/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/timm/models/layers/init.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
Traceback (most recent call last):
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/./scripts/demo.py", line 5, in
from unidepth.models import UniDepthV1, UniDepthV2
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/unidepth/models/init.py", line 1, in
from .unidepthv1 import UniDepthV1
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/unidepth/models/unidepthv1/init.py", line 1, in
from .unidepthv1 import UniDepthV1
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/unidepth/models/unidepthv1/unidepthv1.py", line 17, in
from unidepth.models.unidepthv1.decoder import Decoder
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/unidepth/models/unidepthv1/decoder.py", line 14, in
from unidepth.layers import (MLP, AttentionBlock, ConvUpsample, NystromBlock,
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/unidepth/layers/init.py", line 5, in
from .nystrom_attention import NystromBlock
File "/home/shahao/my_study/PF3plat/UniDepth/UniDepth/unidepth/layers/nystrom_attention.py", line 7, in
from xformers.components.attention import NystromAttention
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/components/init.py", line 15, in
from .attention import Attention, build_attention # noqa
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/components/attention/init.py", line 18, in
from ._sputnik_sparse import SparseCS
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/components/attention/_sputnik_sparse.py", line 9, in
from xformers.ops import masked_matmul
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/ops/init.py", line 8, in
from .fmha import (
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/ops/fmha/init.py", line 10, in
from . import (
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/ops/fmha/triton_splitk.py", line 110, in
from ._triton.splitk_kernels import _fwd_kernel_splitK, _splitK_reduce
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/ops/fmha/_triton/splitk_kernels.py", line 632, in
_fwd_kernel_splitK_autotune[num_groups] = autotune_kernel(
^^^^^^^^^^^^^^^^
File "/home/shahao/anaconda3/envs/Unidepth/lib/python3.11/site-packages/xformers/ops/fmha/_triton/splitk_kernels.py", line 614, in autotune_kernel
kernel = triton.autotune(
^^^^^^^^^^^^^^^^
TypeError: autotune() got an unexpected keyword argument 'use_cuda_graph'
How to solve this problem？（Ask questions sincerely！！！
）
Hey i met the same error as you but i'm not sure your setting is the same as mine.
i split this into several small solutions:
1 cuda 124 not 118: it turns out that triton not supports cuda118 on windows
2 i changed "from xformers.components.attention import NystromAttention" into "from xformers.components.attention import *", probably there's ome structure difference and cause the error that couldn't find the location
3 i used triton 3.2.0 not 3.3.0
hope this could help you!
lpiccinelli-eth
