Skip to content
This repository was archived by the owner on Jul 7, 2023. It is now read-only.

Fixes Universal Transformer not converging bug #1192

Merged
merged 1 commit into from
Nov 1, 2018
Merged

Fixes Universal Transformer not converging bug #1192

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
2 changes: 1 addition & 1 deletion tensor2tensor/models/research/universal_transformer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -416,7 +416,7 @@ def update_hparams_for_universal_transformer(hparams):
# LSTM forget bias for lstm style recurrence.
hparams.add_hparam("lstm_forget_bias", 1.0)
# Uses the memory at the last step as the final output, if true.
hparams.add_hparam("use_memory_as_final_state", True)
hparams.add_hparam("use_memory_as_final_state", False)
# if also add a ffn unit to the transition function when using gru/lstm
hparams.add_hparam("add_ffn_unit_to_the_transition_function", False)

Expand Down