transformers

1,921 Commits 171 Branches 49 Tags

3 Commits

Author	SHA1	Message	Date
Simon Layton	899883644f	Fix test fails and warnings Attention output was in bnij ordering instead of ijbn which everything else will expect. This was an oversight on my part, and keeps the attention inputs/outputs identical to the original code. Also moved back from tensor slicing to index_select in rel_shift_bnij to make the tracer happy.	2019-10-03 12:05:15 -04:00
Simon Layton	9ffda216ec	Fix missed head transpose	2019-10-03 09:23:16 -04:00
Simon Layton	d51b589404	Re-order attention head outputs for better perf Significant performance boost over the original orderings on an already somewhat optimised branch this gave me > 2x end-to-end throughput on a squad xlnet fine-tuning task (batch 8, seq-length 612, fp16)	2019-10-02 12:18:21 -04:00

Author

SHA1

Message

Date

Simon Layton

899883644f

Fix test fails and warnings

Attention output was in bnij ordering instead of ijbn which everything
else will expect. This was an oversight on my part, and keeps the
attention inputs/outputs identical to the original code.

Also moved back from tensor slicing to index_select in rel_shift_bnij to
make the tracer happy.

2019-10-03 12:05:15 -04:00

Simon Layton

9ffda216ec

Fix missed head transpose

2019-10-03 09:23:16 -04:00

Simon Layton

d51b589404

Re-order attention head outputs for better perf

Significant performance boost over the original orderings
on an already somewhat optimised branch this gave me > 2x end-to-end
throughput on a squad xlnet fine-tuning task (batch 8, seq-length 612,
fp16)

2019-10-02 12:18:21 -04:00