How Do I Manually Migrate the tf.train.batch API?
Porting Cause
In TensorFlow, it is not recommended that the tf.train.batch API be used to process input data in queue mode. You are advised to use tf.data.Dataset.batch(batch_size) instead.
If you still want to use tf.train.batch on CANN, set num_threads to 1.
Porting Example
Original script:
1 | (texts, texts_tests, mels, mags, dones) = tf.train.batch([text, texts_test, mel, mag, done], shapes=[(hp.T_x,), (hp.T_x,), ((hp.T_y // hp.r), (hp.n_mels * hp.r)), (hp.T_y, (1 + (hp.n_fft // 2))), ((hp.T_y // hp.r),)], num_threads=2, batch_size=batch_size, capacity=(batch_size * 8), dynamic_pad=False) |
Script after porting:
1 | (texts, texts_tests, mels, mags, dones) = tf.train.batch([text, texts_test, mel, mag, done], shapes=[(hp.T_x,), (hp.T_x,), ((hp.T_y // hp.r), (hp.n_mels * hp.r)), (hp.T_y, (1 + (hp.n_fft // 2))), ((hp.T_y // hp.r),)], num_threads=1, batch_size=batch_size, capacity=(batch_size * 8), dynamic_pad=False) |
Parent topic: Common Operations