Improving Patient Safety Event Report Classification with Parameter Efficient Fine-tuned Language Model

MEng Project 2024-2025

University of Toronto

Focused on enhancing NLP model performance for patient safety applications, significantly improving classification accuracy and reducing computational complexity.

Instructor: Prof. Eldan Cohen

r_values = [16,32,256]
lora_alphas = [32,64,512]
lora_dropouts = [0., 0.1]

for r in r_values:
    for alpha in lora_alphas:
        for dropout in lora_dropouts:
            trainer = run('roberta-large', learning_rate=1e-3, num_train_epochs=10, per_device_train_batch_size=32, r=r, lora_alpha=alpha,lora_dropout=dropout, type = 'lora')