amirhsgh / huggingface-two-sentences-are-similar

Two-Sentences-Are-Similar

It's three different training loop for train your data on pretrained model.

if you want test one data on end of jupyter I show how can use it

first you should tokenize your sentences

sentences = tokneizer({
    "sentence1":"...",
    "sentence2":"..."
})

second you should change it to tensor for you can use it

sentences['input_ids'] = torch.tensor(sentences['input_ids'])
sentences['token_type_ids'] = torch.tensor(sentences['token_type_ids'])
sentences['attention_mask'] = torch.tensor(sentences['attention_mask'])

third you should move it to your device

sentences = {key: value.to('cuda:0') for key, value in sentences.items()}

final you can predict your inputs

with torch.no_grad():
    out = model(sentences['input_ids'].unsqueeze(0), sentences['attention_mask'].unsqueeze(0),
                  sentences['token_type_ids'].unsqueeze(0))
    print(f"the label of predict is : {torch.argmax(out.logits)}")