batch_size_per_gpu and max_seq_length are unexpected args?

[ktfhale]

I'm trying to construct an instance of a BertDataBunch, but the constructor doesn't seem to recognize two of the standard arguments.

from fast_bert.data import BertDataBunch
DATA_PATH = '/home/khale/FastBert/'
LABEL_PATH = '/home/khale/FastBert/'

databunch = BertDataBunch(DATA_PATH, LABEL_PATH,
                          tokenizer='bert-base-cased',
                          train_file='fastbert_train1.csv',
                          val_file='fastbert_val1.csv',
                          label_file='labels.csv',
                          text_col='text',
                          label_col='label',
                          batch_size_per_gpu=16,
                          max_seq_length=512,
                          multi_gpu=False,
                          multi_label=False,
                          model_type='bert')

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.

~/anaconda3/envs/riskenv/lib/python3.7/site-packages/fast_bert/learner_cls.py in from_pretrained_model(dataBunch, pretrained_path, output_dir, metrics, device, logger, finetuned_wgts_path, multi_gpu, is_fp16, loss_scale, warmup_steps, fp16_opt_level, grad_accumulation_steps, multi_label, max_grad_norm, adam_epsilon, logging_steps, freeze_transformer_layers)
    131         model_state_dict = None
    132 
--> 133         model_type = dataBunch.model_type
    134 
    135         if torch.cuda.is_available():

AttributeError: 'BertDataBunch' object has no attribute 'model_type'

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.

[ktfhale]

This can be closed/deleted- the problem was importing BertDataBunch from data, and not data_cls, as the docs actually specify... don't I feel silly.