Adds reranker example

[rodrigonogueira4]

No description provided.

[ronakice]

Can we give a slightly more meaningful example to show that it indeed works? And should we move this to docs/

[rodrigonogueira4]

Regarding exa

Can we give a slightly more meaningful example to show that it indeed works? And should we move this to docs/

Regarding a meaningful example, sure!
Regarding moving to docs/, I think this code needs to be in the main README.md as they can find it more easily.

[ronakice]

LGTM!

[Fatima-200159617]

I am trying to use monobert instead of T5. Here is the code

import torch
from transformers import PreTrainedModel,PreTrainedTokenizer,AutoTokenizer,AutoModel
from pygaggle.rerank.base import Query, Text
from pygaggle.rerank.transformer import SequenceClassificationTransformerReranker
model_name = 'castorini/monobert-large-msmarco'
tokenizer_name = 'castorini/monobert-large-msmarco'
batch_size = 8
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = AutoModel.from_pretrained("castorini/monobert-large-msmarco")
model = model.to(device).eval()
tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
reranker = SequenceClassificationTransformerReranker(model, tokenizer)
query = Query('how old are you?')
doc1 = Text('I am 77 years old')
doc2 = Text('I am hungry')
documents = [doc1,doc2]
scores = [result.score for result in reranker.rerank(query, documents)]
print(scores)
Could you please advice on what is the right tokenizer to use as I am having an issue running it.

[rodrigonogueira4]

Hi Fatima,

Could you try tokenizer_name = 'bert-large-uncased'?

[Fatima-200159617]

I tried I am getting the same error shown below:
ValueError Traceback (most recent call last)

in ()
15 doc2 = Text('I am hungry')
16 documents = [doc1,doc2]
---> 17 scores = [result.score for result in reranker.rerank(query, documents)]
18 print(scores)
19 print(sorted(scores,reverse=True))

1 frames

/content/gdrive/My Drive/Reranking-pygaggle/pygaggle/pygaggle/rerank/transformer.py in rerank(self, query, texts)
117 input_ids = ret['input_ids'].to(self.device)
118 tt_ids = ret['token_type_ids'].to(self.device)
--> 119 output, = self.model(input_ids, token_type_ids=tt_ids)
120 if output.size(1) > 1:
121 text.score = torch.nn.functional.log_softmax(

ValueError: too many values to unpack (expected 1)

Not sure if the tokenizer is the issue.

[Fatima-200159617]

The below code worked for me
#monoBert reranker
import torch
from transformers import BertTokenizer,BertForSequenceClassification
from pygaggle.rerank.base import Query, Text
from pygaggle.rerank.transformer import SequenceClassificationTransformerReranker
model_name = 'castorini/monobert-large-msmarco'
tokenizer_name = 'bert-large-uncased'
batch_size = 8
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = BertForSequenceClassification.from_pretrained("castorini/monobert-large-msmarco")
model = model.to(device).eval()
tokenizer = BertTokenizer.from_pretrained(tokenizer_name)
reranker = SequenceClassificationTransformerReranker(model, tokenizer)
query = Query('how old are you?')
doc1 = Text('I am 77 years old')
doc2 = Text('I am hungry')
doc3=Text('My age is 77')
doc4=Text('I want to sleep early')
documents = [doc1,doc2,doc3,doc4]
scores = [result.score for result in reranker.rerank(query, documents)]
print(scores)

[rodrigonogueira4]

Great, thanks, Fatima!

I've created a pull request that exemplifies how to use the BERT reranker: #59

[Fatima-200159617]

Great, thanks, Fatima!

I've created a pull request that exemplifies how to use the BERT reranker: #59

Thanks a lot for the code.