search

IDEA-Research / DINO

[ICLR 2023] Official implementation of the paper "DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection"
Apache License 2.0
2.08k stars 228 forks source link

undefined symbol: _ZNK3c1010TensorImpl18compute_contiguousEv #233

Open LANDDKPLA opened 8 months ago

LANDDKPLA commented 8 months ago

Failed when Compiling CUDA operators python test.py. Error information is as follows

(/data/Envs/dino39) /data/DINO/models/dino/ops$ python test.py
Traceback (most recent call last):
  File "/data/DINO/models/dino/ops/test.py", line 18, in <module>
    from functions.ms_deform_attn_func import MSDeformAttnFunction, ms_deform_attn_core_pytorch
  File "/data/DINO/models/dino/ops/functions/__init__.py", line 9, in <module>
    from .ms_deform_attn_func import MSDeformAttnFunction
  File "/data/DINO/models/dino/ops/functions/ms_deform_attn_func.py", line 18, in <module>
    import MultiScaleDeformableAttention as MSDA
ImportError: /data/Envs/dino39/lib/python3.9/site-packages/MultiScaleDeformableAttention-1.0-py3.9-linux-x86_64.egg/MultiScaleDeformableAttention.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZNK3c1010TensorImpl18compute_contiguousEv

My environment settings:

NVIDIA GeForce RTX 4090
gcc-9.4.0
cuda-11.8
python-3.9
torch==2.0.0 torchvision==1.15.1

The program were previously able to run python test.py in 2080 Ti, python-3.7, cuda-11.6 but failed to train because cuda out of memory. So I change to 4090 but still failed. What should I do?

meirashaf commented 2 weeks ago

You have the same problem as this issue https://github.com/CASIA-IVA-Lab/DPT/issues/1. Just make sure you install the model up until the Compiling CUDA operators.