evaluate.py

from run import Runner
import sys
import torch

def evaluate(config_name, gpu_id, saved_suffix):
    runner = Runner(config_name, gpu_id)
    model = runner.initialize_model(saved_suffix)

    examples_train, examples_dev, examples_test = runner.data.get_tensor_examples()
    stored_info = runner.data.get_stored_info()

    runner.evaluate(
        model, examples_dev, stored_info, 0, official=False, conll_path=runner.config['conll_eval_path'], predict=True
    )  # Eval dev
    # print('=================================')
    # runner.evaluate(model, examples_test, stored_info, 0, official=False, conll_path=runner.config['conll_test_path'])  # Eval test


if __name__ == '__main__':
    config_name, saved_suffix, gpu_id = sys.argv[1], sys.argv[2], int(sys.argv[3])
    evaluate(config_name, gpu_id, saved_suffix)
    print(
        torch.cuda.max_memory_allocated(device=gpu_id)
    )