Indic-Transformers Bengali DistilBERT

Model description

This is a DistilBERT language model pre-trained on ~6 GB of monolingual training corpus. The pre-training data was majorly taken from OSCAR . This model can be fine-tuned on various downstream tasks like text-classification, POS-tagging, question-answering, etc. Embeddings from this model can also be used for feature-based training.

Intended uses & limitations

from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained('neuralspace-reverie/indic-transformers-bn-distilbert')
model = AutoModel.from_pretrained('neuralspace-reverie/indic-transformers-bn-distilbert')
text = "আপনি কেমন আছেন?"
input_ids = tokenizer(text, return_tensors='pt')['input_ids']
out = model(input_ids)[0]
print(out.shape)
# out = [1, 5, 768] 

The original language model has been trained using PyTorch and hence the use of pytorch_model.bin weights file is recommended. The h5 file for Tensorflow has been generated manually by commands suggested here .