eng_to_hin_torch

MODEL CARD

This model has been pushed to the Hub using the PytorchModelHubMixin integration:

• The pytorch transformer model is for translating English to Hindi
• Model trained using Samanantar v3 dataset

Inferencing Setup:

• Clone Repo
• Create a py/ipynb script in repo's root folder
• Create venv based on requirements file
• Execute below code

import json
import torch

from scripts.transformer import Transformer
from scripts.inference_helper import translate


device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Read index to hindi vocab mapping
with open("./config/index_to_hindi.json", "r") as f:
    index_to_hindi = json.load(f)
    index_to_hindi = {int(k): v for k, v in index_to_hindi.items()}

# Load model
model = Transformer.from_pretrained("./eng_to_hin_model/")
model.eval()
model.to(device)

# Generate translation
eng_sentence = "hello how are you doing"

translation = translate(eng_sentence, model, index_to_hindi)
print(translation)