Downloads · 30 days
0
gpt2-355M-hinglish-translation-instruct
SAURABH257/gpt2-355M-hinglish-translation-instruct
gpt2-355M-hinglish-translation-instruct is a translation model from SAURABH257. Use it when you need text moved from one language to another. The card lists the license as mit.
A fine-tuned GPT-2 Medium model for translating English text into Hinglish (Hindi written using the Latin alphabet).
Access
Public
Updated Jun 22, 2026
Repo size
1.7 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pth1.7 GB · 100%
At a glance
- Task
- Translation
- License
- mit
- Access
- Public
- Created
- Jun 22, 2026
- Updated
- Jun 22, 2026
- SHA
- 4e834fd1
Try a prompt
Base models
Datasets
Model card
From the Hugging Face model README
GPT2-Medium English → Hinglish Translator
A fine-tuned GPT-2 Medium model for translating English text into Hinglish (Hindi written using the Latin alphabet).
Model Overview
This model has been fine-tuned on the findnitai/english-to-hinglish dataset to generate natural Hinglish translations from English input. The model is designed for conversational text, social media content, chatbots, and localization use cases targeting Indian audiences.
Example
Input
How are you doing today?
Output
Aaj tum kaise ho?
Model Details
| Attribute | Value |
|---|---|
| Base Model | GPT-2 Medium |
| Task | English → Hinglish Translation |
| Dataset | findnitai/english-to-hinglish |
| Languages | English, Hinglish |
| License | MIT |
| Framework | Hugging Face Transformers |
| Developed By | Saurabh Prajapati |
Training Dataset
Dataset used:
findnitai/english-to-hinglish
The dataset contains paired English and Hinglish sentences suitable for supervised fine-tuning.
Example:
{
"en": "What are you doing?",
"hi_ng": "Tum kya kar rahe ho?"
}
Intended Use
Direct Use
- English to Hinglish translation
- Conversational AI
- Customer support chatbots
- Social media content generation
- WhatsApp and SMS automation
Downstream Applications
- Multilingual assistants
- Marketing content localization
- Hinglish content generation
- Voice assistant backends
Out-of-Scope Use
This model is not intended for:
- Legal translation
- Medical translation
- Financial advice
- Safety-critical applications
- Professional human translation replacement
Generated outputs should always be reviewed before production deployment.
Quick Start
Installation
pip install transformers torch
Training Procedure
Input Format
Training examples were formatted as:
Translate English to Hinglish:
English: How are you?
Hinglish: Tum kaise ho?
Training Configuration
| Parameter | Value |
|---|---|
| Base Model | GPT-2 Medium |
| Parameters | 355M |
| Memory | 1.7Gb |
| Context_size | 1024 |
| Emb_dim | 1024 |
| n_layers | 24 |
| n_heads | 16 |
| Vocab_size | 50257 |
| Optimizer | AdamW |
| Learning Rate | 5e-5 |
| Precision | FP16 |
| Framework | Google Colab |
| Task | Causal Language Modeling |
Evaluation
Metric
- Accuracy
- 78% Accuracy on the Test Dataset (100 Samples)
Sample Results
| English | Hinglish |
|---|---|
| How many accidents are on the i 5 north freeway? | i 5 north freeway par kitne accidents hue he? |
| Report the weather in New Zealand to me | mere liye New Zealand ka weather report bataiye |
| Remind me every month to pay my mortgage payment | mujhe har mahine har mahine mera mortgage payment karne ke liye yaad dilaye |
The model performs well on common conversational English sentences and generates natural Hinglish outputs.
Limitations
- Hinglish spelling is not standardized.
- Performance may degrade on long paragraphs.
- Technical and domain-specific terminology may not translate well.
- Outputs may occasionally mix English and Hindi terms.
Bias and Risks
This model inherits biases present in:
- GPT-2 pretraining data.
- The English-Hinglish fine-tuning dataset.
Technical Specifications
Architecture
- GPT-2 Medium
- Decoder-only Transformer
- ~355 Million Parameters
Software Stack
- Python
- PyTorch
- Hugging Face Transformers
- Hugging Face Datasets
Acknowledgements
- OpenAI GPT-2
- Hugging Face Transformers
- Hugging Face Datasets
- findnitai/english-to-hinglish dataset
Citation
@misc{prajapati2026gpt2hinglish,
author = {Saurabh Prajapati},
title = {GPT2-Medium English to Hinglish Translator},
year = {2026},
publisher = {Hugging Face}
}
Contact
For questions, issues, or collaboration opportunities, please open an issue on the Hugging Face repository.
- Task
- Translation
- License
- mit
- Languages
- en, hi
- Created
- Jun 22, 2026
- Updated
- Jun 22, 2026