metadata
license: apache-2.0
language:
- en
- zh
metrics:
- accuracy
- f1
- precision
- recall
base_model:
- XiaoEnn/herberta_seq_512_V2
tags:
- TCM
- Classifier
- LLM
- Syndrome Differentiation
datasets:
- XiaoEnn/Syndrome_Differentiation_NK_test
SYD_Model: A Syndrome Differentiation Model fine-tuned based on the herberta pre-trained TCM model, applied in the field of Traditional Chinese Medicine Internal Medicine.
Introduction
Syndrome Differentiation Model_512_v2 is trained based on the pre-trained Chinese herbal medicine model herberta_seq_512_v2 on the Traditional Chinese Medicine Internal Medicine Syndrome Differentiation dataset. The Eval Accuracy, Eval F1, Eval Precision, and Eval Recall reach 0.9454, 0.9293, 0.9221, and 0.9454, respectively, representing improvements of approximately 8.1%, 10.3%, 10.9%, and 8.1% compared to the model trained on the base Roberta model.
DateBase
Extract 321 types of syndrome differentiation and descriptions from the Traditional Chinese Medicine Internal Medicine textbook, and then generate training and test sets.
Model_config
- max_length = 512
- batch_size = 16
- epochs = 26
Results
| Model Name | Eval Accuracy | Eval F1 | Eval Precision | Eval Recall |
|---|---|---|---|---|
| herberta_seq_512_v2 | 0.9454 | 0.9293 | 0.9221 | 0.9454 |