def create_and_check_xlm_token_classif(
self,
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
input_mask,
):
config.num_labels = self.num_labels
model = XLMForTokenClassification(config)
model.to(torch_device)
model.eval()
loss, logits = model(input_ids,
attention_mask=input_mask,
labels=token_labels)
result = {
"loss": loss,
"logits": logits,
}
self.parent.assertListEqual(
list(result["logits"].size()),
[self.batch_size, self.seq_length, self.num_labels])
self.check_loss_output(result)
def init(cls, cf, d_out):
if "bert" in cf.model_pretrained:
model = BertForTokenClassification.from_pretrained(cf.model_pretrained, num_labels=d_out,
output_attentions=False, output_hidden_states=False)
if cf.random_weights is True:
# initiate Bert with random weights
print("randomizing weights")
model = randomize_model(model)
#print(model.classifier.weight.data)
else:
# initiate Bert with pre-trained weights
print("keeping Bert with pre-trained weights")
#print(model.classifier.weight.data)
elif "xlm" in cf.model_pretrained:
model = XLMForTokenClassification.from_pretrained(cf.model_pretrained, num_labels=d_out, output_attentions=False, output_hidden_states=False)
model.d_out = d_out
return model
def create_and_check_xlm_token_classif(
self,
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
):
config.num_labels = self.num_labels
model = XLMForTokenClassification(config)
model.to(torch_device)
model.eval()
result = model(input_ids, attention_mask=input_mask, labels=token_labels)
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
