def pretrain(args,
config,
X_all,
lead_time_idx,
input_dim=None,
onehot_dim=None):
model = ModelWrapper(args, config, args.use_onehot, input_dim, onehot_dim,
lead_time_idx)
model.train((lead_time_idx, X_all))
return model
else:
aggregated_value["score"] = aggregated_value["accuracy"]
if best_val_score[val] < aggregated_value["score"]:
best_val_score[val] = aggregated_value["score"]
aggregated_value["best_score"] = best_val_score[val]
aggregated_value["val"] = val
print(json.dumps(aggregated_value, sort_keys=True, indent=4),
flush=True)
log_f.write(json.dumps(aggregated_value, sort_keys=True) + "\n")
log_f.flush()
print(f"################ {val} iteration end ################")
if model_config["mixed_precision"]:
scaler = torch.cuda.amp.GradScaler()
idx = 0
best_val_score = {val: 0.0 for val in downsteam_task_config["val"]}
for epoch in range(downsteam_task_config["epoch"]):
model.train()
train()
model.eval()
with torch.no_grad():
val()
from model import ModelWrapper model_wrapper = ModelWrapper() model_wrapper.train()