hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
a97601363f2c23b655c1a9679d88b564170159cb | Remove unused StreamExecutor methods.
PiperOrigin-RevId: 560832851 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,261,080,000 |
ff2a008cee079bc6e904d02931ba8d8dded04495 | Remove unused TpuExecutor methods.
PiperOrigin-RevId: 560834283 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,261,412,000 |
4e3f411efc575c618503569cf2b4daf45f92ace9 | Update TopkRewriter to handle tensors with rank > 2
Just add reshapes before and after if necessary.
PiperOrigin-RevId: 560834408 | Kevin Chen | kevinch@google.com | 1,693,261,448,000 |
e5cbfdebf0a94e36a0f5e3d6a3c7e9300532f3c5 | [tosa] Add tf erf to tosa legalization
Add tf to tosa erf legalization, and a test case.
Signed-off-by: Tai Ly <tai.ly@arm.com>
Change-Id: I75e419757942bb40469a5990fc6d86dfbe65812e | Tai Ly | tai.ly@arm.com | 1,692,821,554,000 |
020f07be427ee45505ab38e3960266f7b408e900 | Move NoneTensor and NoneTensorSpec into framework.
PiperOrigin-RevId: 560843952 | Fiona Lang | flang@google.com | 1,693,263,758,000 |
b853ca433151dfe73e3ee13e9334b89d380d5fa1 | [Tosa] Fix legalization of BroadcastTo
fix legalization of BroadcastTo by:
- for floats: add with -0.0 (instead of 0.0)
- for integer types other than I32, insert cast to I32, before adding zero,
then cast back to output type
fixed lit tests for above changes as well
Signed-off-by: Tai Ly <tai.ly@arm.com>
Cha... | Tai Ly | tai.ly@arm.com | 1,691,701,248,000 |
214f2d7e14a933587b8d768e1e41ae7a606e5cff | Add escape hatch for users who do not want to enable coordination service by default.
PiperOrigin-RevId: 560849974 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,265,252,000 |
deac401f45ea48ac9eb0d6fcf737fd5886da4993 | Remove unused TpuMemoryLimit method.
PiperOrigin-RevId: 560855873 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,266,829,000 |
4eb9a591e07b9aa6638a1f7d3c992eb8c1d9b9ae | Better support for reversed sharding strategies (generated due to the presence of the kReverse HLO op). This implies handling some cases where we coudl not previosuly compute resharding costs.
PiperOrigin-RevId: 560857276 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,267,237,000 |
88aa9f0092dc521d317f5d45a116d1818317f0ea | Improve the documentation of the `has_rank` field in the Tensor object.
PiperOrigin-RevId: 560859240 | Jared Junyoung Lim | jaredlim@google.com | 1,693,267,784,000 |
0dd7ed1e3b65f58067789fef4c2352b7ebcd8e07 | Change `unfold_batch_matmul`'s default value to false in tfl_pass_config to be consistent with python.
PiperOrigin-RevId: 560859910 | Haoliang Zhang | haoliang@google.com | 1,693,267,989,000 |
fb5d72325a40f2fef3c6bed0e8a1eac7d667107b | Remove tfl-standard-pipeline since it's not used by anyone.
PiperOrigin-RevId: 560866971 | Haoliang Zhang | haoliang@google.com | 1,693,270,077,000 |
17b3a838ad1924638f292c42938c949a0ce3bd3b | Add MatMul/Dot support to quantized_function_library uniform quantized.
Added quantize_matmul_{fn, with_bias_and_relu(6), with_relu(6)_fn}.
PiperOrigin-RevId: 560871764 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,271,343,000 |
fa8459d72eaf7de504de5ace1de4c7a5417dbef8 | Adds EdgeTPU Model Info
PiperOrigin-RevId: 560873460 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,271,796,000 |
d7516276d068b3ad11108532a67784c336dba359 | Add xprof traceme producer and consumer for CollectiveGatherV2 op
PiperOrigin-RevId: 560874391 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,272,175,000 |
2cbebc0797d7b59c609e2ae5422f97b0c32f16ca | Add a catchall case for estimating resharding costs by fully materializing the replicated tensor before resharding to the final target sharding. This is an overestimation, and should handle most cases.
