modeling

class GPT2Model(vocab_size, hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, hidden_act='gelu', hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, type_vocab_size=16, initializer_range=0.02, pad_token_id=0)

Bases: paddlenlp.transformers.gpt2.modeling.GPT2PretrainedModel

The base model of gpt2.

forward(input_ids, position_ids=None, attention_mask=None, use_cache=False, cache=None)

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

• *inputs (tuple) – unpacked tuple arguments

• **kwargs (dict) – unpacked dict arguments

class GPT2PretrainedModel(name_scope=None, dtype='float32')

Bases: paddlenlp.transformers.model_utils.PretrainedModel

An abstract class for pretrained GPT2 models. It provides GPT2 related model_config_file, resource_files_names, pretrained_resource_files_map, pretrained_init_configuration, base_model_prefix for downloading and loading pretrained models. See PretrainedModel for more details.

init_weights(layer)

Initialization hook

base_model_class

alias of paddlenlp.transformers.gpt2.modeling.GPT2Model

class GPT2ForPretraining(gpt2)

Bases: paddlenlp.transformers.gpt2.modeling.GPT2PretrainedModel

The pretraining model of GPT2.

It returns some logits and cached_kvs.

forward(input_ids, position_ids=None, attention_mask=None, masked_positions=None, use_cache=False, cache=None)

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

• *inputs (tuple) – unpacked tuple arguments

• **kwargs (dict) – unpacked dict arguments

class GPT2PretrainingCriterion

Bases: paddle.fluid.dygraph.layers.Layer

Criterion for GPT2.

It calculates the final loss.

forward(prediction_scores, masked_lm_labels, loss_mask)

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

• *inputs (tuple) – unpacked tuple arguments

• **kwargs (dict) – unpacked dict arguments