Open 1274085042 opened 6 months ago
https://github.com/pytorch/pytorch/blob/v2.1.2/aten/src/ATen/native/transformers/cuda/flash_attn/fmha_api.cpp#L247-L248 cc @jbschlosser @bhosmer @cpuhrsch @erichan1 @drisspg @mikaylagawarecki
torch version == v2.1.2
If you wanted to update this to check that cu_seq_len_q is contiguous that would be sweet and I can review
🐛 Describe the bug
https://github.com/pytorch/pytorch/blob/v2.1.2/aten/src/ATen/native/transformers/cuda/flash_attn/fmha_api.cpp#L247-L248
cc @jbschlosser @bhosmer @cpuhrsch @erichan1 @drisspg @mikaylagawarecki
Versions
torch version == v2.1.2