aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
...
* | | | | | Simplify _get_grad_fn_name and other minor fixes.Gravatar Saurabh Saxena2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211519628
* | | | | | Fix CRS combiner for spatial partitioningGravatar HyoukJoong Lee2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211519250
* | | | | | Optimize CuboidConvolutionBackwardKernel (Conv3D kernel backprop).Gravatar Eugene Zhulenev2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * simplify contraction by collapsing inner dims into single dimension * get rid of expensive reverse op ~5X improvement when compiled with AVX. PiperOrigin-RevId: 211518363
* | | | | | Support <4D tensor inputs for pad/padv2Gravatar Jared Duke2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Fixes #21266 PiperOrigin-RevId: 211515918
* | | | | | Removed old dynamic learning rate support code.Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211514287
* | | | | | Disable rtti for builtin TFLite kernelsGravatar Jared Duke2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211514002
* | | | | | Fix bugs in backward convolution benchmarks + add labels.Gravatar Eugene Zhulenev2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211510051
* | | | | | Create layer norm LSTM custom Op.Gravatar Jian Li2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211505721
* | | | | | Minor tweaks to TFLite API docsGravatar Jared Duke2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211505612
* | | | | | Merge pull request #21061 from lgeiger:tfgan-fix-summaryGravatar TensorFlower Gardener2018-09-04
|\ \ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211502883
* | | | | | | Add more data fields to step proto.Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211501909
* | | | | | | Enable TensorRT in ci docker build.Gravatar Guangda Lai2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211500190
* | | | | | | Add IsValidSignature method to signature_def_utilsGravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211498364
* | | | | | | Fix the tensorrt dependency order in tensorflow/contrib/BUILD.Gravatar Guangda Lai2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211496364
* | | | | | | [TF:XLA] Bump open source llvm revision to r341347Gravatar Sanjoy Das2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211496283
* | | | | | | Extend hoisting monotonic functions out of min/max reductions to all ↵Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | monotonic unary functions. Add the ability to flip Max <-> Min if the function is non-increasing, e.g. Max(Neg(x)) => Neg(Min(x)). PiperOrigin-RevId: 211490436
* | | | | | | [XLA] Make kConvolution, kDot HLO attributes mandatoryGravatar David Majnemer2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | HLO transformations would forget to propagate the feature depth attribute. Making these attributes mandatory, while slightly less convenient for tests, makes HLO transformations more robust. PiperOrigin-RevId: 211490160
* | | | | | | Support zeros_like for nested TensorLists.Gravatar Saurabh Saxena2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211489741
* | | | | | | Merge pull request #21905 from ↵Gravatar TensorFlower Gardener2018-09-04
|\ \ \ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | wdirons:21833_disable_gpu_test_scatter_add_ndim_op_test PiperOrigin-RevId: 211489137
* | | | | | | | Fix LazyAdamOptimizer for sparse updates on resource variables.Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211488610
* | | | | | | | Removed redundant std::string -> string conversions.Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211487989
* | | | | | | | Merge pull request #21989 from perfinion:codeownersGravatar TensorFlower Gardener2018-09-04
|\ \ \ \ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211480449
* | | | | | | | | This CL adds extra tests for `contrib.eager.metrics` that check eager ↵Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | metrics combined with while loops. PiperOrigin-RevId: 211479604
* | | | | | | | | Added a new eager C API TFE_NewContextFromSession(), where TFE_NewContext willGravatar Mingsheng Hong2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | get an owned device mgr from the input session. One use case is in S4TF, we run a graph session to enqueue a tensor into a fifo queue, and then call TFE_Execute() on a dequeue op over the same queue, as a way to transfer a tensor from TF to host (tensor tranfer in the other direction also works). To make this work, we need TFE_Context and the the TF_Session to use the same ResourceMgr object (attached to a Device, which is in turn owned by DeviceMgr), so that both can access the fifo queue resource op. PiperOrigin-RevId: 211471075
* | | | | | | | | Merge pull request #22015 from lc0:improve-examplesGravatar TensorFlower Gardener2018-09-04
|\ \ \ \ \ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211469413
* | | | | | | | | | Fix Split, convert kernel signature to use runtime shapes.Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211459453
* | | | | | | | | | Minor fix in the WALS estimator test, to pass the correct shape of the ↵Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | transposed matrix. PiperOrigin-RevId: 211453816
* | | | | | | | | | Internal change.Gravatar Dan Moldovan2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211450476
| | | | | * | | | | Merge branch 'master' into ma_easgdGravatar Xin Jin2018-09-04
| | | | | |\ \ \ \ \ | |_|_|_|_|/ / / / / |/| | | | | | | | |
* | | | | | | | | | Use LLD when building with downloaded GPU toolchainGravatar Ilya Biryukov2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This improves build times when using the downloaded clang toolchain. Additionally, remove '-B/usr/bin' flags from the cuda CROSSTOOL when using the downloaded toolchain. It was forcing 'clang' to first search for the linker in '/usr/bin', preventing downloaded LLD from being selected. PiperOrigin-RevId: 211430374
* | | | | | | | | | compat: Update forward compatibility horizon to 2018-09-04Gravatar A. Unique TensorFlower2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211423600
* | | | | | | | | | [XLA:GPU] Make sure the input for ThenMemcpy stays alive when it executesGravatar Benjamin Kramer2018-09-04
