- 15 4月, 2020 40 次提交
-
-
由 Yuanzhong Xu 提交于
<5xi32> to <5x1xi32> is a reshape, not a transpose PiperOrigin-RevId: 306531112 Change-Id: I1e5541bc43997eda222837691bcbad7107f57982
-
由 Andy Ly 提交于
Variable reformatting is currently not supported with model parallelism. PiperOrigin-RevId: 306529947 Change-Id: I49f83bc1c76762d7bfb4607905c65e413f7bf278
-
由 Chuan He 提交于
PiperOrigin-RevId: 306529070 Change-Id: I625b01323eae9d5860665abcda6157b087745ad5
-
由 Karim Nosir 提交于
PiperOrigin-RevId: 306527646 Change-Id: I8d5582f12097ef94d0d9d8640f5a5b2cb48df3d3
-
由 Andy Ly 提交于
Update is autogenerated from TensorFlow op registry. PiperOrigin-RevId: 306526875 Change-Id: I95b3c4945f225343428fb005fa78fd84e277809c
-
由 A. Unique TensorFlower 提交于
_mirrored_update In this way it's easier to modify the arugments, which is needed to make the return type another DistributedVariable. PiperOrigin-RevId: 306525155 Change-Id: I8ea762fe555827ff8b4061109f7a1db884a9a910
-
由 Andrew Audibert 提交于
This way, tests don't need to explicitly depend on local credentials, which can cause Bazel issues due to depending on cc libraries from tf_py_test. PiperOrigin-RevId: 306524831 Change-Id: I67dd59348691608845b149e2d0dc5f1190e69d9f
-
由 Gunhan Gulsoy 提交于
TFlite build requirements break in pip smoke test in bazel 3.0 PiperOrigin-RevId: 306522753 Change-Id: Ib941511af53fad225bbe75b477ee5e77ffff5050
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306519238 Change-Id: Idd97a4de4fb0b5a88d905b5e846a76294d2e5951
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306518815 Change-Id: Ice094b324d7368914e1d0feecfd7bc129e629a4b
-
由 Andrew Audibert 提交于
PiperOrigin-RevId: 306518696 Change-Id: I118f6599d93b149ae21f3eab63595008ded2692f
-
由 Berkin Ilbeyi 提交于
spaces. PiperOrigin-RevId: 306516644 Change-Id: I11acd0e941f990aaee6e8f6db760a623ed148969
-
由 Reed Wanderman-Milne 提交于
Now the format is identical to as if a LossScaleOptimzier is not used, except that the loss scale is saved with a LossScaleOptimizer. This allows saving checkpoints with a LossScaleOptimizer and restoring without a LossScaleOptimizer, and vice versa. Checkpoints with LossScaleOptimizers created in older versions of TensorFlow can still be loaded. New checkpoints saved will use the new format. PiperOrigin-RevId: 306511555 Change-Id: Ie316ab8c4fbfec7babd6f7803d337799d0ff10a5
-
由 Feng Liu 提交于
This is to run the element-wise operation first and then move the tensor elments. This is particular useful when the element-wise operations can be fused to the preceding ops. PiperOrigin-RevId: 306510101 Change-Id: Ic055f0f70c3ca10325b54092700fbaa26c4b3c9e
-
由 Nat Jeffries 提交于
Create keyword spotting benchmark. Keyword model has scrambled weights, so it will only function as a benchmark for performance optimizations. PiperOrigin-RevId: 306508532 Change-Id: I056b546fb2e61165132c4c04fc95ec572ec94c48
-
由 Akshay Modi 提交于
PiperOrigin-RevId: 306503450 Change-Id: Ia54c1c2bd5a46eec1795a13914129fc601a4d09e
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306503296 Change-Id: Ia979339155466f99da5d306eca87ac6abb626d92
-
由 Gunhan Gulsoy 提交于
PiperOrigin-RevId: 306502691 Change-Id: I1cf0ffa12e6800cf691e5c189fd28806d8ed8cae
-
由 Robert David 提交于
PiperOrigin-RevId: 306497434 Change-Id: I1b94e40f5ea12e5f2c9fbb73be534b12016e76fe
-
由 Lucy Fox 提交于
