Instructions to use soyuj/deeper-impact with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use soyuj/deeper-impact with Transformers:
# Load model directly from transformers import AutoTokenizer, DeepImpact tokenizer = AutoTokenizer.from_pretrained("soyuj/deeper-impact") model = DeepImpact.from_pretrained("soyuj/deeper-impact") - Notebooks
- Google Colab
- Kaggle
Add ONNX export (onnx/model.onnx) and model-card section
#1
by AlJohri - opened
Adds a single-file ONNX export of the model under onnx/model.onnx (~436 MB) for inference with ONNX Runtime, plus an 'ONNX' section on the model card documenting the input/output contract and how a term's impact is recovered.
The file was produced with the export script proposed in the companion PR basnetsoyuj/DeeperImpact#17 (src/deep_impact/scripts/export_onnx.py) and verified to match the PyTorch model within max |diff| ~ 6e-6 over a padded batch of documents.