Spaces:
Running on Zero
Running on Zero
Commit ·
669c9d8
1
Parent(s): d13df9d
Install Sol-Attn TVM runtime
Browse files- 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`).
|