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.
Fixes ```WARNING:tensorflow:Model failed to serialize as JSON. Ignoring... Layer AttentionWeightedAverage has arguments in `__init__` and therefore must override `get_config`.``` Superfluous but gets rid of a warning, from what I understand...
Using a kwargs dict to add `recurrent_activation='sigmoid'` when no GPUs are present. I also removed the `K.backend()` check, doesn't make sense to me when you're using Keras from the tensorflow...
Hi there, First of all great library, I've been having lots of fun with it! Is it expected behaviour that the vocab.json file doesn't get updated once it's been created...
How can I set the length to my generated text?
Hi! This is a great library, thanks for making and maintaining it. It's fun to play with and the code is easy to follow. :) I'm trying to train a...
I have a use case where I _only_ want to save the model weights at the end of the training. This PR adds a check for save_epochs
What does 128/256 mean for the LSTM? How does this effect the results? What is the optimal rnn layers and batch size? What does word level do?
If adding space around each punctuation put after `train_new_model`, the Tokenizer didn't count punctuation as a whole word and didn't separate the sentence `"this sentence?"` to `["this", "sentence", "?"]` but...
In `textgenrnn/textgenrnn/textgenrnn.py` line 226 and 248, there is an inconsistency of dropout parameter where at line 226, the `dropout` parameter is used if there is `context_labels` and `multi_gpu` is `True`...