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])
token_type_ids = None
if self.use_token_type_ids:
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
mc_token_ids = None
if self.use_mc_token_ids:
mc_token_ids = ids_tensor([self.batch_size, self.num_choices], self.seq_length)
sequence_labels = None
token_labels = None
choice_labels = None
if self.use_labels:
sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
choice_labels = ids_tensor([self.batch_size], self.num_choices)
config = GPTJConfig(
vocab_size=self.vocab_size,
n_embd=self.hidden_size,
n_layer=self.num_hidden_layers,
n_head=self.num_attention_heads,
intermediate_size=self.intermediate_size,
hidden_act=self.hidden_act,
hidden_dropout_prob=self.hidden_dropout_prob,
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
n_positions=self.max_position_embeddings,
type_vocab_size=self.type_vocab_size,
initializer_range=self.initializer_range,
bos_token_id=self.bos_token_id,
eos_token_id=self.eos_token_id,
pad_token_id=self.pad_token_id,
rotary_dim=self.rotary_dim,
return_dict=True,
)
head_mask = ids_tensor([self.num_hidden_layers, self.num_attention_heads], 2)
return (
config,
input_ids,
input_mask,
head_mask,
token_type_ids,
mc_token_ids,
sequence_labels,
token_labels,
choice_labels,
)
def get_config(self):
return GPTJConfig(
vocab_size=self.vocab_size,
n_embd=self.hidden_size,
n_layer=self.num_hidden_layers,
n_head=self.num_attention_heads,
intermediate_size=self.intermediate_size,
hidden_act=self.hidden_act,
hidden_dropout_prob=self.hidden_dropout_prob,
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
n_positions=self.max_position_embeddings,
type_vocab_size=self.type_vocab_size,
initializer_range=self.initializer_range,
use_cache=True,
bos_token_id=self.bos_token_id,
eos_token_id=self.eos_token_id,
pad_token_id=self.pad_token_id,
)
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 = GPTJConfig(
vocab_size=self.vocab_size,
n_embd=self.hidden_size,
n_layer=self.num_hidden_layers,
n_head=self.num_attention_heads,
n_positions=self.max_position_embeddings,
use_cache=False,
bos_token_id=self.bos_token_id,
eos_token_id=self.eos_token_id,
pad_token_id=self.pad_token_id,
rotary_dim=self.rotary_dim,
)
return (config, input_ids, input_mask)
def get_large_model_config(self):
return GPTJConfig.from_pretrained("EleutherAI/gpt-j-6B")