textgenrnn
textgenrnn copied to clipboard
minimaxir
Easily train your own text-generating neural network of any size and complexity on any text dataset with a few lines of code.
How is fix this mistake ?
Traceback (most recent call last): File "C:/Users/telfo/Downloads/reddit.py", line 1, in import textgenrnn as textgen File "C:\Users\telfo\AppData\Local\Programs\Python\Python36\lib\site-packages\textgenrnn\__init__.py", line 1, in from .textgenrnn import textgenrnn File "C:\Users\telfo\AppData\Local\Programs\Python\Python36\lib\site-packages\textgenrnn\textgenrnn.py", line 1, in from tensorflow.keras.callbacks...
``` fasttext_model = fasttext.load_model('./fasttext/cc.en.300.bin') model_vocab = fasttext_model.get_words() zero_embedded = np.zeros((1,300)) embedding_matrix_v2 = np.zeros((len(model_vocab)+1,300)) embedding_matrix_v2[0,:] = zero_embedded for i in range(len(model_vocab)): embedding_matrix_v2[i+1,:] = fasttext_model.get_word_vector(model_vocab[i]) input = Input(shape=(cfg['max_length'],), name='input') embedded = Embedding(len(model_vocab)+1,...
I was wondering what's the biggest file textgenrnn can support well. I have a huge file and was wondering whether it would be worth it to just split it up...
Can you explain more about the gen parameter in .fit as I am unable to understand that how we are calculating loss based upon input and target vector?  
I expected to be choosing more from words than characters in interactive mode, particularly because the output from non-interactive `generate()` looks great. Is there something I am doing wrong?
I am trying to use textgenrnn but it trains for about an hour and then it does [this](https://pastebin.com/sDysPnX0) i tried shrinking my training data from 50000 lines to 25000 lines...
Hi, I would like to calculate "perplexity" and "accuracy" of my trained model. Is there any way to do that?
My dataset is a list of subject lines from IT tickets and the only thing I have done to them is remove blank lines and white space. Most of my...
