def evaluate(self, eval_data, output_dir):
"""
Evaluates the model on eval_data.
Utility function to be used by the eval_model() method. Not intended to be used directly.
"""
tokenizer = self.tokenizer
device = self.device
model = self.model
args = self.args
eval_output_dir = output_dir
results = {}
if isinstance(eval_data, str):
with open(eval_data, 'r') as f:
eval_examples = json.load(f)
else:
eval_examples = eval_data
eval_dataset, examples, features = self.load_and_cache_examples(
eval_examples, evaluate=True, output_examples=True)
eval_sampler = SequentialSampler(eval_dataset)
eval_dataloader = DataLoader(eval_dataset,
sampler=eval_sampler,
batch_size=args["eval_batch_size"])
eval_loss = 0.0
nb_eval_steps = 0
preds = None
out_label_ids = None
model.eval()
all_results = []
for batch in tqdm(eval_dataloader, disable=args['silent']):
batch = tuple(t.to(device) for t in batch)
with torch.no_grad():
inputs = {
'input_ids': batch[0],
'attention_mask': batch[1],
}
if args['model_type'] != 'distilbert':
inputs['token_type_ids'] = None if args[
'model_type'] == 'xlm' else batch[2]
example_indices = batch[3]
if args['model_type'] in ['xlnet', 'xlm']:
inputs.update({'cls_index': batch[4], 'p_mask': batch[5]})
outputs = model(**inputs)
for i, example_index in enumerate(example_indices):
eval_feature = features[example_index.item()]
unique_id = int(eval_feature.unique_id)
if args['model_type'] in ['xlnet', 'xlm']:
# XLNet uses a more complex post-processing procedure
result = RawResultExtended(
unique_id=unique_id,
start_top_log_probs=to_list(outputs[0][i]),
start_top_index=to_list(outputs[1][i]),
end_top_log_probs=to_list(outputs[2][i]),
end_top_index=to_list(outputs[3][i]),
cls_logits=to_list(outputs[4][i]))
else:
result = RawResult(unique_id=unique_id,
start_logits=to_list(outputs[0][i]),
end_logits=to_list(outputs[1][i]))
all_results.append(result)
prefix = 'test'
if not os.path.isdir(output_dir):
os.mkdir(output_dir)
output_prediction_file = os.path.join(
output_dir, "predictions_{}.json".format(prefix))
output_nbest_file = os.path.join(
output_dir, "nbest_predictions_{}.json".format(prefix))
output_null_log_odds_file = os.path.join(
output_dir, "null_odds_{}.json".format(prefix))
if args['model_type'] in ['xlnet', 'xlm']:
# XLNet uses a more complex post-processing procedure
all_predictions, all_nbest_json, scores_diff_json = write_predictions_extended(
examples, features, all_results, args['n_best_size'],
args['max_answer_length'], output_prediction_file,
output_nbest_file, output_null_log_odds_file, eval_data,
model.config.start_n_top, model.config.end_n_top, True,
tokenizer, not args['silent'])
else:
all_predictions, all_nbest_json, scores_diff_json = write_predictions(
examples, features, all_results, args['n_best_size'],
args['max_answer_length'], False, output_prediction_file,
output_nbest_file, output_null_log_odds_file,
not args['silent'], True, args['null_score_diff_threshold'])
return all_predictions, all_nbest_json, scores_diff_json
def evaluate(self, eval_data, output_dir):
"""
Evaluates the model on eval_data.
Utility function to be used by the eval_model() method. Not intended to be used directly.
"""
tokenizer = self.tokenizer
device = self.device
model = self.model
args = self.args
if isinstance(eval_data, str):
with open(eval_data, "r") as f:
eval_examples = json.load(f)
else:
eval_examples = eval_data
eval_dataset, examples, features = self.load_and_cache_examples(
eval_examples, evaluate=True, output_examples=True
)
eval_sampler = SequentialSampler(eval_dataset)
eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args["eval_batch_size"])
model.eval()
all_results = []
for batch in tqdm(eval_dataloader, disable=args["silent"]):
batch = tuple(t.to(device) for t in batch)
with torch.no_grad():
inputs = {
"input_ids": batch[0],
"attention_mask": batch[1],
}
if args["model_type"] != "distilbert":
inputs["token_type_ids"] = None if args["model_type"] == "xlm" else batch[2]
example_indices = batch[3]
if args["model_type"] in ["xlnet", "xlm"]:
inputs.update({"cls_index": batch[4], "p_mask": batch[5]})
outputs = model(**inputs)
for i, example_index in enumerate(example_indices):
eval_feature = features[example_index.item()]
unique_id = int(eval_feature.unique_id)
if args["model_type"] in ["xlnet", "xlm"]:
# XLNet uses a more complex post-processing procedure
result = RawResultExtended(
unique_id=unique_id,
start_top_log_probs=to_list(outputs[0][i]),
start_top_index=to_list(outputs[1][i]),
end_top_log_probs=to_list(outputs[2][i]),
end_top_index=to_list(outputs[3][i]),
cls_logits=to_list(outputs[4][i]),
)
else:
result = RawResult(
unique_id=unique_id, start_logits=to_list(outputs[0][i]), end_logits=to_list(outputs[1][i]),
)
all_results.append(result)
prefix = "test"
os.makedirs(output_dir, exist_ok=True)
output_prediction_file = os.path.join(output_dir, "predictions_{}.json".format(prefix))
output_nbest_file = os.path.join(output_dir, "nbest_predictions_{}.json".format(prefix))
output_null_log_odds_file = os.path.join(output_dir, "null_odds_{}.json".format(prefix))
if args["model_type"] in ["xlnet", "xlm"]:
# XLNet uses a more complex post-processing procedure
(all_predictions, all_nbest_json, scores_diff_json,) = write_predictions_extended(
examples,
features,
all_results,
args["n_best_size"],
args["max_answer_length"],
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
eval_data,
model.config.start_n_top,
model.config.end_n_top,
True,
tokenizer,
not args["silent"],
)
else:
all_predictions, all_nbest_json, scores_diff_json = write_predictions(
examples,
features,
all_results,
args["n_best_size"],
args["max_answer_length"],
False,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
not args["silent"],
True,
args["null_score_diff_threshold"],
)
return all_predictions, all_nbest_json, scores_diff_json
