Skip to content

Prophetnet optimization #9453

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
patrickvonplaten merged 4 commits into from
Jan 7, 2021
Merged

Prophetnet optimization #9453

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
561abb8
Vectorized `ngram_attention_bias` calculation
guillaume-be Jan 7, 2021
287f9cb
updated formatting with black
guillaume-be Jan 7, 2021
3becee4
Further optimization
guillaume-be Jan 7, 2021
15542f5
one (last) optimization
guillaume-be Jan 7, 2021
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
12 changes: 7 additions & 5 deletions src/transformers/models/prophetnet/modeling_prophetnet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,13 +171,15 @@ def ngram_attention_bias(sequence_length, ngram, device, dtype):
"""
This function computes the bias for the predict stream
"""
bias = torch.ones((ngram, sequence_length, 2 * sequence_length), device=device, dtype=dtype) * float("-inf")
left_block = torch.ones((ngram, sequence_length, sequence_length), device=device, dtype=dtype) * float("-inf")
right_block = left_block.detach().clone()
# create bias
for stream_idx in range(ngram):
for i in range(sequence_length):
bias[stream_idx, i, sequence_length + i] = 0
bias[stream_idx, i, : max(i - stream_idx, 0) + 1] = 0
return bias
right_block[stream_idx].fill_diagonal_(0, wrap=False)
left_block[stream_idx].triu_(-stream_idx + 1)

left_block[:, :, 0] = 0
return torch.cat([left_block, right_block], dim=2)


def compute_relative_buckets(num_buckets, max_distance, relative_positions, is_bidirectional=False):
Expand Down