tree 2858d5c673cb678ffbb31314b7a0d55421a4b033
parent 28de72c482898a358bcce333c3fc91663cae267f
author harsh-nod <harsh@nod-labs.com> 1706597138 -0800
committer GitHub <noreply@github.com> 1706597138 -0800
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJluJsSCRC1aQ7uu5UhlAAAgVEQAIWAWaiU5AgEphuruoILicxy
 pC5q6i0NLwolzvKNK+sEX7BfZfqi+E69CL8jX13o8X0vtQm4Yhmz2nAFXP9WUXrS
 iyWAzs+zzX+F+Ubzukl6osMu4StK6poxNpnjVqM9cQM3mjS/xVzxAaB05QLBgjPj
 xxBpyJ0elNLS6YPiaMzkgpnuwyg8/pGe8HTZyMNgmb2I2GT1fhVT2Pj5WjY1PCh7
 SUptwEathSPVJUqI04t01UEwcgm8weAbl3nqtyN1kjM2T866f6DUyKWEyQLJJMlZ
 WA+73kh3VGo27C5cyPKaSMPFzQc/riD8uiozj6l62BSIp2ReSGhlaM0Py2CqAokZ
 UVjB/XZONpaGbD1Env35dOUQJkdZ6UDtObYvlBSAo7xLEvOP/YWZG1e0tlO3U35m
 7bQfqx0rshkzW4hY4KBdF0XoDmNB/9iC7o5g1TyCUzOvBLHZuYSJOJYgqCiSPVhT
 lLcVtuTjkOIFBNPk/DFtL1O96DvjDq4Pf2xjVzDch7z7MuYPiT11qGKIA+V1LwHh
 vaElgSUT9lKZamZUTWMHaiHOs3sCe9I3s1gP4ppZ4BdE4cTMMb7OxVpfzKO0/kRJ
 rjMiWKgvh4ODMUbqsuvWqbmnlbd/8jVteayLUWm+cl2qPSpxqkqHEWaMaOH9zKjw
 gtHt3gbh5pTFH0e4wsPo
 =5oS5
 -----END PGP SIGNATURE-----
 

[LinalgExt] Add optional transpose_v attribute to attention (#16254)

In the attention operator, the second contraction is a matrix
multiplication. Since many hardware vendors natively support matrix
multiplication with b transposed, a common benchmark is to assume that
the second operand of the second contraction is transposed, result in a
matrix multiplication with b transposed. This PR adds that optional
attribute and updates the tile and decompose pass to handle the case.