Jayveersinh-Raj's picture
Update README.md
fcf65fa verified
metadata
language:
  - hi
metrics:
  - bleu
  - rouge

Model discription

Hindi Summarization model. It summarizes a hindi paragraph.

Base model

  • mt5-small

How to use

from transformers import AutoTokenizer
from transformers import AutoModelForSeq2SeqLM, Seq2SeqTrainingArguments, Seq2SeqTrainer

checkpoint = "Jayveersinh-Raj/hindi-summarizer-small"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)

# Input paragraph for summarization
input_sentence = "<sum> your hindi paragraph"

# Tokenize the input sentence
input_ids = tokenizer.encode(input_sentence, return_tensors="pt").to("cuda")

# Generate predictions
with torch.no_grad():
   output_ids = model.generate(input_ids, max_new_tokens=200)

# Decode the generated output
output_sentence = tokenizer.decode(output_ids[0], skip_special_tokens=True)

# Print the generated output
print("Input:", input_sentence)
print("Summarized:", output_sentence)

Evaluation

  • Rogue1: 0.38
  • BLUE: 0.35