| # Fill-Mask PyTorch Model (Camembert) |
|
|
| This model is a `fill-mask` model that was trained using the PyTorch framework and the Hugging Face Transformers library. It was utilized in Hugging Face's NLP course as an introductory model. |
|
|
| ## Model Description |
|
|
| This model uses the `camembert` architecture, a variant of the RoBERTa model adapted for French. It's designed for the fill-mask task, where a portion of input text is masked and the model predicts the missing token. |
|
|
| ## Features |
|
|
| - **PyTorch**: The model was implemented and trained using the PyTorch deep learning framework, which allows for dynamic computation graphs and is known for its flexibility and efficiency. |
| - **Safetensors**: The model utilizes Safetensors, a Python library that provides safer operations for PyTorch Tensors. |
| - **Transformers**: The model was built using the Hugging Face Transformers library, a state-of-the-art NLP library that provides thousands of pre-trained models and easy-to-use implementations of transformer architectures. |
| - **AutoTrain Compatible**: This model is compatible with Hugging Face's AutoTrain, a tool that automates the training of transformer models. |
|
|
| ## Usage |
|
|
| ```python |
| from transformers import CamembertForMaskedLM, CamembertTokenizer |
| |
| tokenizer = CamembertTokenizer.from_pretrained('model-name') |
| model = CamembertForMaskedLM.from_pretrained('model-name') |
| |
| inputs = tokenizer("Le camembert est <mask>.", return_tensors='pt') |
| outputs = model(**inputs) |
| predictions = outputs.logits |
| predicted_index = torch.argmax(predictions[0, mask_position]).item() |
| predicted_token = tokenizer.convert_ids_to_tokens([predicted_index])[0] |
| |
