|
@@ -279,7 +279,7 @@ def flash_attn_unpadded_kvpacked_func(q, kv, cu_seqlens_q, cu_seqlens_k, max_seq
|
|
|
(they might not have the right scaling).
|
|
|
deterministic: bool. Whether or not to ensure deterministic execution.
|
|
|
Return:
|
|
|
- out: (total, nheads, headdim).
|
|
|
+ out: (total_q, nheads, headdim).
|
|
|
softmax_lse [optional, if return_attn_probs=True]: (batch_size, nheads, seqlen). The
|
|
|
logsumexp of each row of the matrix QK^T * scaling (e.g., log of the softmax
|
|
|
normalization factor).
|
|
@@ -315,7 +315,7 @@ def flash_attn_unpadded_func(q, k, v, cu_seqlens_q, cu_seqlens_k, max_seqlen_q,
|
|
|
(they might not have the right scaling).
|
|
|
deterministic: bool. Whether or not to ensure deterministic execution.
|
|
|
Return:
|
|
|
- out: (total, nheads, headdim).
|
|
|
+ out: (total_q, nheads, headdim).
|
|
|
softmax_lse [optional, if return_attn_probs=True]: (batch_size, nheads, seqlen). The
|
|
|
logsumexp of each row of the matrix QK^T * scaling (e.g., log of the softmax
|
|
|
normalization factor).
|