Recurrent layer to treat 2nd and the rest of inputs as initial_states #7691
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
wanasit
force-pushed
the
recurrent_inputs_as_init_states
branch
from
aef0616
to
9601c67
Compare
fchollet
previously approved these changes
Aug 24, 2017
tests/keras/layers/recurrent_test.py
Outdated
| initial_state = [np.random.random((num_samples, units)) | ||
| for _ in range(num_states)] | ||
| targets = np.random.random((num_samples, units)) | ||
| model.fit([main_inputs] + initial_state, targets) |
There was a problem hiding this comment.
Use train_on_batch to make the test lighter. Similarly I suggest you use 2 instead of num_samples
There was a problem hiding this comment.
Switching to train_on_batch, but num_samples is already defined as 2 at the beginning of the file, isn't it.
fchollet
reviewed
Aug 24, 2017
tests/test_model_saving.py
Outdated
| @keras_test | ||
| def test_saving_recurrent_layer_with_init_state(): | ||
| VECTOR_SIZE = 8 | ||
| INPUT_LENGTH = 20 |
There was a problem hiding this comment.
No need for all caps variable names here.
There was a problem hiding this comment.
changing to lower case
|
I believe |
|
I've never used the class. If you are not hurry, I may look into it this weekend. |
dschwertfeger
added a commit
to dschwertfeger/keras
that referenced
this pull request
Oct 8, 2017
…github.com:dschwertfeger/keras into tensorboard-callback-embeddings-from-layer-outputs * 'tensorboard-callback-embeddings-from-layer-outputs' of github.com:dschwertfeger/keras: Update for PPM format (keras-team#7750) Style fix Style fix Prepare 2.0.8 release. Update TFRecord example. Update backend docs. Style fix. Update README.md and setup.py. Minor optimizer consistency fixes. Recurrent layer to treat 2nd and the rest of inputs as initial_states (keras-team#7691) added 2 missing scripts to the README in the examples folder (keras-team#7734) Fixes and tests wrt step-wise training. Fix PEP8 (keras-team#7731) fix image_ocr example steps_per_epoch and validation_steps params (keras-team#7696) Remove requirement that batch_size divides len(x_train) in Variational Autoencoders examples. (keras-team#7716) Small fixes in training.py
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Hi. I'm submitting a PR for a bug reported in #7612.
I think the Recurrent layer should treat a part of the inputs as
initial_states. I think this is also what you want to do according to the API doc. If you have a different design, please suggest.