This model cannot handle extremely large dataset

[RayXu14]

Just to point out that use
tf.convert_to_tensor -> tf.train.slice_input_producer -> tf.train.shuffle_batch
will meet an error

ValueError: Cannot create a tensor proto whose content is larger than 2GB.

if dataset is too large

[4pal]

it means that you have very long sentences in your datasets which consumer a lot of memory during baching. you need to summarize your dataset line by line using an intelligent summarize like textrank which is supervised --doesn't require training. Then try this summarized dataset ..it can even be 10 GB and you won't have any problem.

[xlniu]

you can use feed_dict or tf.data