Problem in converting pytorch model( pytorch_model.bin) to model.pt

working on colab
Project (folder containing model)
├── pytorch_model.bin
├── bert_config.json
├── vocab.txt
├──model.ckpt-231879.meta
├──model.ckpt-231879.index
└── model.ckpt-231879.data-00000-of-00001

import torch
from pytorch_pretrained_bert import BertTokenizer, BertModel, BertForNextSentencePrediction, BertForQuestionAnswering
PRE_TRAINED_MODEL_NAME_OR_PATH = 'Project '
model = BertForQuestionAnswering.from_pretrained( PRE_TRAINED_MODEL_NAME_OR_PATH, cache_dir=None, from_tf=False, state_dict=None)
torch.save(model.state_dict(), 'model.pt')
model.load_state_dict(torch.load('model.pt'))
model.eval()

ERROR :- RuntimeError: Trying to create tensor with negative dimension -1: [-1, 768]

i think i have done something wrong here please help.