search

CPJKU / audio_sheet_retrieval

Learning Audio–Sheet Music Correspondences for Cross-Modal Retrieval and Piece Identification
25 stars 6 forks source link

Troubles when starting training #2

Open YangLuYang opened 4 years ago

YangLuYang commented 4 years ago 
Starting training...
Traceback (most recent call last):
  File "run_train.py", line 118, in <module>
    refinement_patience=refinement_patience)
  File "/home/yangluyang/asr_master/audio_sheet_retrieval/audio_sheet_retrieval/utils/train_dcca_pool.py", line 376, in fit
    for epoch in train(iter_funcs, data, train_batch_iter, valid_batch_iter, fit_cca):
  File "/home/yangluyang/asr_master/audio_sheet_retrieval/audio_sheet_retrieval/utils/train_dcca_pool.py", line 205, in train
    batch_res = iter_funcs['train'](*train_input)
...
...
ValueError: Input dimension mis-match. (input[0].shape[2] = 7, input[1].shape[2] = 10)
Apply node that caused the error: Elemwise{mul}(Reshape{4}.0, Elemwise{sub,no_inplace}.0)
Toposort index: 1086
Inputs types: [TensorType(float64, 4D), TensorType(float64, 4D)]
Inputs shapes: [(100, 32, 7, 12), (100, 32, 10, 12)]
Inputs strides: [(21504, 672, 96, 8), (30720, 960, 96, 8)]
Inputs values: ['not shown', 'not shown']
Outputs clients: [[Sum{axis=[0, 2, 3], acc_dtype=float64}(Elemwise{mul}.0)]]

I have installed audio_sheet_retrieval master branch and msmd tismir2018branch。