sachin18449/hindi-english-wordpiece-nmt

Model Card for English-to-Hindi Transformer Model

This model is a custom Transformer-based neural machine translation (NMT) model trained to translate English sentences to Hindi.
It follows the original Attention Is All You Need architecture and has been fine-tuned on a parallel English-Hindi dataset.

---

Model Details

Model Description

This is a sequence-to-sequence model with an Encoder-Decoder Transformer architecture.
It is trained for English → Hindi machine translation using PyTorch and Hugging Face 🤗 Transformers.

• Developed by: Sachin Kumar
• Funded by: Self / Personal Research
• Shared by: Sachin Kumar
• Model type: Transformer (Encoder-Decoder) for Neural Machine Translation
• Language(s): English (source), Hindi (target)
• License: Apache-2.0
• Finetuned from model: Trained from scratch using transformers (no pre-trained checkpoint)

Model Sources

• Repository: Hugging Face Model Link
• Paper (base architecture): Attention Is All You Need

---

Uses

Direct Use

The model can be used as-is for:

• Translating English text into Hindi.
• Building English→Hindi chatbots, translation services, or multilingual applications.

Downstream Use

It can be further fine-tuned for:

• Domain-specific translation (medical, legal, technical).
• Multi-task setups (e.g., summarization + translation).

Out-of-Scope Use

• Not designed for speech-to-text or Hindi-to-English translation.
• Avoid using it for sensitive or critical decision-making tasks (legal, medical, etc.) without human review.

---

Bias, Risks, and Limitations

• Translations may contain biases present in the training data (e.g., gender stereotypes).
• Hindi grammar and morphology are complex; rare words or idiomatic phrases may be mistranslated.
• Not guaranteed to handle code-mixed (Hindi-English) input well.

Recommendations

• Always review translations before production use.
• Consider post-editing for professional/official translations.

---

How to Get Started with the Model

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

model_name = "sachin18449/hindi-english-wordpiece-nmt"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

input_text = "How are you?"
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_length=50)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))