Questions about document handling.

[chenyu-l]

I want to know what changes have been made to the two CSV files（audio_test.csv、audio_train.csv ）（in data/s2e/） to generate modified_ df_ test. csv、modified_ df_ train. csv ？ Because I met a mistake using 'torch.LongTensor(data[1])' directly for the original two CSV files when run lstm_classfier.py . And you didn't tell us how to generate the two new CSV file.

[Demfier]

The modified version oversamples data points for surprise and fear (as they are underrepresented). A 90/10 train/test split is then created.

[llitmus]

I was faced with the same problem as chenyu-I, and I did the split, and got modified_ df_ test. csv、modified_ df_ train. csv. but there's another problem: too many dimensions 'str' when running the lstm_classifier.

[Demfier]

@llitmus could you paste the full error log? It's difficult to for me to assist otherwise.

[llitmus]

This is my error when running the lstm_classifier. Something wrong in utils.py. Maybe I did the split false.

ValueError Traceback (most recent call last) ~/multimodal-speech-emotion-recognition-master/lstm_classifier/s2e/lstm_classifier.py in 49 optimizer = optim.Adam(model.parameters(), lr=config['learning_rate']) 50 ---> 51 train_batches = load_data() 52 test_pairs = load_data(test=True) 53

~/multimodal-speech-emotion-recognition-master/lstm_classifier/s2e/utils.py in load_data(batched, test, file_dir) 28 output_batch.append(e[1]) 29 batches.append([torch.FloatTensor(input_batch), ---> 30 torch.LongTensor(output_batch)]) 31 return batches 32

ValueError: too many dimensions 'str'

This is my split code. I add it in the 4th .ipynb
modified_df_train,modified_df_test = train_test_split(df, test_size=0.10)
modified_df_train.to_csv('../data/s2e/modified_df_train.csv', index=False)
modified_df_test.to_csv('../data/s2e/modified_df_test.csv', index=False)

print(modified_df_train.shape, modified_df_test.shape)

the output is alright . (1913, 10) (213, 10)

And I'm trying to use data/s2e/audio_train.csv in the lstm_classifier but failed either

Thank u so much!

[llitmus]

Oh! I just sovled the problem by droping the col 'wav_file' .hhh. But there's a loss of predit_proba.py .hhh in s2e. I copied the combined predit py ^-^ Anyway,thanks a lot!