def evaluate(model, data_loader, args):
model.eval()
meter = Meter()
with torch.no_grad():
for batch in data_loader:
batch = tuple(t.to(args.device) for t in batch)
_, items = calc_loss(model, batch)
meter.add(*items)
return meter.average(), meter.print_str(False)
def evaluate_kcBert(model, data_loader, args):
model.eval()
meter = Meter()
with torch.no_grad():
for batch in data_loader:
batch = tuple(t.to(args.device) for t in batch)
noise_input_ids, clean_input_ids, noise_mask, clean_mask = batch
outputs = model(noise_input_ids,
labels=clean_input_ids,
attention_mask=noise_mask)
loss = outputs[0]
bsz = clean_input_ids.size(0)
items = [loss.data.item(), bsz, clean_mask.sum().item()]
meter.add(*items)
return meter.average(), meter.print_str(False)