metadata

pipeline_tag: text-generation
tags:
  - model_hub_mixin
  - pytorch_model_hub_mixin

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

Library: https://huggingface.co/Aananda-giri/GPT2-Nepali/
Docs: [More Information Needed]

GPT-2-Nepali-512 Model

512 represents context length
This repository contains a custom GPT-2 model trained on Nepali text. Follow the instructions below to use this model for text generation.

How to Use the Model

Download the Required Code
Save the model_code.py file in the same directory where you'll run the script.
Install Required Libraries
Ensure you have the necessary libraries installed:
```
pip install transformers torch
```

Run the Following Code
Here's an example to load the model and generate text:

import torch
from model_code import GPTModel, generate_and_print_sample
from transformers import PreTrainedTokenizerFast

# Load the tokenizer
tokenizer = PreTrainedTokenizerFast.from_pretrained("Aananda-giri/NepaliBPE")

# Define the starting text
start_context = "रामले भात"

# Load the pre-trained model
loaded_model = GPTModel.from_pretrained("Aananda-giri/GPT2-Nepali")

# Move the model to the appropriate device (CPU or GPU)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
loaded_model.to(device)

# Generate text
generate_and_print_sample(
    loaded_model, tokenizer, device, start_context
)

Additional Notes

Tokenizer: The model uses a pre-trained tokenizer available at Aananda-giri/NepaliBPE. Ensure this is downloaded and accessible during runtime.
Dependencies: This code requires transformers (by Hugging Face) and torch (PyTorch). Install them if not already installed.
Device Compatibility: The script automatically detects if a CUDA-enabled GPU is available and utilizes it for faster inference. If not, it defaults to the CPU.

Example Output

Input:

रामले भात

Generated Text:

रामले भात खाएर सन्तोष माने। ऊ आफ्ना साथीहरूसँग रमाइलो गरिरहेको थियो।

Let me know if you'd like further assistance!