Skip to content

Forward input keys to decoder. #944

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
markblee merged 1 commit into from
Jan 23, 2025
Merged

Forward input keys to decoder. #944

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
20 changes: 5 additions & 15 deletions axlearn/common/causal_lm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -475,22 +475,12 @@ def predict(self, input_batch: dict[str, Tensor]) -> dict[str, Tensor]:
hidden_states: a float Tensor of shape [batch_size, seq_len, hidden_dim]
"""
self._constrain_input_batch(input_batch)
input_ids: Tensor = input_batch["input_ids"]
token_type_ids: Optional[Tensor] = input_batch.get("token_type_ids")
input_segment_ids: Optional[Tensor] = input_batch.get("input_segment_ids")
input_positions: Optional[Tensor] = input_batch.get("input_positions")
# TODO(markblee): Simplify by using consistent naming between `input_positions` and
# `positions`, `input_segment_ids` and `segment_ids`.
# Decoder hidden states: [batch_size, target_len, hidden_dim].
decoder_output = self.decoder(
# TODO(markblee): Simplify by using consistent naming between `input_positions` and
# `positions`, `input_segment_ids` and `segment_ids`.
input_batch=dict(
input_ids=input_ids,
token_type_ids=token_type_ids,
input_segment_ids=input_segment_ids,
positions=input_positions,
),
)
return decoder_output
decoder_batch = {**input_batch}
decoder_batch["positions"] = input_batch.get("input_positions")
return self.decoder(input_batch=decoder_batch)

def _metrics(
self, input_batch: Nested[Tensor], *, predict_outputs: Nested[Tensor]
Expand Down
Loading