aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
* Adding NodeDef names to error messages for better debuggability.Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | The format used is as follows: {{node <node_name>}} PiperOrigin-RevId: 206370355
* Use constant buffer allocations for XLA:CPUGravatar Sanjoy Das2018-07-27
| | | | | | | | This is simpler than the corresponding change to XLA:GPU because on XLA:CPU all instructions are codegened so we can always embed a pointer to the constant global variable directly in the generated LLVM IR. PiperOrigin-RevId: 206363887
* Always lock resource variables in training ops.Gravatar Alexandre Passos2018-07-27
| | | | PiperOrigin-RevId: 206362555
* Fix SavedModelEstimator docstring formatting.Gravatar Katherine Wu2018-07-27
| | | | PiperOrigin-RevId: 206361654
* Update functionality of --allow_nudging_weights_to_use_fast_gemm_kernel.Gravatar Nupur Garg2018-07-27
| | | | PiperOrigin-RevId: 206354203
* Update backward pass to save memory in graph mode.Gravatar Xuechen Li2018-07-27
| | | | PiperOrigin-RevId: 206352708
* [XLA] Use se:: rather than stream_executor:: in a few places.Gravatar Justin Lebar2018-07-27
| | | | | | No functional change. PiperOrigin-RevId: 206352602
* Support nested inputs for bidirectional_dynamic_rnn.Gravatar Rui Zhao2018-07-27
| | | | PiperOrigin-RevId: 206347779
* [XLA:GPU] Only add the cubin if it is availableGravatar Benjamin Kramer2018-07-27
| | | | | | | | It's only non-empty if we were able to run ptxas. If the PTX is going to be JIT'ed by the driver it won't be around. Loading an empty cubin will result in a fatal error. PiperOrigin-RevId: 206341931
* Merge pull request #20671 from yongtang:06052018-tan-complexGravatar TensorFlower Gardener2018-07-27
|\ | | | | | | PiperOrigin-RevId: 206341656
* | [XLA] Parse window and dim_labels that appear on a custom call.Gravatar Justin Lebar2018-07-27
| | | | | | | | | | | | XLA:GPU uses a custom-call with window/dim_labels to represent a call to cudnn. PiperOrigin-RevId: 206339219
* | Always use either kAllNCHW or kAllNHWC for GPU convolution layout assignment.Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | PiperOrigin-RevId: 206338966
* | Bug fixes and 16 bit matmul added in CXX11/src/FixedPoint.Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | PiperOrigin-RevId: 206335619
* | Quick FusedBatchNorm performance regression fix:Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | | | | | | | | | | | When not compiled with "--config=opt", or when compiling with "--config=opt --distinct_host_configuration=false" (to skip host-specific optimizations), the following code incurs casting overhead even when T == U, y.reshape(rest_by_depth).device(d) = x_shifted.template cast<T>(); The fix: explicitly avoid calling cast<T>() if T == U. PiperOrigin-RevId: 206332285
* | Improve restore performance for large checkpoints.Gravatar Russell Power2018-07-27
| | | | | | | | | | | | When loading large tensors, the cost of creating a new BundleReader is small relative to the load time for the Tensor. When reading from network storage, using a threadpool for large tensor loads allows us to push expensive operations (alloc, fetch, checksum) to separate cores. PiperOrigin-RevId: 206330021
* | Properly call PrepareToUpdateVariable in resource strided slice assign.Gravatar Alexandre Passos2018-07-27
| | | | | | | | PiperOrigin-RevId: 206327963
* | Add support for len calls on tensor lists.Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | PiperOrigin-RevId: 206325816
* | Merge pull request #20744 from npanpaliya:resize_bilinear_test-fix-ppc64le-20361Gravatar TensorFlower Gardener2018-07-27
|\ \ | | | | | | | | | PiperOrigin-RevId: 206325357
* \ \ Merge pull request #20711 from SneakyFish5:copied-buffer-fixGravatar TensorFlower Gardener2018-07-27
|\ \ \ | | | | | | | | | | | | PiperOrigin-RevId: 206323345
* | | | Fix typo in FileWriter docs.Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206320196
* | | | Adding core estimator for ranking.Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206318440
* | | | Add parameter server distribution.Gravatar Yuefeng Zhou2018-07-27
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206289143
* | | | Automated rollback of commit d4cb01f242dc3ff0f7b0aae7284def46281755f2Gravatar A. Unique TensorFlower2018-07-27
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206281287
* | | | Set the correct context when calling cudnnCreate.Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | When running with multiple devices, using the wrong context will lead to a check-fail when trying to set a stream that has been created with a different context. This resolves a check-fail on resnet50 with 8 GPUs. PiperOrigin-RevId: 206274741
* | | | Have the SavedModel loader use Session's Make/Run/ReleaseCallable() API ↵Gravatar Christopher Olston2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | instead of Run(), to avoid leaving behind non-GC'ed state after model initialization. PiperOrigin-RevId: 206266841
* | | | SpaceToBatchND should pad with zero_point when inference_type is uint8Gravatar Suharsh Sivakumar2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206265356
* | | | Layers/Networks now have a default implementation for static shape inference ↵Gravatar Francois Chollet2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | in eager execution (when possible). Also move `build` implementation for subclassed networks from Model to Network (where it belongs) and slightly refactor it to minimize code duplication. PiperOrigin-RevId: 206260286
* | | | Add build option to exclude TensorRT from cuda build.Gravatar Guangda Lai2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206252639
