-
Download PyTorch: Official PyTorch Website
or using command line:
conda install pytorch torchvision -c pytorch # MacOS Binaries dont support CUDA, install from source if CUDA is needed
-
With pip:
pip install pytorch-pretrained-bert
pip3 install --upgrade gensim
Ten answers choices are given. Select the best answer to complete the sentence.
-
Example
Create a text file and save in folder as follows:
1 Longman , the third , looked at the others over his shoulder . 2 Goldband , the fourth , had a gold sash round his waist ; and little Playman did nothing at all , and was the more proud . 3 There was too much ostentation , and so I came away . ' 4 ` And now we are sitting and shining here ! ' 5 said the bit of bottle-glass . 6 At that moment more water came into the gutter ; it streamed over the edges and washed the bit of bottle-glass away . 7 ` Ah ! 8 now he has been promoted ! ' 9 said the Darning-needle . ' 10 I remain here ; I am too fine . 11 But that is my pride , which is a sign of respectability ! ' 12 And she sat there very proudly , thinking lofty thoughts . ' 13 I really believe I must have been born a sunbeam , I am so fine ! 14 It seems to me as if the sunbeams were always looking under the water for me . 15 Ah , I am so fine that my own mother can not find me ! 16 If I had my old eye which broke off , I believe I could weep ; but I ca n't -- it is not fine to weep ! ' 17 One day two street-urchins were playing and wading in the gutter , picking up old nails , pennies , and such things . 18 It was rather dirty work , but it was a great delight to them . 19 ` Oh , oh ! ' 20 cried out one , as he pricked himself with the Darning-needle ; ` he is a fine fellow though ! ' ' 21 I XXXXX not a fellow ; I am a young lady ! ' am am|born|find|picking|pricked|said|sat|seems|streamed|thinking
- The XXXXX in the 21th sentence is the target we want to predict.
- Also, the example in 21th sentence 'am' is the answer in this question. You could change it whether the token you want.
- The end of the 21th sentence are given 10 probable selections.
-
Results
- BERT model: https://github.com/huggingface/pytorch-pretrained-BERT
- gensim data: https://github.com/RaRe-Technologies/gensim-data