OverLordGoldDragon / see-rnn

RNN and general weights, gradients, & activations visualization in Keras & TensorFlow
MIT License
179 stars 21 forks source link

tf 2.5, keras functional api, tf.data.Dataset #51

Closed max1mn closed 3 years ago

max1mn commented 3 years ago

Hello, this PR adds tensorflow 2.5.0 support with keras functional api. Also, input_data can be packed (input, labels) tf.data.Dataset.

tf.data.Dataset example: tf_dataset = tf.data.Dataset.from_generator(generator_func, output_types=output_types, output_shapes=output_shapes) tf_dataset = tf_dataset.batch(16) grads = get_gradients(model, 1, tf_dataset, None)

OverLordGoldDragon commented 3 years ago

Thanks for the PR.