Author: Ashish Kumar, Head of Platforms, AI & Data at Tata Group · ashishkumar.ai
from transformers import AutoTokenizer, AutoModelWithLMHead, AutoModelForCausalLM
import torch
if torch.cuda.is_available():
device = torch.device("cuda")
else :
device = "cpu"
tokenizer = AutoTokenizer.from_pretrained("Ashishkr/grammar_correction")
model = AutoModelForCausalLM.from_pretrained("Ashishkr/grammar_correction").to(device)
input_query="what be the reason for everyone leave the company"
query= "<|startoftext|> " + input_query + " ~~~"
input_ids = tokenizer.encode(query.lower(), return_tensors='pt').to(device)
sample_outputs = model.generate(input_ids,
do_sample=True,
num_beams=1,
max_length=128,
temperature=0.9,
top_p= 0.7,
top_k = 5,
num_return_sequences=3)
corrected_sentences = []
for i in range(len(sample_outputs)):
r = tokenizer.decode(sample_outputs[i], skip_special_tokens=True).split('||')[0]
r = r.split('~~~')[1]
if r not in corrected_sentences:
corrected_sentences.append(r)
print(corrected_sentences)