i am trying to train a masked language model from scratch. i use before code to create the roberta model architecture. but when I compare it with RobertaLM, I found it does not have the GELU activation layer. could someone help explain how to correctly do this? thanks
config = RobertaConfig(
vocab_size= 50265,
max_position_embeddings=514,
num_attention_heads=12,
num_hidden_layers=12,
type_vocab_size=1,
)
model = RobertaForMaskedLM(config=config)
the roberta layer looks like this:
------ a lot of layers…
(encoder): RobertaEncoder(
(layer): ModuleList(
(0): RobertaLayer(
(attention): RobertaAttention(
(self): RobertaSelfAttention(
(query): Linear(in_features=768, out_features=768, bias=True)
(key): Linear(in_features=768, out_features=768, bias=True)
(value): Linear(in_features=768, out_features=768, bias=True)
(dropout): Dropout(p=0.1, inplace=False)
)
(output): RobertaSelfOutput(
(dense): Linear(in_features=768, out_features=768, bias=True)
(LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
(dropout): Dropout(p=0.1, inplace=False)
)
)
(intermediate): RobertaIntermediate(
(dense): Linear(in_features=768, out_features=3072, bias=True)
--------- missing gelu activation layer here
)
---------- a lot of layers…
when i load a pretrained it looks like this:
-------- a lot of layers…
(encoder): RobertaEncoder(
(layer): ModuleList(
(0): RobertaLayer(
(attention): RobertaAttention(
(self): RobertaSelfAttention(
(query): Linear(in_features=768, out_features=768, bias=True)
(key): Linear(in_features=768, out_features=768, bias=True)
(value): Linear(in_features=768, out_features=768, bias=True)
(dropout): Dropout(p=0.1, inplace=False)
)
(output): RobertaSelfOutput(
(dense): Linear(in_features=768, out_features=768, bias=True)
(LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
(dropout): Dropout(p=0.1, inplace=False)
)
)
(intermediate): RobertaIntermediate(
(dense): Linear(in_features=768, out_features=3072, bias=True)
(intermediate_act_fn): GELUActivation()
)
-----------a lot of layers…
