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

Merge pull request #1695 from huggingface/models_inputs_embeds

Browse Source 
model forwards can take an inputs_embeds param
This commit is contained in:
Julien Chaumond
2019-11-05 09:55:28 -05:00
committed by GitHub
parent a44f112fb9 00337e9687
commit 7daacf00df
23 changed files with 387 additions and 158 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
transformers/modeling_openai.py
View File

@@ -322,6 +322,10 @@ OPENAI_GPT_INPUTS_DOCSTRING = r""" Inputs:
Mask to nullify selected heads of the self-attention modules.
Mask values selected in ``[0, 1]``:
``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
**inputs_embeds**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, embedding_dim)``:
Optionally, instead of passing ``input_ids`` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix.
"""
@add_start_docstrings("The bare OpenAI GPT transformer model outputting raw hidden-states without any specific head on top.",
@@ -373,14 +377,22 @@ class OpenAIGPTModel(OpenAIGPTPreTrainedModel):
for layer, heads in heads_to_prune.items():
self.h[layer].attn.prune_heads(heads)
def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None):
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None):
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
elif input_ids is not None:
input_shape = input_ids.size()
input_ids = input_ids.view(-1, input_shape[-1])
elif inputs_embeds is not None:
input_shape = inputs_embeds.size()[:-1]
else:
raise ValueError("You have to specify either input_ids or inputs_embeds")
if position_ids is None:
# This was used when we had a single embedding matrice from position and token embeddings
# start = self.config.vocab_size + self.config.n_special
# end = start + input_ids.size(-1)
# position_ids = torch.arange(start, end, dtype=torch.long, device=input_ids.device)
position_ids = torch.arange(input_ids.size(-1), dtype=torch.long, device=input_ids.device)
position_ids = position_ids.unsqueeze(0).expand_as(input_ids)
# Code is different from when we had a single embedding matrice from position and token embeddings
device = input_ids.device if input_ids is not None else inputs_embeds.device
position_ids = torch.arange(input_shape[-1], dtype=torch.long, device=device)
position_ids = position_ids.unsqueeze(0).view(-1, input_shape[-1])
# Attention mask.
if attention_mask is not None:
@@ -413,11 +425,8 @@ class OpenAIGPTModel(OpenAIGPTPreTrainedModel):
else:
head_mask = [None] * self.config.n_layer
input_shape = input_ids.size()
input_ids = input_ids.view(-1, input_ids.size(-1))
position_ids = position_ids.view(-1, position_ids.size(-1))
inputs_embeds = self.tokens_embed(input_ids)
if inputs_embeds is None:
inputs_embeds = self.tokens_embed(input_ids)
position_embeds = self.positions_embed(position_ids)
if token_type_ids is not None:
token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1))
@@ -495,13 +504,14 @@ class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel):
def get_output_embeddings(self):
return self.lm_head
def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None,
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None,
labels=None):
transformer_outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask)
head_mask=head_mask,
inputs_embeds=inputs_embeds)
hidden_states = transformer_outputs[0]
lm_logits = self.lm_head(hidden_states)
@@ -587,13 +597,14 @@ class OpenAIGPTDoubleHeadsModel(OpenAIGPTPreTrainedModel):
def get_output_embeddings(self):
return self.lm_head
def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None,
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None,
mc_token_ids=None, lm_labels=None, mc_labels=None):
transformer_outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask)
head_mask=head_mask,
inputs_embeds=inputs_embeds)
hidden_states = transformer_outputs[0]
lm_logits = self.lm_head(hidden_states)