FlashMLA An open library of optimized attention kernels for efficiently running models that use multi-head latent attention on GPUs.
Visit productCosigns Received
Be the first Cosign
Put your name behind FlashMLA.
Add it to your Products I use list and share why you use it.