def test(model, args, eval_examples, eval_features, device):
print("***** Eval *****")
RawResult = collections.namedtuple(
"RawResult", ["unique_id", "start_logits", "end_logits"])
output_prediction_file = os.path.join(args.checkpoint_dir,
"predictions_test.json")
output_nbest_file = output_prediction_file.replace('predictions', 'nbest')
all_input_ids = torch.tensor([f['input_ids'] for f in eval_features],
dtype=torch.long)
all_input_mask = torch.tensor([f['input_mask'] for f in eval_features],
dtype=torch.long)
all_segment_ids = torch.tensor([f['segment_ids'] for f in eval_features],
dtype=torch.long)
all_example_index = torch.arange(all_input_ids.size(0), dtype=torch.long)
eval_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
all_example_index)
eval_dataloader = DataLoader(eval_data,
batch_size=args.n_batch,
shuffle=False)
model.eval()
all_results = []
print("Start evaluating")
for input_ids, input_mask, segment_ids, example_indices in tqdm(
eval_dataloader, desc="Evaluating"):
input_ids = input_ids.to(device)
input_mask = input_mask.to(device)
segment_ids = segment_ids.to(device)
with torch.no_grad():
batch_start_logits, batch_end_logits = model(
input_ids, segment_ids, input_mask)
for i, example_index in enumerate(example_indices):
start_logits = batch_start_logits[i].detach().cpu().tolist()
end_logits = batch_end_logits[i].detach().cpu().tolist()
eval_feature = eval_features[example_index.item()]
unique_id = int(eval_feature['unique_id'])
all_results.append(
RawResult(unique_id=unique_id,
start_logits=start_logits,
end_logits=end_logits))
write_predictions(eval_examples,
eval_features,
all_results,
n_best_size=args.n_best,
max_answer_length=args.max_ans_length,
do_lower_case=True,
output_prediction_file=output_prediction_file,
output_nbest_file=output_nbest_file)
tmp_result = get_eval(args.test_file, output_prediction_file)
print(tmp_result)
'prediction_epoch' + str(i) + '.json')
output_nbest_file = os.path.join(
args.checkpoint_dir,
'nbest_epoch' + str(i) + '.json')
write_predictions(
dev_examples,
dev_data,
all_results,
n_best_size=args.n_best,
max_answer_length=args.max_ans_length,
do_lower_case=True,
output_prediction_file=
output_prediction_file,
output_nbest_file=output_nbest_file)
tmp_result = get_eval(args.dev_file,
output_prediction_file)
tmp_result['STEP'] = global_steps
print_rank0(tmp_result)
with open(args.log_file, 'a') as aw:
aw.write(
json.dumps(str(tmp_result)) + '\n')
if float(tmp_result['F1']) > best_f1:
best_f1 = float(tmp_result['F1'])
if float(tmp_result['EM']) > best_em:
best_em = float(tmp_result['EM'])
if float(tmp_result['F1']) + float(
tmp_result['EM']) > best_f1_em:
best_f1_em = float(
tmp_result['F1']) + float(
def evaluate(model, args, eval_examples, eval_features, device, global_steps,
best_f1, best_em, best_f1_em):
print("***** Eval *****")
RawResult = collections.namedtuple(
"RawResult", ["unique_id", "start_logits", "end_logits"])
output_prediction_file = os.path.join(
args.checkpoint_dir, "predictions_steps" + str(global_steps) + ".json")
output_nbest_file = output_prediction_file.replace('predictions', 'nbest')
all_input_ids = torch.tensor([f['input_ids'] for f in eval_features],
dtype=torch.long)
all_input_mask = torch.tensor([f['input_mask'] for f in eval_features],
dtype=torch.long)
all_segment_ids = torch.tensor([f['segment_ids'] for f in eval_features],
dtype=torch.long)
all_example_index = torch.arange(all_input_ids.size(0), dtype=torch.long)
eval_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
all_example_index)
eval_dataloader = DataLoader(eval_data,
batch_size=args.n_batch,
shuffle=False)
model.eval()
all_results = []
print("Start evaluating")
for input_ids, input_mask, segment_ids, example_indices in tqdm(
eval_dataloader, desc="Evaluating"):
input_ids = input_ids.to(device)
input_mask = input_mask.to(device)
segment_ids = segment_ids.to(device)
with torch.no_grad():
batch_start_logits, batch_end_logits = model(
input_ids, segment_ids, input_mask)
for i, example_index in enumerate(example_indices):
start_logits = batch_start_logits[i].detach().cpu().tolist()
end_logits = batch_end_logits[i].detach().cpu().tolist()
eval_feature = eval_features[example_index.item()]
unique_id = int(eval_feature['unique_id'])
all_results.append(
RawResult(unique_id=unique_id,
start_logits=start_logits,
end_logits=end_logits))
write_predictions(eval_examples,
eval_features,
all_results,
n_best_size=args.n_best,
max_answer_length=args.max_ans_length,
do_lower_case=True,
output_prediction_file=output_prediction_file,
output_nbest_file=output_nbest_file)
tmp_result = get_eval(args.dev_file, output_prediction_file)
tmp_result['STEP'] = global_steps
with open(args.log_file, 'a') as aw:
aw.write(json.dumps(tmp_result) + '\n')
print(tmp_result)
if float(tmp_result['F1']) > best_f1:
best_f1 = float(tmp_result['F1'])
if float(tmp_result['EM']) > best_em:
best_em = float(tmp_result['EM'])
if float(tmp_result['F1']) + float(tmp_result['EM']) > best_f1_em:
utils.torch_save_model(model,
args.checkpoint_dir, {
'f1': float(tmp_result['F1']),
'em': float(tmp_result['EM'])
},
max_save_num=1)
model.train()
return best_f1, best_em
