**Describe the bug** None of the memory efficient attention kernels are working with Whisper implementation. **To Reproduce** Run whisper with paged_attention/flash attention (splash attention is working but internally falls back to vanilla)
Describe the bug
None of the memory efficient attention kernels are working with Whisper implementation.
To Reproduce
Run whisper with paged_attention/flash attention (splash attention is working but internally falls back to vanilla)