File size: 982 Bytes
5ccb6e2 54791f4 5ccb6e2 c564920 5ccb6e2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 |
---
license: afl-3.0
---
# Question generation using T5 transformer trained on SQuAD
<h2> <i>Input format: context: "..." answer: "..." </i></h2>
Import the pretrained model as well as tokenizer:
```
from transformers import T5ForConditionalGeneration, T5Tokenizer
model = T5ForConditionalGeneration.from_pretrained('AbhilashDatta/T5_qgen-squad_v1')
tokenizer = T5Tokenizer.from_pretrained('AbhilashDatta/T5_qgen-squad_v1')
```
Then use the tokenizer to encode/decode and model to generate:
```
input = "context: My name is Abhilash Datta. answer: Abhilash"
batch = tokenizer(input, padding='longest', max_length=512, return_tensors='pt')
inputs_batch = batch['input_ids'][0]
inputs_batch = torch.unsqueeze(inputs_batch, 0)
ques_id = model.generate(inputs_batch, max_length=100, early_stopping=True)
ques_batch = [tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in ques_id]
print(ques_batch)
```
Output:
```
['what is my name']
``` |