peter-wang321
commented
2 months ago I am very grateful for the author's response to my questions, which gave me important insights on reproducing the results of this repo.
And now I did it. The following are my settings about the environment, which is more important to keep it exactly consistent.
[GeminiFusion]$ python3 -c "import torch; print(torch.version)" 1.8.0+cu111 [GeminiFusion]$ python3 -V Python 3.6.13 :: Anaconda, Inc. [GeminiFusion]$ pip list Package Version
addict 2.4.0 certifi 2021.5.30 charset-normalizer 2.0.12 cycler 0.11.0 dataclasses 0.8 filelock 3.4.1 huggingface-hub 0.4.0 idna 3.7 importlib-metadata 4.8.3 importlib-resources 5.4.0 kiwisolver 1.3.1 matplotlib 3.3.4 mmcv-full 1.6.0 mmdet 2.25.0 numpy 1.19.5 opencv-python 4.5.1.48 packaging 21.3 Pillow 8.4.0 pip 21.2.2 pycocotools 2.0.7 pyparsing 3.1.2 python-dateutil 2.9.0.post0 PyYAML 6.0.1 requests 2.27.1 setuptools 58.0.4 six 1.16.0 terminaltables 3.1.10 timm 0.6.12 torch 1.8.0+cu111 torchaudio 0.8.0 torchvision 0.9.0+cu111 tqdm 4.64.1 typing_extensions 4.1.1 urllib3 1.26.19 wheel 0.37.1 yapf 0.32.0 zipp 3.6.0
Besides the environments, I also made two changes: First, deleting the "batch_first" since it is not supported in pytorch 1.8.0, and it is "batch_first=False" in the code. Second, replace (np.bool) with (bool) since it is not supported by Numpy 1.19.5.
That is what I did to reproduce it. I am not sure whether the changes will incur other potential issues, but now I have reproduced the results as reported in the Repo.
Many thanks to the authors.
JiaDingCN
Great work! Thanks for releasing the code. However, I cannot run the code successfully based on the instructions. Have you tested it by rebuilding a new conda environment? If you succeed in reproducing the results based on this repo., could you provide me with environments by running the following commands?
$ python3 -c "import torch; print(torch.__version__)" $ python3 -V $ pip list
Thanks a lot!