Issues
- 6
- 3
- 2
Can we build fa with torch 2.3?
#954 opened - 1
H20 compatibility
#953 opened - 1
- 1
Allow causal mask alignment configuration
#951 opened - 5
- 2
flash decoding algorithm numerical error
#949 opened - 1
can not install
#948 opened - 1
Three-dimensional local attention
#947 opened - 1
Relative postitions
#946 opened - 8
[bug] build is verrrrrrrrrrrrrrrrrrrry slow
#945 opened - 1
flash attention是否支持RTX8000
#944 opened - 2
build failed under miniconda3
#943 opened - 0
您好,如何在日志增加输出Tokens/gpu/s和TFLOPS
#942 opened - 23
- 3
- 1
Does it support Swin Transformer
#939 opened - 1
- 0
- 2
need cp312 whl!
#935 opened - 4
- 4
- 2
flash-attention imported, not running
#932 opened - 10
flash_attn_2_cuda.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN3c104cuda9SetDeviceEi
#931 opened - 2
Does flash-attention2 support L40?
#930 opened - 6
Import Error
#928 opened - 0
- 1
- 0
error in named_apply()
#925 opened - 7
Any plans to support tree attention mask?
#924 opened - 2
- 1
- 1
why flash can't accelerate on A40 machine?
#921 opened - 5
- 12
- 0
Sparse Masking (for Graphs)
#918 opened - 2
Adding support for sqrt of softmax scores
#917 opened - 3
- 9
- 1
- 1
- 5
- 1
- 1
How to use the _flash_attn_forward func
#908 opened - 1
- 1
- 2
- 1
- 1