|
Issue Title
|
State
|
Comments
|
Created Date
|
Updated Date
|
Closed Date
|
|
Saving a model with SparseCategoricalCrossentropy loss breaks subsequent training
|
open
|
0 |
2023-01-31 |
2023-01-25 |
- |
|
InvalidArgumentError: Value for attr 'dtype' of bfloat16 is not in the list of allowed values: uint8, int32, int64, half, float, double
|
open
|
4 |
2023-01-31 |
2023-01-25 |
- |
|
What are tf.IndexedSlices and dense gradients? Why do a keras optimizer distinguish between them?
|
open
|
1 |
2023-01-30 |
2023-02-01 |
- |
|
Updating network weights using only a specific portion of batch samples
|
open
|
0 |
2023-01-28 |
2023-01-25 |
- |
|
Broken normalization layer save and load: very different results when using a model after it has been dumped to disk and reloaded if it uses normalization layers
|
open
|
5 |
2023-01-27 |
2023-02-01 |
- |
|
Output shape argument description in tf.keras.layers.MultiHeadAttention possibly wrong.
|
open
|
0 |
2023-01-26 |
2023-01-25 |
- |
|
Keras generates non-unique names for stateful LSTM's states
|
open
|
0 |
2023-01-25 |
2023-02-01 |
- |
|
keras.layers.Bidirectional constructor ignores trainable argument, causing models with trainable=False layers to incorrectly be cloned
|
closed
|
1 |
2023-01-24 |
2023-02-01 |
2023-01-27 |
|
Found a possible security concern
|
open
|
0 |
2023-01-22 |
2023-01-25 |
- |
|
Memory leak when using tf.Model and tf.Model.fit() in a loop. clear_session() does not help
|
open
|
3 |
2023-01-22 |
2023-01-25 |
- |
|
Two RandomFlip layers with same seed do not return the same result
|
open
|
3 |
2023-01-21 |
2023-02-01 |
- |
|
`amsgrad=True` of Adam optimizer causes InvalidArgumentError on TPU
|
closed
|
5 |
2023-01-21 |
2023-02-01 |
2023-01-30 |
|
model.fit() uses same chunk of validation data for all evaluation steps
|
open
|
2 |
2023-01-20 |
2023-01-25 |
- |
|
keras.layers.Conv2D with dilation_rate > 1 becomes SpaceToBatchNd-Conv2D-BatchToSpaceNd in tflite conversion
|
open
|
2 |
2023-01-19 |
2023-01-25 |
- |
|
keras.layers.Bidirectional is missing argument "states", which is defined for keras.layers.RNN
|
open
|
3 |
2023-01-19 |
2023-02-01 |
- |
|
Dividing MNIST image into blocks (without overlapping) and run simple CNN through different blocks. Found 4 input samples and 60000 target samples.
|
open
|
2 |
2023-01-19 |
2023-01-25 |
- |
|
Gradients are [None,...] in text GAN model
|
open
|
4 |
2023-01-18 |
2023-01-25 |
- |
|
keras.preprocessing.image.DirectoryIterator should perform the same checks as keras.preprocessing.image.DataFrameIterator
|
open
|
0 |
2023-01-18 |
2023-01-25 |
- |
|
cannot find reference 'keras in __init__ py __init__ py
|
open
|
2 |
2023-01-18 |
2023-02-01 |
- |
|
Missing methods to easily access reset_state and states within keras.Model
|
open
|
7 |
2023-01-18 |
2023-01-25 |
- |
|
`test_model_with_brackets_in_shape` test in vis_utils_test.py is failing
|
closed
|
5 |
2023-01-17 |
2023-02-01 |
2023-01-27 |
|
Model save issues with custom optimizer - DPKerasSGDOptimizer
|
open
|
2 |
2023-01-17 |
2023-01-25 |
- |
|
Feature extraction in pretrained model: different result from keras 2.3 to keras 2.4
|
open
|
2 |
2023-01-16 |
2023-01-25 |
- |
|
Enable the discussion section.
