Can you share the CMU-MOSEI features you use?

[dzr1026]

Dear author, should you share the CMU-MOSEI features you use? I am mainly concerned with the audio modality, as this is one of the few works that shows the result of a single modality of audio. I would like to do some work on your feature set so that it can be compared with your work, thank you very much.

[Coding511]

@dzr1026 Have you reproduced the results ?

[Coding511]

@dzr1026 I am getting this error while reading pickle file in colab:

KeyError                                  Traceback (most recent call last)
[/content/gdrive/MyDrive/Multimodal/Multimodal/train.py in <module>
     58         WA2 = torch.empty(10, 1)
     59     for index in range(D):
---> 60         train_data, train_data1, train_label, test_data, test_data1, Test_label = read_data_IE(cross, index)
     61         best = 0
     62         # if os.path.exists(model_path +'IE_sess{}_{}.ckpt'.format((index + 1), D)):

3 frames
[/content/gdrive/MyDrive/Multimodal/Multimodal/read_data.py in read_data_IE(cross, index)
      8 
      9 def read_data_IE(cross, index):
---> 10     S, T, Label = joblib.load('./features/IEMOCAP_ST.pkl')
     11     if cross == '5':
     12         print(

[/usr/local/lib/python3.9/dist-packages/joblib/numpy_pickle.py in load(filename, mmap_mode)
    585                     return load_compatibility(fobj)
    586 
--> 587                 obj = _unpickle(fobj, filename, mmap_mode)
    588     return obj

[/usr/local/lib/python3.9/dist-packages/joblib/numpy_pickle.py  in _unpickle(fobj, filename, mmap_mode)
    504     obj = None
    505     try:
--> 506         obj = unpickler.load()
    507         if unpickler.compat_mode:
    508             warnings.warn("The file '%s' has been gen

Can any one including author solve this please? I was putting this result in my comparison table, unfortunately cant reproduce the results :(

[dzr1026]

@dzr1026 I am getting this error while reading pickle file in colab:

KeyError                                  Traceback (most recent call last)
[/content/gdrive/MyDrive/Multimodal/Multimodal/train.py in <module>
     58         WA2 = torch.empty(10, 1)
     59     for index in range(D):
---> 60         train_data, train_data1, train_label, test_data, test_data1, Test_label = read_data_IE(cross, index)
     61         best = 0
     62         # if os.path.exists(model_path +'IE_sess{}_{}.ckpt'.format((index + 1), D)):

3 frames
[/content/gdrive/MyDrive/Multimodal/Multimodal/read_data.py in read_data_IE(cross, index)
      8 
      9 def read_data_IE(cross, index):
---> 10     S, T, Label = joblib.load('./features/IEMOCAP_ST.pkl')
     11     if cross == '5':
     12         print(

[/usr/local/lib/python3.9/dist-packages/joblib/numpy_pickle.py in load(filename, mmap_mode)
    585                     return load_compatibility(fobj)
    586 
--> 587                 obj = _unpickle(fobj, filename, mmap_mode)
    588     return obj

[/usr/local/lib/python3.9/dist-packages/joblib/numpy_pickle.py  in _unpickle(fobj, filename, mmap_mode)
    504     obj = None
    505     try:
--> 506         obj = unpickler.load()
    507         if unpickler.compat_mode:
    508             warnings.warn("The file '%s' has been gen

Can any one including author solve this please? I was putting this result in my comparison table, unfortunately cant reproduce the results :(

I can read "IEMOCAP_ST.pkl" using joblib, I just use his extracted features and do not reproduce the author's code.

[Coding511]

Can you spot the error here in my case?

[xiaomi1024]

Dear author, should you share the CMU-MOSEI features you use? I am mainly concerned with the audio modality, as this is one of the few works that shows the result of a single modality of audio. I would like to do some work on your feature set so that it can be compared with your work, thank you very much.

The features of CMU-MOSEI I used were provided by this dataset provider (https://github.com/A2Zadeh/CMU-MultimodalSDK). I have tried to process the features of this dataset from the raw data, but failed. Because I found the timestamps of many samples are incorrect and there are quite a lot of data involved, I had to give up and choose the feature set from datasets.

[xiaomi1024]

Can you spot the error here in my case?

I see that the error is a problem with reading the data, but this problem is normal in my case. Because we re-uploaded the data before and modified some of the code, is it because it is not the latest version?

[Coding511]

@xiaomi1024 I have downloaded the pickle file in drive and then run the train.py. It is still not working. Is anybody trying it on colab?