aijadugar/gujarati-nanogpt

Gujarati NanoGPT

Gujarati character-level NanoGPT model trained on a Gujarati text corpus.

Model Details

• Vocabulary Size: 5581
• Layers: 4
• Heads: 4
• Embedding Size: 96
• Block Size: 128

Dataset

Custom Gujarati Multimodal AI Dataset.

Usage

import torch
import pickle

with open("meta.pkl", "rb") as f:
    meta = pickle.load(f)

stoi = meta["stoi"]
itos = meta["itos"]

model = torch.load("pytorch_model.bin")

Author

Bari Ankit Vinod