search

jiegzhan / multi-class-text-classification-cnn-rnn

Classify Kaggle San Francisco Crime Description into 39 classes. Build the model with CNN, RNN (GRU and LSTM) and Word Embeddings on Tensorflow.
https://www.kaggle.com/c/sf-crime/data
Apache License 2.0
599 stars 262 forks source link

train.py fails with best_model.ckpt not found #22

Open hilmij opened 7 years ago

hilmij commented 7 years ago

Hi, I am trying the example with Python 3.6.1 and TensorFlow 1.2.1 on Windows 10.

I am getting the following error when I run "python train.py ./data/train.csv.zip ./training_config.json".

CRITICAL:root:Saved model ./checkpoints_1501717661/model-2700 at step 2700 CRITICAL:root:Best accuracy 0.997291996203733 at step 2700 CRITICAL:root:Training is complete, testing the best model on x_test and y_test INFO:tensorflow:Restoring parameters from ./checkpoints_1501717661/model-2700 INFO:tensorflow:Restoring parameters from ./checkpoints_1501717661/model-2700 CRITICAL:root:Accuracy on test set: 0.9972894482090997 Traceback (most recent call last): File "train.py", line 161, in train_cnn_rnn() File "train.py", line 151, in train_cnn_rnn os.rename(path, trained_dir + 'best_model.ckpt') FileNotFoundError: [WinError 2] The system cannot find the file specified: './checkpoints_1501717661/model-2700' -> './trained_results_1501717661/best_model.ckpt'

I ran the train.py couple of times now. Same error. Please help me to solve this issue.

Thanks, Hilmi.

tcxdgit commented 7 years ago

Have you ever changed the original code files?If you have changed, maybe you can send your codes to me. I'll try to help you.

At 2017-08-03 08:44:13, "Hilmi Jauffer" notifications@github.com wrote:

Hi, I am trying the example with Python 3.6.1 and TensorFlow 1.2.1 on Windows 10.

I am getting the following error when I run "python train.py ./data/train.csv.zip ./training_config.json".

CRITICAL:root:Saved model ./checkpoints_1501717661/model-2700 at step 2700 CRITICAL:root:Best accuracy 0.997291996203733 at step 2700 CRITICAL:root:Training is complete, testing the best model on x_test and y_test INFO:tensorflow:Restoring parameters from ./checkpoints_1501717661/model-2700 INFO:tensorflow:Restoring parameters from ./checkpoints_1501717661/model-2700 CRITICAL:root:Accuracy on test set: 0.9972894482090997 Traceback (most recent call last): File "train.py", line 161, in train_cnn_rnn() File "train.py", line 151, in train_cnn_rnn os.rename(path, trained_dir + 'best_model.ckpt') FileNotFoundError: [WinError 2] The system cannot find the file specified: './checkpoints_1501717661/model-2700' -> './trained_results_1501717661/best_model.ckpt'

I ran the train.py couple of times now. Same error. Please help me to solve this issue.

Thanks, Hilmi.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

hilmij commented 7 years ago

Thanks for the reply. I did not change any files. Just trying to run as it is.

hilmij commented 7 years ago

I tested the same with Ubuntu (Python 3.5.2, TensorFlow 1.2.1). Same result.

Please let me know if you need more info to figure this issue.

INFO:tensorflow:Restoring parameters from ./checkpoints_1501820994/model-2500 INFO:tensorflow:Restoring parameters from ./checkpoints_1501820994/model-2500 CRITICAL:root:Accuracy on test set: 0.9972780593360288 Traceback (most recent call last): File "train.py", line 161, in train_cnn_rnn() File "train.py", line 151, in train_cnn_rnn os.rename(path, trained_dir + 'best_model.ckpt') FileNotFoundError: [Errno 2] No such file or directory: './checkpoints_1501820994/model-2500' -> './trained_results_1501820994/best_model.ckpt'

jiegzhan commented 7 years ago

https://github.com/jiegzhan/multi-class-text-classification-cnn-rnn/blob/master/train.py#L141

comment all the code after line 141, take a look at the checkpoint directory.

Then change the code after line 141 accordingly. Essentially you wanna save the best model to a file for prediction in future.

matevisky commented 7 years ago

fix line 151: from: os.rename(path, trained_dir + 'best_model.ckpt') to: os.rename(path + '.ckpt', trained_dir + 'best_model.ckpt')