|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- qiaojin/PubMedQA |
|
language: |
|
- en |
|
metrics: |
|
- accuracy |
|
base_model: |
|
- google-bert/bert-base-uncased |
|
--- |
|
|
|
# GEM_PubMedQA Model Card |
|
|
|
This model card provides an overview of the GEM_PubMedQA model, a finetuned implementation of the GEM architecture designed for the PubMedQA dataset. |
|
|
|
## Purpose |
|
|
|
The GEM_PubMedQA model was developed to assess the performance of the GEM architecture on domain-specific datasets, with a focus on healthcare. The PubMedQA dataset, a key benchmark in this field, was selected to evaluate its effectiveness. |
|
|
|
## Key Details |
|
|
|
- **License**: Apache-2.0 |
|
- **Dataset**: qiaojin/PubMedQA |
|
- **Language**: English |
|
- **Metrics**: Accuracy: 92.5% |
|
- **Base Model**: google-bert/bert-base-uncased |
|
|
|
## Model Details |
|
|
|
The GEM_PubMedQA model is built on the GEM architecture and finetuned from the `google-bert/bert-base-uncased` model using the PubMedQA dataset. The training was performed with the following parameters: |
|
|
|
- **Number of epochs**: 5 |
|
- **Batch size**: 128 |
|
- **Learning rate**: 2e-5 |
|
- **Maximum sequence length**: 128 |
|
- **Gradient accumulation steps**: 2 |
|
- **Cluster size**: 256 |
|
- **Threshold**: 0.65 |
