train code has some errors

[hearthewind]

It works fine when denovo from the provided model, (model.h5). When training our own model, there is an error

Train mode runs successfully, however, if we used the trained model to do denovo, the following exception.

File "denovo.py", line 228, in peps, ppeps, scores, pscores, ppms, _ = denovo(model, spectra, args.batch_size) File "denovo.py", line 181, in denovo pep, pos, positional_score = post_correction(rst, ms, c) File "denovo.py", line 73, in post_correction pep = topep(seq) File "/home/m/git/PepNet/utils.py", line 172, in topep raise(e) File "/home/m/git/PepNet/utils.py", line 170, in topep return ''.join(map(lambda n: id2amino[n], seq)).strip("[]") File "/home/m/git/PepNet/utils.py", line 170, in return ''.join(map(lambda n: id2amino[n], seq)).strip("[]") KeyError: 24

The root of the problem is the output shape of the model

` for rst, sp in zip(predictions, spectra): ms, c = sp['mass'], sp['charge']

    print('rst', rst.shape)

    # run post correction
    try:
        pep, pos, positional_score = post_correction(rst, ms, c)
    except KeyError as e:
        print('rst', rst.shape)
        raise(e)

    predict_peps.append(pep)
    positional_scores.append(positional_score)
    scores.append(np.prod(positional_score))`

In denovo.py, the code section above, if we use the provided model, rst has shape (32,23), which is correct. However, if we use our own trained model, the shape of rst is (10000, 32)