MMSA icon indicating copy to clipboard operation
MMSA copied to clipboard

安装后遇到一些问题

Open Guoxu233 opened this issue 3 years ago • 6 comments

SENA_run is not loaded due to missing dependencies. This is ok if you are not using M-SENA.

请问这个问题怎么解决呢? 麻烦解答一下 感谢

Guoxu233 avatar Aug 07 '22 07:08 Guoxu233

您好,出现“SENA_run is not loaded due to missing dependencies. This is ok if you are not using M-SENA.”怎么解决? 还出现了“RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspac es). Use .reshape(...) instead. ”这样的错误

IndustryZoe avatar Aug 08 '22 13:08 IndustryZoe

“SENA_run is not loaded due to missing dependencies. This is ok if you are not using M-SENA.”

这个不影响正常使用。后续会改进提示方式。

第二个问题需要详细信息。

FlameSky-S avatar Aug 12 '22 03:08 FlameSky-S

Traceback (most recent call last): File "", line 1, in File "D:\anaconda3\envs\pytorch1.12\lib\site-packages\MMSA\run.py", line 186, in MMSA_run torch.cuda.set_device(args['device']) File "D:\anaconda3\envs\pytorch1.12\lib\site-packages\torch\cuda_init_.py", line 314, in set_device torch._C._cuda_setDevice(device) RuntimeError: CUDA error: invalid device ordinal CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. 请问这个要怎么解决呢

Guoxu233 avatar Aug 22 '22 08:08 Guoxu233

“SENA_run is not loaded due to missing dependencies. This is ok if you are not using M-SENA.”

这个不影响正常使用。后续会改进提示方式。

第二个问题需要详细信息。

您好,请问一下,我把MMSA中的一个模型单独拎出来,然后使用ch-sims数据集训练该模型,使用的也是MMSA中的data_loader.py中的代码得到MMDataSets,最后跑出来的有些结果中corr会变成nan,导致最后的TEST的corr也是nan,比如:“TEST-(LF-DNN) >> Mult_acc_2: 0.6937 Mult_acc_3: 0.5427 Mult_acc_5: 0.2123 F1_score: 0.5682 MAE: 0.5893 Corr: nan Loss: 0.5889”这是为什么?在处理这个问题时,MMSA包里具体哪个部分考虑了这个?

IndustryZoe avatar Sep 08 '22 09:09 IndustryZoe

请问可以将自己的模型按照框架已有模型的格式加载进这个框架进行使用吗

Chi-Tian avatar Sep 21 '22 07:09 Chi-Tian

@Chi-Tian 欢迎将自己已发表论文中的模型集成进MMSA框架,可以提PR。如果是模型尝试(未发表)的话也可以使用MMSA框架进行调参,只要格式没问题,是可以使用的。

Columbine21 avatar Sep 21 '22 11:09 Columbine21