No loop matching the specified signature 报错，训练了5000步后界面退出

[wing158]

File "D:\miniconda3\envs\ernerf\lib\site-packages\torch\utils\tensorboard\summary.py", line 380, in make_histogram cum_counts = np.cumsum(np.greater(counts, 0, dtype=np.int32)) TypeError: No loop matching the specified signature and casting was found for ufunc greater。 环境 ：windows 3090 24g

[yihua7]

Hi, It seems to be an environmental issue of your python. Could you please provide a more detailed error traceback?

[yihua7]

Windows is not recommended as all our code is evaluated on Ubuntu. The command without GUI may solve your problem:

CUDA_VISIBLE_DEVICES=0 python train_gui.py --source_path YOUR/PATH/TO/DATASET/jumpingjacks --model_path outputs/jumpingjacks --deform_type node --node_num 512 --is_blender --eval --gt_alpha_mask_as_scene_mask --local_frame --W 800 --H 800 --random_bg_color --white_background

[yihua7]

It seems to be related to the compatibility of your tensorboard version. Can you try (1) removing codes of tensorboard logging or (2) switching tensorboard version to 2.14.0?

[wing158]

1.12.1+cu113, python 3.10 ，tensorboard-2.8.0

[yihua7]

It seems to be related to the compatibility of your tensorboard version. Can you try (1) removing codes of tensorboard logging or (2) switching tensorboard version to 2.14.0?

Could you try these two ways? It should be an environmental issue.

[wing158]

谢谢，我先按方法1 移走tensorboard log

[yihua7]

Thanks for pointing out the missing model functionality. I just commited a version with the save model function (2e843fdb48d87137dfab1484bacadef2b89b0b0c). If any other questions arise, please feel free to contact me~

[yihua7]

Hello, do you need any further help with this issue?

[wing158]

按方法处理代码已正常运行，谢谢