BERT-36L-CSL Sepsis Diag is a machine learning model designed to predict in-hospital mortality related to sepsis using ICD diagnosis text. The model employs a BERT-style Transformer architecture, specifically configured with six encoder branches and six layers per branch. It is implemented in PyTorch and is available on Hugging Face.
Key technical specifications include a total of 21,997,698 parameters, a vocabulary size of 4,805, a hidden size of 128, four attention heads, a feed-forward dimension of 2,048, and a maximum sequence length of 30 tokens.
To use the model, users can load it with PyTorch and the provided BertClassifier class, initializing it with the specified vocabulary size and encoder configuration. Inference involves tokenizing input text and passing it through the model to obtain prediction logits, from which the predicted outcome can be derived. The model is built specifically for sepsis diagnosis tasks.
No information is provided regarding the intended user audience, licensing, or pricing.
In the Other AI space, BERT 36L CSL Sepsis Diag takes a focused approach. It focuses on predicting in-hospital sepsis mortality from patient diagnosis text using machine learning. BERT 36L CSL Sepsis Diag is an open-source project aimed at medical researchers. The project is open source (Open Source). It runs on the web, API, and the command line.
It is developed by fansen, and the product first shipped in 2024. Among its 5 catalogued features are sepsis prediction, ICD text input, and BERT architecture.
Latest indexed changes and source events
fansen/BERT-36L-CSL-Sepsis-Diag verified by the PulseGate indexer
Other apps tracked under the same category.