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | asynchronously By the time memcpy actually executes the source on stack would've gone out of scope. Also drop an unneeded BlockHostUntilDone from the autotune code. PiperOrigin-RevId: 211422876
* | | | | | | | | | Introduce auxiliary input and allow "cross-linking" in the bidirectional RNN Op.Gravatar A. Unique TensorFlower2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This introduces a connection between forward and backward cells across subsequent layers when stacking bidirectional RNN Ops on top of each other. In more detail: Previously, the Op had only one input that was fed into the layer in the following way: INPUT (INPUT_REVERSED) | | --------------------- | FW_RNN BW_RNN | <----- bidi-RNN cell (with one input / two outpus) --------------------- | | FW_OUT BW_OUT Now, the Op can have an (optional) auxiliary input in the following way: AUX_INPUT (AUX_INPUT_REVERSED) | | INPUT | (INPUT_R'D.)| | | | | ----------------------- | \ / \ / | | FW_RNN BW_RNN | <----- bidi-RNN cell (with 2 inputs / 2 outpus) ----------------------- | | FW_OUT BW_OUT When stacking these Ops, previously, only the following flow was allowed: Input / \ FW_RNN1 BW_RNN1 | | | | FW_RNN2 BW RNN2 | | | | FW_RNN3 BW_RNN3 \ / Output With the introduction of an auxiliary input to the bidi-RNN layer, the forward (FW_RNNi) output of the ith layer is fed into as the input to the next layer (hence, inputs to both FW_RNN{i+1} and BW_RNN{i+1}) and the backward output is fed as the auxiliary inputs to both FW_RNN{i+1} and BW_RNN{i+1}). This way, the stacking can be changed to allow for the "cross-linking" between subsequent layer in the following way: Input / \ FW_RNN1 BW_RNN1 | \ / | | / \ | FW_RNN2 BW RNN2 | \ / | | / \ | FW_RNN3 BW_RNN3 \ / Output PiperOrigin-RevId: 211401475
* | | | | | | | | | [TF:XLA] Bump open source llvm revision to r341305Gravatar Sanjoy Das2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211387503
* | | | | | | | | | Update unidirectional sequential LSTM to support state API.Gravatar A. Unique TensorFlower2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211378182
* | | | | | | | | | Update bidirectional sequential LSTM to support state API.Gravatar A. Unique TensorFlower2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211378028
* | | | | | | | | | Fix a test.Gravatar Shashi Shekhar2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211377977
* | | | | | | | | | [XLA:GPU] Flush out any pending work before starting autotuneGravatar Benjamin Kramer2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The autotune code assumes a clean slate, but there might be things from previous program executions still pending on the streams owned by the executor. Do a full host-device sync before autotuning to flush out any pending work. I'm still somewhat confused on how autotune can interfere with other buffers. There might be more things going wrong ... PiperOrigin-RevId: 211369162
* | | | | | | | | | Fix floating point ordering in the Sort HLO op in the GPU backend.Gravatar Adrian Kuegel2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | We use the same trick that is used in the TPU backend. PiperOrigin-RevId: 211344106
* | | | | | | | | | Call Cudnn also for grouped convolutions.Gravatar Adrian Kuegel2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Cudnn supports grouped convolutions, so we don't need the ConvolutionFeatureGroupConverter pass and can instead set the group_count parameter on the cudnn custom calls. PiperOrigin-RevId: 211339551
* | | | | | | | | | Update downloadable clang to r340427Gravatar Ilya Biryukov2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211339000
* | | | | | | | | | compat: Update forward compatibility horizon to 2018-09-03Gravatar A. Unique TensorFlower2018-09-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211323840
* | | | | | | | | | Rollforward of rollback:Gravatar A. Unique TensorFlower2018-09-02
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Reinstate the use of integral-exponent power function MathUtil::IPow, but make sure to use a floating point base, so as to compute the result using floating point arithmetic. This behaviour is equivalent to, but faster than, std::pow. Note that care must be taken to convert the base to double, which we effect by providing an explicit template type argument for MathUtil::IPow. PiperOrigin-RevId: 211290304
* | | | | | | | | | [XLA] Simplify effective scalar iota to zeroGravatar David Majnemer2018-09-02
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Happened to observe this come up in a linear algebra workload. PiperOrigin-RevId: 211290278
* | | | | | | | | | compat: Update forward compatibility horizon to 2018-09-02Gravatar A. Unique TensorFlower2018-09-02
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211257009
* | | | | | | | | | Merge pull request #20237 from Intel-tensorflow:nhasabni/gatherndGravatar TensorFlower Gardener2018-09-01
|\ \ \ \ \ \ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211226585
| | * | | | | | | | | Fix minor typosGravatar Sergii Khomenko2018-09-01
| |/ / / / / / / / / |/| | | | | | | | |
* | | | | | | | | | Only watch tensors on the current tape rather than all of them.Gravatar Tom Hennigan2018-09-01
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This allows fine grained control over recording in some cases, for example the following where we want d2y but not d2z: x1 = tf.Variable(2.0, trainable=False) x2 = tf.Variable(2.0, trainable=False) with tf.GradientTape() as tape1: with tf.GradientTape() as tape2: tape1.watch(x1) tape2.watch([x1, x2]) y = x1 ** 3 z = x2 ** 2 dy, dz = tape2.gradient([y, z], [x1, x2]) d2y, d2z = tape1.gradient([dy, dz], [x1, x2]) assert d2z is None PiperOrigin-RevId: 211206506
* | | | | | | | | | Automated rollback of commit 9e2ce8f4c483e68309a60dc89739bb1b79b4a12eGravatar A. Unique TensorFlower2018-09-01
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 211204708
* | | | | | | | | | [XLA] Remove remaining StringPiece references.Gravatar Benjamin Kramer2018-09-01
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | StringPiece and string_view are the same now, no need to convert between them. PiperOrigin-RevId: 211195959
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-09 07:37:29 +0000