mrfakename commited on
Commit
669c9d8
·
1 Parent(s): d13df9d

Install Sol-Attn TVM runtime

Browse files
Files changed (1) hide show
  1. requirements.txt +3 -0
requirements.txt CHANGED
@@ -24,6 +24,9 @@ comfy-kitchen==0.2.26
24
  sol-attn @ git+https://github.com/NVlabs/Sana.git@46031940ba8af5d18054217e571149579424c0b1#subdirectory=techniques/sparse_backends
25
  nvidia-cutlass-dsl>=4.5.0
26
  cuda-python>=13.0
 
 
 
27
  # No `kernels` pin on purpose: the Hub attention backends want `kernels>=0.12.3`, and that version breaks
28
  # transformers 5.8.0 at import.
29
  # PyAV muxes the generated soundtrack onto the frames (`encode_video`).
 
24
  sol-attn @ git+https://github.com/NVlabs/Sana.git@46031940ba8af5d18054217e571149579424c0b1#subdirectory=techniques/sparse_backends
25
  nvidia-cutlass-dsl>=4.5.0
26
  cuda-python>=13.0
27
+ # Sol-Attn compiles its CuTe callable with `--enable-tvm-ffi`, but its package currently leaves this transitive
28
+ # runtime undeclared. Without it the SM120 path compiles up to dispatch and then falls back to dense attention.
29
+ apache-tvm-ffi>=0.1.12
30
  # No `kernels` pin on purpose: the Hub attention backends want `kernels>=0.12.3`, and that version breaks
31
  # transformers 5.8.0 at import.
32
  # PyAV muxes the generated soundtrack onto the frames (`encode_video`).