chunml / text-generator Goto Github PK

Hi Trung, I am trying to implement your code, but there is one line I couldn't understand. In the RNN_utils.py, Line 38 and Line 45. I think "X[i] = input_sequence" should be executed after the whole loop of “for j in range(seq_length)” is finished. In other words, "X[i] = input_sequence" should not be included inside the "for" loop. So does y[i] in Line 45. Is my understanding right? Thanks.

I'm trying to implement the recurrent neural network using Keras but with Tensorflow as the back-end instead of Theano. Is this possible? Any gotchas or tips/tricks are appreciated. Thank you!

Tried to load the model and continue training, it seems like it starts with high loss and the model seems to be totally untrained (start over at epoch 0). The previous trained work seems to be all gone. No error message from loading. Any clue? I upgraded to Keras 2.0.8 from .6, verified the setup of the network (using the same setup code) but no help.

Thanks for sharing.

Are the default parameters in the code the same as those used in your post?
And how large is your training data (Harry Potter series)?

Traceback (most recent call last):

],,NNVEEE,EKKkkkEKkkkqqqqqqqssqsq---rrrr!!!zKKrrrPkkzzKKkkrrKkkkrkKkkqqqqqqqssqsq---rrrr!!!zKKrrrPkk  File "C:/Users/username/AppData/Local/Programs/Python/Python36/text-generator-master/text-generator-master/recurrent_keras.py", line 54, in <module>
    model.load_weights('')
  File "C:\Users\usernameAppData\Local\Programs\Python\Python36\lib\site-packages\keras\models.py", line 721, in load_weights
    f = h5py.File(filepath, mode='r')
  File "C:\Users\username\AppData\Local\Programs\Python\Python36\lib\site-packages\h5py\_hl\files.py", line 269, in __init__
    fid = make_fid(name, mode, userblock_size, fapl, swmr=swmr)
  File "C:\Users\username\AppData\Local\Programs\Python\Python36\lib\site-packages\h5py\_hl\files.py", line 99, in make_fid
    fid = h5f.open(name, flags, fapl=fapl)
  File "h5py\_objects.pyx", line 54, in h5py._objects.with_phil.wrapper
  File "h5py\_objects.pyx", line 55, in h5py._objects.with_phil.wrapper
  File "h5py\h5f.pyx", line 78, in h5py.h5f.open
ValueError: Invalid file name (invalid file name)

Having this error when trying to force the program to load weights. Otherwise, it defaults to epoch 0 and learning progress is lost. The program is still able to create the HDF5 files as intended, it just doesn't load them.

Running the code gives me the following error:

Data length: 3204677 characters
Vocabulary size: 94 characters
Traceback (most recent call last):
File "recurrent_keras.py", line 36, in
X, y, VOCAB_SIZE, ix_to_char = load_data(DATA_DIR, SEQ_LENGTH)
File "/home/boris/text-generator/RNN_utils.py", line 30, in load_data
X = np.zeros((len(data)/seq_length, seq_length, VOCAB_SIZE))
MemoryError

If I make the training data file smaller (e.g. 200kb) it works fine.