* | | | Simulate eager variable resoration in tf.keras.Model.load_weights when graph ↵Gravatar Allen Lavoie2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | building Previously, the first Model build after load_weights (e.g. a predict()) would trigger restore ops, and any variables added later (e.g. slot variables from an added optimizer) would not be restored when graph building. This change makes behavior consistent between eager execution and graph building by running new restore ops as they come in. PiperOrigin-RevId: 206251879
* | | | Add low and high as properties to quantized distribution.Gravatar Dustin Tran2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206249977
* | | | Update backward pass to save memory in eager mode.Gravatar Xuechen Li2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206249965
* | | | Use `SaveableObject.restore` in init_from_checkpoint instead of `assign`.Gravatar Priya Gupta2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206245967
* | | | Implement constant buffer allocation for XLA:GPUGravatar Sanjoy Das2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This CL teaches XLA:GPU to use "normal" buffer assignment for constant instructions. Constant instructions are mapped to a BufferAllocation, like all other instructions, except the storage for this buffer is allocated statically as a global in the generated PTX. This CL does not change how we access the constants -- in IrEmitterUnnested::BuildKernelThunk (used for top level computations) and in HloToIrBindings::EmitBasePointersForHlos (used for nested computations) we bind the kConstant instructions to the llvm::GlobalVariable backing them. So users of constant instructions still access the globals corresponding to the constants directly. However, we no longer emit the constant literals inline. Instead we emit a constant with a zero initializer and then memcpy in the contents of the literal when we load the CUBIN/PTX. This works around compile time issues in LLVM and ptxas caused by large constants. We also populate `BufferAllocations` with the device pointers for the constant globals. This is at least needed for TupleThunk today because TupleThunk wants the addresses for the sub-buffers on the host. I'm not sure if there are other places in XLA:GPU that rely on there being an entry in BufferAllocations for every BufferAllocation. PiperOrigin-RevId: 206243319
* | | | Fix: When sample_weight_mode is a list/dict set default sample_weight values ↵Gravatar Pavithra Vijay2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | so that we do not require sample_weight to be set during training/eval PiperOrigin-RevId: 206242625
* | | | Remove tensorflow/opensource_only/Core as it is not used by tensorflow.Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206240947
* | | | [TF:XLA] Align the two implementations of CanShareOperandBufferWithUser.Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Eventually (when TuplePointsToAnalysis is removed), there will be only one implementation left. Also, use early return instead of else-if to make the code less indented. PiperOrigin-RevId: 206240067
* | | | Avoid using cuda_fp16.h for fp16 definition.Gravatar Gunhan Gulsoy2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206238991
* | | | [TF:XLA] Remove unused parameter.Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206237934
* | | | [tf.data] Switch `batch_and_drop_remainder` to use the fused op.Gravatar Derek Murray2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | This change also switches `padded_batch_and_drop_remainder` to use the corresponding fused op. PiperOrigin-RevId: 206236616
* | | | Don't make remote copy call when both send/recv devices are the same.Gravatar Akshay Modi2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206236233
* | | | Increase the input dimension size from 4 to 6 to address the RetinaNet modelGravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206235660
* | | | Keras Tensorboard Callback - enable metrics logging in Eager modeGravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206235264
* | | | Test tf.constant directly (still doesn't work with scalars)Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206224062
* | | | This change adds an attribute to the _HostCast op to have the same interface ↵Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | as the Cast op. PiperOrigin-RevId: 206218592
* | | | [XLA] Fix BF16 propagation type adjustmentGravatar Yuanzhong Xu2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | AllUsersConsumeBF16() incorrectly used ValueTypeAfterChange() for the current value being checked, but it should be the original type. Also fusion computation should be adjusted as soon as the fusion root is adjusted. There was also redundant work for while computations. Now removed. PiperOrigin-RevId: 206216822
* | | | [XLA:GPU] Don't fuse random things into fusions that become a cublas callGravatar Benjamin Kramer2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | Only transpose and broadcast are valid. I think this used to work because we didn't emit cublas calls for fused dots until recently. PiperOrigin-RevId: 206213730
* | | | Full set of unittests for the Eager delegate.Gravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206211243
* | | | ResourceVariables shouldn't need twice the memory when initializing.Gravatar Alexandre Passos2018-07-26
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is safe because all ops which write to resource variables check whether there are other outstanding references to the buffer and copy if that's the case. So we can safely reuse the buffer of initializer tensors even in weird cases such as initializing from a constant (which should never be mutated) or using the same tensor to initialize multiple variables. PiperOrigin-RevId: 206211065
* | | | Delete unused code in CXX11/src/NeuralNetworksGravatar A. Unique TensorFlower2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206209252
* | | | Restore tower local variables correctly in init_from_checkpoint.Gravatar Priya Gupta2018-07-26
| | | | | | | | | | | | | | | | PiperOrigin-RevId: 206208637
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-01 15:22:21 +0000