PG Field Guide
Learn
Techniques
Emerging
PRs
← Back to Architecture
Fused QKV projection
Architecture
Used in
1 PRs
Best BPB
1.1570
Avg BPB
1.1570
Submissions
PR #640
by CiprianFlorin-Ifrim
RECORD
1.1570
Hyperparameters Across PRs
pr_number
parameters
640
—