Modify HLO DynamicSlice op definition such that start indices are a Variadic list rather than tensor. This allows for DynamicSlice ops where the start indices are the result of op computations which cannot be folded at compile time. Aside from modifying the op definition, this requires updating: - the TF Slice to HLO DynamicSlice legalization - the HLO DynamicSlice to Slice canonicalization - the HLO importer - the IREE lowering from HLO DynamicSlice to VMLA I also added a test for the HLO DynamicSlice exporter, as there wasn't one previously. PiperOrigin-RevId: 306495966 Change-Id: I5a080c63eaeedfceda034987cdf16177393ed3e8
-
由 Ran Chen 提交于
_mirrored_update In this way it's easier to modify the arugments, which is needed to make the return type another DistributedVariable. PiperOrigin-RevId: 306493210 Change-Id: I330a19ebb3045a2fc5e47b79e49c3ab45d1e8887
-
由 Robert David 提交于
PiperOrigin-RevId: 306493152 Change-Id: I1805afc68e4ffb37a076bec79e0e51eac9110336
-
由 A. Unique TensorFlower 提交于
Refactors memory_space_assignment so that it is easier for subclasses to implement different algorithms. PiperOrigin-RevId: 306492169 Change-Id: I6af9366706ec36d84c7a121f163cf2e55ffc8518
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306491528 Change-Id: Ie611bfcd826e62473385c2e4aea2e5305fe1eb07
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 306484877 Change-Id: Ibcf72e35c40a8a5d3570168d8e48aca77b86e36f
-
由 Robert David 提交于
PiperOrigin-RevId: 306484150 Change-Id: Ia2b1194a96ce5ba5eb781258afd958613b58a11f
-
由 A. Unique TensorFlower 提交于
so that we won't need to import all the dependencies. This CL simply copies the existing code into the new file. PiperOrigin-RevId: 306480426 Change-Id: I672a14c3edeeab975b5e75c18d632c8d71ecead4
-
由 Yunxing Dai 提交于
Unknown ranked tensor are expressed as xla shape T[0] as used in other places across the tf2xla bridge. I don't quite like it but fixing that is outside the scope of this cl. PiperOrigin-RevId: 306476269 Change-Id: I67eb6f93f38059003549a26f123dbcc299aa57d0
-
由 Feng Liu 提交于
PiperOrigin-RevId: 306472779 Change-Id: Iaf079951f492235c27ecdb5146849f3345a8276a
-
由 Amit Patankar 提交于
PiperOrigin-RevId: 306472255 Change-Id: Id1c760d984311a1037c71b167d1ccbb110bff9ce
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306465949 Change-Id: I6ace4b5e11e890a7cdd2b6f117491a2ccb55997c
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306464817 Change-Id: Id9a1ff1edce2b7bd0341963576aa2060c3e41899
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 306463186 Change-Id: I9114c574c1871d08d6b54cedd468db46b26d86ae
-
由 Scott Zhu 提交于
See https://github.com/tensorflow/tensorflow/issues/38403. The issue only occurs when the initial state is feed as a input tensor. PiperOrigin-RevId: 306462596 Change-Id: I6d2c7758d1801df61c01ec8c34357f146ec3f437
-
由 Ran Chen 提交于
PiperOrigin-RevId: 306461298 Change-Id: I7dfa6c99ac5820ccd0262ea984ddbfad0742deab
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 306458409 Change-Id: I4c887b93a782b68a8dc1facedc6596b36afffc84
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 306455732 Change-Id: I37bbc35282151422501501172af0d4c3d8982abb
-
由 Ran Chen 提交于
It's not supported and the test is failing. CentralStorage is essentially ParameterServer. PiperOrigin-RevId: 306455178 Change-Id: Ia201b4da4962067660e877bce01115619c08e57b
-
由 Mihai Maruseac 提交于
PiperOrigin-RevId: 306453507 Change-Id: I4dd5fbb04f8047dfd8bbdb3ae8765923949ceb7d
-
由 Mihai Maruseac 提交于
PiperOrigin-RevId: 306452288 Change-Id: Ib93fba77a4f3b9f141d04cc32662606fae2eede9
-