Search results

installation dependency issue

Status: Open.

#15 In alexzhang13/flashattention2-custom-mask;

· invoker-LL opened on Jan 4, 2025

AssertionError: First input (fp16) and second input (fp32) must have the same dtype!

Status: Open.

#14 In alexzhang13/flashattention2-custom-mask;

· 3bobo opened on Oct 4, 2024

Code fails the backward pass

Status: Open.

#13 In alexzhang13/flashattention2-custom-mask;

· RishikeshMagar opened on Sep 26, 2024

Questions about bf16 support and correctness check

Status: Open.

#12 In alexzhang13/flashattention2-custom-mask;

· xiabingquan opened on Sep 11, 2024

Implementing Standard Interface for _flash_attn_forward and _flash_attn_backward Functions

Status: Open.

#11 In alexzhang13/flashattention2-custom-mask;

· chenlidar opened on Sep 9, 2024

Support float trainable masks

Status: Open.

#10 In alexzhang13/flashattention2-custom-mask;

· michal-kustosz opened on Sep 6, 2024

pytorch 12.4

Status: Open.

#8 In alexzhang13/flashattention2-custom-mask;

· alita-moore opened on Aug 20, 2024

precision issue

Status: Open.

#7 In alexzhang13/flashattention2-custom-mask;

· dyhBUPT opened on Aug 19, 2024

illegal memory access with seqlen = 2048 and dimension = 64

Status: Open.

#2 In alexzhang13/flashattention2-custom-mask;

· Uwwal opened on Aug 1, 2024

Comparison to FlexAttention

Status: Open.

#1 In alexzhang13/flashattention2-custom-mask;

· michaelfeil opened on Jul 21, 2024