def prepare_config_and_inputs(self):
input_ids = ids_tensor([self.batch_size, self.seq_length],
self.vocab_size)
input_mask = None
if self.use_input_mask:
input_mask = random_attention_mask(
[self.batch_size, self.seq_length])
if input_mask is not None:
batch_size, seq_length = input_mask.shape
rnd_start_indices = np.random.randint(1,
seq_length - 1,
size=(batch_size, ))
for batch_idx, start_index in enumerate(rnd_start_indices):
input_mask[batch_idx, :start_index] = 1
input_mask[batch_idx, start_index:] = 0
config = CLIPTextConfig(
vocab_size=self.vocab_size,
hidden_size=self.hidden_size,
num_hidden_layers=self.num_hidden_layers,
num_attention_heads=self.num_attention_heads,
intermediate_size=self.intermediate_size,
dropout=self.dropout,
attention_dropout=self.attention_dropout,
max_position_embeddings=self.max_position_embeddings,
initializer_range=self.initializer_range,
)
return config, input_ids, input_mask
def get_config(self):
return CLIPTextConfig(
vocab_size=self.vocab_size,
hidden_size=self.hidden_size,
num_hidden_layers=self.num_hidden_layers,
num_attention_heads=self.num_attention_heads,
intermediate_size=self.intermediate_size,
dropout=self.dropout,
attention_dropout=self.attention_dropout,
max_position_embeddings=self.max_position_embeddings,
initializer_range=self.initializer_range,
)
def test_load_vision_text_config(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
# Save CLIPConfig and check if we can load CLIPVisionConfig from it
with tempfile.TemporaryDirectory() as tmp_dir_name:
config.save_pretrained(tmp_dir_name)
vision_config = CLIPVisionConfig.from_pretrained(tmp_dir_name)
self.assertDictEqual(config.vision_config.to_dict(), vision_config.to_dict())
# Save CLIPConfig and check if we can load CLIPTextConfig from it
with tempfile.TemporaryDirectory() as tmp_dir_name:
config.save_pretrained(tmp_dir_name)
text_config = CLIPTextConfig.from_pretrained(tmp_dir_name)
self.assertDictEqual(config.text_config.to_dict(), text_config.to_dict())
def prepare_config_and_inputs(self):
input_ids = ids_tensor([self.batch_size, self.seq_length],
self.vocab_size)
input_mask = None
if self.use_input_mask:
input_mask = random_attention_mask(
[self.batch_size, self.seq_length])
config = CLIPTextConfig(
vocab_size=self.vocab_size,
hidden_size=self.hidden_size,
num_hidden_layers=self.num_hidden_layers,
num_attention_heads=self.num_attention_heads,
intermediate_size=self.intermediate_size,
dropout=self.dropout,
attention_dropout=self.attention_dropout,
max_position_embeddings=self.max_position_embeddings,
initializer_range=self.initializer_range,
)
return config, input_ids, input_mask
