Skip to content

Add inputs_embeds support when generating with GPT-J #21575

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gante merged 4 commits into from
Feb 13, 2023
Merged

Add inputs_embeds support when generating with GPT-J #21575

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
34552c3
gptj accepts input_embeds
dimitry12 Feb 10, 2023
af4bb9c
Merge branch 'huggingface:main' into dzmitry/gptj_input_embeds
dimitry12 Feb 10, 2023
f72cbe0
fixed formatting
dimitry12 Feb 10, 2023
299ba6e
Merge branch 'dzmitry/gptj_input_embeds' of github.com:dimitry12/tran…
dimitry12 Feb 10, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 19 additions & 9 deletions src/transformers/models/gptj/modeling_gptj.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -771,7 +771,7 @@ def get_output_embeddings(self):
def set_output_embeddings(self, new_embeddings):
self.lm_head = new_embeddings

def prepare_inputs_for_generation(self, input_ids, past_key_values=None, **kwargs):
def prepare_inputs_for_generation(self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs):
token_type_ids = kwargs.get("token_type_ids", None)
# only last token for inputs_ids if past is defined in kwargs
if past_key_values:
Expand All @@ -790,14 +790,24 @@ def prepare_inputs_for_generation(self, input_ids, past_key_values=None, **kwarg
position_ids = position_ids[:, -1].unsqueeze(-1)
else:
position_ids = None
return {
"input_ids": input_ids,
"past_key_values": past_key_values,
"use_cache": kwargs.get("use_cache"),
"position_ids": position_ids,
"attention_mask": attention_mask,
"token_type_ids": token_type_ids,
}

# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
if inputs_embeds is not None and past_key_values is None:
model_inputs = {"inputs_embeds": inputs_embeds}
else:
model_inputs = {"input_ids": input_ids}

model_inputs.update(
{
"past_key_values": past_key_values,
"use_cache": kwargs.get("use_cache"),
"position_ids": position_ids,
"attention_mask": attention_mask,
"token_type_ids": token_type_ids,
}
)

return model_inputs

@add_start_docstrings_to_model_forward(GPTJ_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
@add_code_sample_docstrings(
Expand Down