Tri Dao 74b0761ff7 [FA3] BF16 forward vor 5 Monaten
..
cutlass @ 756c351b49 74b0761ff7 [FA3] BF16 forward vor 5 Monaten
flash_attn 40e534a7f6 Implement cache_leftpad vor 5 Monaten
ft_attention 50896ec574 Make nvcc threads configurable via environment variable (#885) vor 9 Monaten
fused_dense_lib 50896ec574 Make nvcc threads configurable via environment variable (#885) vor 9 Monaten
fused_softmax 50896ec574 Make nvcc threads configurable via environment variable (#885) vor 9 Monaten
layer_norm 50896ec574 Make nvcc threads configurable via environment variable (#885) vor 9 Monaten
rotary 50896ec574 Make nvcc threads configurable via environment variable (#885) vor 9 Monaten
xentropy 50896ec574 Make nvcc threads configurable via environment variable (#885) vor 9 Monaten