Unable to download scibert-nli after latest changes

Covid-19 Semantic Browser: Browse Covid-19 & SARS-CoV-2 Scientific Papers with Transformers 🦠 📖

Covid-19 Semantic Browser is an interactive experimental tool leveraging a state-of-the-art language model to search relevant content inside the COVID-19 Open Research Dataset (CORD-19) recently published by the White House and its research partners. The dataset contains over 44,000 scholarly articles about COVID-19, SARS-CoV-2 and related coronaviruses.

Various models already fine-tuned on Natural Language Inference are available to perform the search:

scibert-nli, a fine-tuned version of AllenAI's SciBERT [1].
biobert-nli, a fine-tuned version of BioBERT by J. Lee et al. [2]
covidbert-nli, a fine-tuned version of Deepset's CovidBERT.
clinicalcovidbert-nli, a fine-tuned version of @manueltonneau's ClinicalCovidBERT.

All models are trained on SNLI [3] and MultiNLI [4] using the sentence-transformers library [5] to produce universal sentence embeddings [6]. Embeddings are subsequently used to perform semantic search on CORD-19.

Currently supported operations are:

Browse paper abstract with interactive queries.
Reproduce SciBERT-NLI, BioBERT-NLI and CovidBERT-NLI training results.

Setup

Python 3.6 or higher is required to run the code. First, install the required libraries with pip, then download the en_core_web_sm language pack for spaCy and data for NLTK:

pip install -r requirements.txt
python -m spacy download en_core_web_sm
python -m nltk.downloader punkt

Using the Browser

First of all, download a model fine-tuned on NLI from HuggingFace's cloud repository.

python scripts/download_model.py --model scibert-nli

Second, download the data from the Kaggle challenge page and place it in the data folder.

Finally, simply run:

python scripts/interactive_search.py

to enter the interactive demo. Using a GPU is suggested since the creation of the embeddings for the entire corpus might be time-consuming otherwise. Both the corpus and the embeddings are cached on disk after the first execution of the script, and execution is really fast after embeddings are computed.

Use the interactive demo as follows:

Reproducing Training Results for Transformers

First, download a pretrained model from HuggingFace's cloud repository.

python scripts/download_model.py --model scibert

Second, download the NLI datasets used for training and the STS dataset used for testing.

python scripts/get_finetuning_data.py

Finally, run the finetuning script by adjusting the parameters depending on the model you intend to train (default is scibert-nli).

python scripts/finetune_nli.py

The model will be evaluated against the test portion of the Semantic Text Similarity (STS) benchmark dataset at the end of training. Please refer to my model cards for additional references on parameter values.