast-bert by cloning the repo and using pip, though I've done so into a conda environment (never ideal, but fast-bert is the final package installed in building the environment, and the only package I've used pip with, so that should limit potential problems). Any help would be much appreciated.