search

minimaxir / textgenrnn

Easily train your own text-generating neural network of any size and complexity on any text dataset with a few lines of code.
Other
4.94k stars 751 forks source link

random keyerrors #29

Open ihavetoomanyquestions opened 6 years ago

ihavetoomanyquestions commented 6 years ago

Hi,

When I train my models, I randomly get the following key error. Any idea what might be causing the problem?

Traceback (most recent call last): File "train.py", line 39, in word_level=cfg['model_config']['word_level']) File "textgenrnn/textgenrnn.py", line 262, in train_new_model *kwargs) File "textgenrnn/textgenrnn.py", line 195, in train_on_texts validation_steps=val_steps File "miniconda2/envs/py36/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 91, in wrapper return func(args, kwargs) File "miniconda2/envs/py36/lib/python3.6/site-packages/keras/engine/training.py", line 1426, in fit_generator initial_epoch=initial_epoch) File "miniconda2/envs/py36/lib/python3.6/site-packages/keras/engine/training_generator.py", line 229, in fit_generator callbacks.on_epoch_end(epoch, epoch_logs) File "miniconda2/envs/py36/lib/python3.6/site-packages/keras/callbacks.py", line 77, in on_epoch_end callback.on_epoch_end(epoch, logs) File "textgenrnn-s/textgenrnn/utils.py", line 174, in on_epoch_end max_gen_length=self.max_gen_length) File "textgenrnn-s/textgenrnn/textgenrnn.py", line 94, in generate_samples self.generate(n, temperature=temperature, kwargs) File "textgenrnn-s/textgenrnn/textgenrnn.py", line 83, in generate max_gen_length) File "textgenrnn-s/textgenrnn/utils.py", line 69, in textgenrnn_generate next_char = indices_char[next_index] KeyError: 0

minimaxir commented 6 years ago

What's happening is that the model is trying to select a value representing an unknown char/word. I thought I put safeguards against this but apparently that was insufficient.

ihavetoomanyquestions commented 6 years ago

is there a quick fix I can do to the code, or is it more complicated? (would a try/catch somewhere at least prevent the script to crash?) because no matter what i do, retrain/use another data etc. it always come to that point where it crashes :(

Adattilio commented 5 years ago

I am getting this as well. I am using a cleaned up, shorter version of the cornell movie dialog corpus.

nm17 commented 4 years ago

@minimaxir the problem is still there, any ETC on this issue?