SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

[LED Test] fix common inputs pt for flaky pt-tf led test (#9459)

Browse Source 
* fix common inputs pt flakey led

* fix other tests correspondingly
This commit is contained in:
Patrick von Platen
2021-01-07 12:29:03 +01:00
committed by GitHub
parent ae5a32bb0d
commit a400fe8931
1 changed files with 13 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
tests/test_modeling_led.py
View File

@@ -110,7 +110,8 @@ class LEDModelTester:
# because its local attention only attends to `self.attention_window + 1` locations
# (assuming no token with global attention, otherwise the last dimension of attentions
# is x + self.attention_window + 1, where x is the number of tokens with global attention)
self.encoder_key_length = self.attention_window + 1
# x is set to 1
self.encoder_key_length = self.attention_window + 2
# because of padding `encoder_seq_length`, is different from `seq_length`. Relevant for
# the `test_attention_outputs` and `test_hidden_states_output` tests
@@ -149,6 +150,10 @@ class LEDModelTester:
def prepare_config_and_inputs_for_common(self):
config, inputs_dict = self.prepare_config_and_inputs()
global_attention_mask = torch.zeros_like(inputs_dict["input_ids"])
global_attention_mask[:, -1] = 1
inputs_dict["global_attention_mask"] = global_attention_mask
return config, inputs_dict
def create_and_check_decoder_model_past_large_inputs(self, config, inputs_dict):
@@ -196,9 +201,11 @@ class LEDModelTester:
encoder.save_pretrained(tmpdirname)
encoder = LEDEncoder.from_pretrained(tmpdirname).to(torch_device)
encoder_last_hidden_state_2 = encoder(inputs_dict["input_ids"], attention_mask=inputs_dict["attention_mask"])[
0
]
encoder_last_hidden_state_2 = encoder(
inputs_dict["input_ids"],
attention_mask=inputs_dict["attention_mask"],
global_attention_mask=inputs_dict["global_attention_mask"],
)[0]
self.parent.assertTrue((encoder_last_hidden_state_2 - encoder_last_hidden_state).abs().max().item() < 1e-3)
@@ -390,7 +397,8 @@ class LEDModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
)
out_len = len(outputs)
correct_outlen = 5
# global attention outputs are added as well => so +1 here
correct_outlen = 6
# loss is at first position
if "labels" in inputs_dict: