Update the to_categorical in the data_utils.py #950
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
* add contrastive loss * add contrastive loss
The asarray function of numpy will keep ndarray's shape, so the original to_categorical function cannot return the right answer when y is a ndarray with 2-dimension or more. A reshape function is added for changing ndarray to 1-d array. And add a warning when the input array is 3-dimension or more.
* fix bidirectional_rnn working with TF 1.2 (resolves #818) * fix bidirectional RNN, ensure backward compatibility
* add contrastive loss * add contrastive loss * Fix Using a as a Python is not allowed
tensorflow:tf.variable_op_scope(values, name, default_name) is deprecated, use tf.variable_scope(name, default_name, values)
* Fix Grayscale Image Shape When using grayscale images, making the shape [None, width, height, 1] will allow you to use it directly with an input_layer, since it's a 4-D tensor, as apposed to [None, width, height]. Color images are 4-D by default ([None, width, height, 3]) * Fix Grayscale Image Shape Changed the addition to work for all versions of python (tuple unpacking like that is only 3.x) * Fix Typo Added a paren
The code are so ugly, so I format it.
* [Docs] Convolution layers: Typo fixes The markdown output should be less confused now. * [Docs] Ftrl Proximal optimizer: Typo fix Another missing `backtick`.
In the example, we point to the legacy seq2seq class in Tensorflow. In summaries.py, the format of the tag has changed in Tensorflow; this breaks the seq2seq example, so we hotfix the particular tag that's a problem in the seq2seq example.
* Improve to_categorical function nb_classes not required anymore * Refactoring examples
temporary add the older 'nb_classes' arg for older code compatibility.
* Adding upscore3D layer * Changing upscore2d to be consistent with upscore3d.
* Create VGG19.py * Update README.md
* Renamed tflearn.losses to regularizers. The functions within the renamed module are not losses but regularizers. * Replaced logging.warn with logging.warning. Logging.warn is now deprecated. * Extracted method from duplicate code in dnn.py.
* initializations: lazily import xavier and variance scaling from tf.contrib * variables: use vendored copy of tensorflow's add_arg_scope * data_utils: replace VocabularyProcessor with a lazy-loading proxy
* fixed directory mismatch in cifar10 loaddata * Update cifar10.py
… and build_hdf5_image_dataset() (#1086)
"Truncating type '%s' not understood" should have 'truncating', not 'padding'
* Fixing termlogs for R2 When using R2 as metric, training step displays 'val_acc' instead of 'val_R2' * Fixing termlogs for R2 While using R2 as metric, termlog displays 'val_acc' instead of 'val_R2'
hard sigmoid is faster to compute than sigmoid function
Now the regular training model saver won't delete the "best validation accuracy" models.
GLUEs are nonconvex, nonmonotonic unlike ReLU or ELU. Reference: Gaussian Error Linear Units (GELUs), Hendrycks et. al, 2018.
* add tf2 support * cleanup * update to 0.5.0
* add tf2 support * cleanup * update to 0.5.0 * fix & update setup
* Added fashion_mnist dataset * Added triplet loss
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.
I think it has a mistake in this function:
When I test the codes below from the Quickstart from tflearn :
//---------------------split line --------------------------------
import numpy as np
import tflearn
Download the Titanic dataset
from tflearn.datasets import titanic
titanic.download_dataset('titanic_dataset.csv')
Load CSV file, indicate that the first column represents labels
from tflearn.data_utils import load_csv
data, labels = load_csv('titanic_dataset.csv', target_column=0,
categorical_labels=True, n_classes=2)
//---------------------split line --------------------------------
the python IDE shows me the error message:
return (y[:, None] == np.unique(y)).astype(np.float32)
TypeError: list indices must be integers or slices, not tuple
I checked the source codes of the file "data_utils.py", the error was found in the function "to_categorical",
it should use "y = np.array(y)" before "return (y[:, None] == np.unique(y)).astype(np.float32) " because y should be a array rathen than a list.