|
closed
|
3 |
2023-01-16 |
2023-01-25 |
2023-01-24 |
|
keras.models.clone_model creates different weight names for RNN layers
|
open
|
6 |
2023-01-16 |
2023-01-25 |
- |
|
RuntimeError: `merge_call` called while defining a new graph or a tf.function.
|
closed
|
25 |
2023-01-15 |
2023-01-25 |
2023-01-26 |
|
Non adaptable FeatureSpace features
|
open
|
1 |
2023-01-14 |
2023-01-25 |
- |
|
New optimizers fail to load CUDA installed through conda
|
open
|
3 |
2023-01-13 |
2023-01-25 |
- |
|
Functional model computes wrong output signature in mixed_fp16
|
open
|
3 |
2023-01-13 |
2023-02-01 |
- |
|
Updates to pydot dependency in `plot_model`
|
closed
|
3 |
2023-01-12 |
2023-01-25 |
2023-01-27 |
|
tf.keras.mixed_precision.LossScaleOptimizer causes Graph execution error when using tfa.optimizers.MultiOptimizer and mixed_precision
|
open
|
9 |
2023-01-12 |
2023-01-25 |
- |
|
Not sure how to solve this bug: "WARNING:tensorflow:Detecting that an object or model or tf.train.Checkpoint."
|
closed
|
1 |
2023-01-12 |
2023-02-01 |
2023-01-12 |
|
Issues in tf.keras.utils.timeseries_dataset_from_array
|
open
|
4 |
2023-01-12 |
2023-01-31 |
- |
|
Sequential models should let you add multiple layers at once
|
closed
|
7 |
2023-01-11 |
2023-01-31 |
2023-01-26 |
|
`GroupNormalization` layer does not compute broadcast shape correctly for specified axis
|
open
|
2 |
2023-01-07 |
2023-01-27 |
- |
|
Cant slice along batch for MixtureSameDistribution produced by Keras layer
|
open
|
5 |
2023-01-07 |
2023-01-27 |
- |
|
ImageDataGenerator Issue: "height_shift_range" shifts the image horizontally and "width_shift_range" shifts the image vertically
|
open
|
2 |
2023-01-06 |
2023-01-27 |
- |
|
Mirrored strategy keras model.fit failure
|
open
|
11 |
2023-01-05 |
2023-01-28 |
- |
|
’model.load_weight‘ does not work in a particular framework's subclasses model and does not report an error
|
closed
|
4 |
2023-01-04 |
2023-01-28 |
2023-01-12 |
|
`expand_nested=True` not working same for `model.summary` and `tf.keras.utils.plot_model`
|
closed
|
5 |
2023-01-04 |
2023-01-28 |
2023-01-27 |
|
`timeseries_dataset_from_array` gives empty output on edge case
|
open
|
2 |
2023-01-04 |
2023-01-27 |
- |
|
Adam (and other adaptive algorithms) use epsilon the wrong way
|
closed
|
6 |
2023-01-04 |
2023-01-27 |
2023-01-26 |
|
Computing gradients for KerasTensors and allowing other operations
|
closed
|
3 |
2023-01-02 |
2023-01-27 |
2023-01-17 |
|
No shape mismatch error with different shapes of outputs and targets with tf.keras.losses.BinaryCrossentropy()
|
closed
|
6 |
2023-01-02 |
2023-01-27 |
2023-01-26 |
|
invalid syntax trainer.trainModel()
|
closed
|
7 |
2023-01-01 |
2023-01-27 |
2023-01-16 |
|
ConvNeXt `classifier_activation=softmax` returns linear layer
|
closed
|
3 |
2022-12-31 |
2023-01-27 |
2023-01-02 |
|
Augmentation layers only active during training!
|
closed
|
8 |
2022-12-31 |
2023-02-01 |
2023-01-12 |
|
using a custom layer, loaded_model cannot give the same predicted value compared with original model [results not reproducible]
|
open
|
5 |
2022-12-30 |
2023-01-27 |
- |
|
Transform label to ordinal_regression label
|
closed
|
5 |
2022-12-30 |
2023-01-26 |
2023-01-26 |
|
Serializing a `tf.keras.Model` in half-precision
|
open
|
11 |
2022-12-30 |
2023-01-25 |
- |