Add weightless hyper-connection fusion operators #32687
Add HyperConnectionMix CUDA contrib operator
3b5660cd
Format HyperConnectionMix CUDA kernel
c12532dd
Remove incorrect HyperConnectionMix operator
0be28087
Add weightless hyper-connection fusion operators
6ce82512
Split hyper-connection kernels by operator
e31e472a
Add dedicated hyper-connection shape inference
f51ae49a
Allow flattened HyperConnection pre-mix gates
ccda2376
Fix hyper-connection CI and review issues
853b48a2
Fix provider build for branchwise RMS norm
b4eeea8a
Optimize CUDA branchwise RMS normalization
ccfeb4ae
Fix CUDA type mapping for branchwise RMS norm
0590a549
Login to write a write a comment.
Login via GitHub