def log_params(curr_log_dir, index, root_handler, params):
"""
Logs the parameters given.
"""
logger = utils.create_logger(curr_log_dir, index=index)
logger.addHandler(root_handler)
logger.info("Current log dir: " + curr_log_dir)
logger.info('Training with parameter combination ' + str(index))
logger.info("With parameters: " + pformat(params[index]))
logger.info("")
batch = training_pipeline.request_batch(train_request)
curr_loss.append(batch.loss)
if len(curr_loss) % pipeline_params['save_every'] == 0:
loss = loss + curr_loss
np.save(history_path, loss, allow_pickle=True)
curr_loss = []
if __name__ == '__main__':
dataset, exp, index, _ = script_utils.get_args()
logdir, expirement_dir = script_utils.get_logdir(dataset, exp)
curr_log_dir = script_utils.make_dirs(logdir, index)
root_logger = utils.create_logger(logdir, name='root')
root_handler = root_logger.handlers[0]
root_logger.info("Starting experiment %s with dataset %s",
expirement_dir.split('-')[0], dataset)
params = json.load(open(logdir + "/param_dict.json"))
contrastive_params, _, model_params, _ = script_utils.get_params(params)
root_logger.info("Parameter dict: %s", pformat(contrastive_params[index]))
root_logger.info("")
pipeline = Contrastive
model = script_utils.get_model(index, model_params, root_logger)