Haiping Chen
271dcefc15
fix keras model predict return result.
2 years ago
Haiping
197224fd74
Add pb model save ( #976 )
* Add check for dims of x and y in model.fit.
* Init the serialization of keras pb model.
* Add more facilities to the saved model framework.
* Add ListWrapper and ITrackable, and revise implmentations.
* Add serialized attributes.
* Implement layer serializations.
* Add lacked implementations (mainly MultiDeviceSaver).
* Support autograph.to_graph under graph mode.
* Add more implementations to the pb model save.
* Add more implementations to the keras part of pb model save.
* Refine some code after merge.
* Add two simple sequential test case of pb model save.
* Implement serializing attributes other keras arg definitions.
* Add alexnet pb save test.
* Check and refine the code.
---------
Co-authored-by: AsakusaRinne <AsakusaRinne@gmail.com>
2 years ago
Haiping Chen
c5cdf2c540
Fixed model.fit return results. #927
2 years ago
Haiping Chen
4f88109ae3
fix _self_tracked_trackables.
2 years ago
Haiping Chen
321ddfc13e
Fix Model.build.
2 years ago
Haiping Chen
a432807c23
Remove _layers in Layer.
2 years ago
Haiping Chen
2584b684e4
Fix conv_net.load_weights #956
2 years ago
Haiping Chen
49876bb952
Change model.build parameter to input_shape.
2 years ago
Haiping Chen
2adfcd2cf9
Fix Sequential model.summary missing layers. #960
2 years ago
AsakusaRinne
10f66f0baf
Fix IO Exception from keras.fit.
2 years ago
Haiping Chen
9b11d45906
Fix NeuralNetXorKeras accuracy. #952
2 years ago
hlx1120@outlook.com
430ac93137
Add Attention support and test it
3 years ago
Ahmed Elsayed
ce3ddb2321
Initial regularizers.
3 years ago
Oceania2018
d1fc44dcef
Add InboundLayers to Node
3 years ago
Oceania2018
18d2512ee5
fix keras sequential.
3 years ago
Oceania2018
ebfb3275cf
clean code.
3 years ago
Oceania2018
a4a99bb08f
add tf.linalg.global_norm #857
3 years ago
Oceania2018
87fa204247
ndarray mask set value.
3 years ago
vshakhlin
0e13a3a05d
Check if callContext is null or callContext.Value is null create a new ThreadLocal with CallContext.
3 years ago
Oceania2018
b048b62436
fix Embedding layer.
3 years ago
Oceania2018
67a70bf800
Remove _isCreatedInGraphMode in Tensor
3 years ago
Oceania2018
a0cc4735e5
rename SafeOpHandle to SafeEagerOpHandle
3 years ago
Oceania2018
3f0b1f9ef7
fix Multi Threaded unit test.
3 years ago
Oceania2018
8dd1eecdbe
remove unnecessary variables.
3 years ago
Oceania2018
cb7e0170eb
mean_absolute_percentage_error
4 years ago
Oceania2018
1460ec899b
fix TensorFlowOpLayer memory leak.
4 years ago
Oceania2018
e10e2807af
Release v0.60.2, change to .net standard 2.0.
4 years ago
Oceania2018
94601f5b89
throw exception from SetData
4 years ago
Oceania2018
35bcda9fd8
_CastGrad
4 years ago
Oceania2018
734430d48c
slice assign works
4 years ago
Oceania2018
22effd6553
Using Shape instead of TensorShape.
4 years ago
Oceania2018
8784c31cb3
change tensor shape to Shape.
4 years ago
Oceania2018
3c7207c251
fix GetDataType.
4 years ago
Oceania2018
3231500660
linspace and meshgrid.
4 years ago
Oceania2018
f3eb31fb87
Built-in numpy api.
4 years ago
Oceania2018
8048b6258a
Add RepeatDataSetCrash test.
4 years ago
Oceania2018
3d943a10c1
IDataset cardinality.
4 years ago
Oceania2018
a994a86794
Add mean_absolute_error to metrics.
4 years ago
Oceania2018
500f0c0cca
Fix preserve_cardinality for ParallelMapDataset.
4 years ago
Oceania2018
a7df67b13c
Fix timeseries_dataset_from_array.
4 years ago
Oceania2018
11fdd9a8ee
Remove Range class due to conflict with .net 5.
4 years ago
Oceania2018
954462f4f8
KerasObjectLoader
4 years ago
Oceania2018
96e0067061
Fixed Expand1d for keras 1d input.
4 years ago
Oceania2018
f0030ca9bb
reset_metrics for every epoch.
4 years ago
Oceania2018
b54cbaa772
Fix binary_accuracy for keras.
4 years ago
lsylusiyao
3309b036f6
change standard output (Console) to a custom one
4 years ago
Oceania2018
bbc2e98a51
IndexLookup, Accumulator.
4 years ago
Oceania2018
4ef675faf9
Consolidate MapDataset function.
4 years ago
Oceania2018
6a8665f111
Add Conv2DTranspose #735
4 years ago
lsylusiyao
178cdbc3e7
Add verbose function
4 years ago