PiperOrigin-RevId: 560877637 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,273,550,000 |
3d9c899f380b006e92c4c2d605212759751ecb05 | Integrate LLVM at llvm/llvm-project@80e5b46e4523
Updates LLVM usage to match
[80e5b46e4523](https://github.com/llvm/llvm-project/commit/80e5b46e4523)
PiperOrigin-RevId: 560886207 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,277,197,000 |
408eb11adf89cca5d61f0e10f07d90e8fa470f44 | Factor out the quantized operand pattern from `ComposeUniformQuantizedDotGeneralOpWithTwoQuantizedActivations` for clarity.
Creates a new `MatchQuantizedOperand` method for matching this pattern. It could be used for both the lhs and rhs operands of `stablehlo.dot_general` when both operands are quantized.
PiperOrigi... | Dan Suh | dansuh@google.com | 1,693,278,331,000 |
a1c5ca040224da7cee47cc1d99b35252501e85d7 | Add capability to read human readable sharding.
PiperOrigin-RevId: 560900997 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,282,187,000 |
f6ecbdcb94ae4db19163296f62230eb6e06119d7 | Fix embedding ops test for TF1 mode | Ben Barsdell | bbarsdell@nvidia.com | 1,693,282,759,000 |
ba80bde1dcde852a5a808381b79e842241dfe822 | Add missing import deps for embedding lookup tests | Ben Barsdell | bbarsdell@nvidia.com | 1,693,282,876,000 |
fe6c1072e392e2ea2682e9d9d20f30d3e2d70c3a | Bump compat horizon for new embedding ops | Ben Barsdell | bbarsdell@nvidia.com | 1,693,283,037,000 |
82fd8da6440a19870d6f6ec38622998a9583a178 | Internal Code Change
PiperOrigin-RevId: 560907781 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,284,812,000 |
509b907564c8c2f8d6f32025e60c447ae81c44c5 | Pass the control address into CreateWorkerStreamInterface().
PiperOrigin-RevId: 560907989 | Kuangyuan Chen | chky@google.com | 1,693,284,887,000 |
1558995f6541c0f58a12a774644cd6d88ce7e374 | Internal Code Change
PiperOrigin-RevId: 560908572 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,284,986,000 |
dae07ac18292d73a8a8ad1d61abb9ad2b79d7fa4 | Internal Code Change
PiperOrigin-RevId: 560911385 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,285,666,000 |
01071fa1b8d26b0aad8da75793e045ad826316d8 | Internal Code Change
PiperOrigin-RevId: 560915655 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,286,921,000 |
d693e62b3f3f4ee4a92a15ccac04a788cbafd56b | Internal Code Change
PiperOrigin-RevId: 560919872 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,288,035,000 |
5dd082dce85b4e05beed75a9b020b4900a5c05f8 | [xla:gpu2] Remove CUDA graph support via custom modules
We are targeting CUDA HAL to automatically build graphs from command buffers.
PiperOrigin-RevId: 560929220 | Eugene Zhulenev | ezhulenev@google.com | 1,693,290,508,000 |
aa6c5a8b65e74366eccf41fc56c4a24b07ddf49b | lite:stablehlo: add serialization for stablehlo ops: iota, compare, dynamic_slice, dynamic_update_slice, pad
PiperOrigin-RevId: 560929493 | Zichuan Wei | zichuanwei@google.com | 1,693,290,590,000 |
941af5871738642aab574585d8fb2ed72441fd57 | Add tests and cleanup for multi-platform support in XlaCallModule.
XlaCallModule already had support for multi-platform modules, even
though JAX serialization does not yet export such modules. Here we
add a test for combining multi-platform with shape polymorphism.
We also clean up the implementation. Previously the ... | George Necula | necula@google.com | 1,693,292,203,000 |
df536b6ddaf74419c1db414382f248a9b6b42893 | Internal Code Change
PiperOrigin-RevId: 560937374 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,292,892,000 |
3156d031721699fd524fb1ae807cc5ba5a93f329 | Update rules_python version to 0.25.0 | Namrata Bhave | Namrata.Bhave@ibm.com | 1,693,296,797,000 |
fedaa2a5d32f454dd3608e496bc30609dd82327f | [xla:gpu] Replace GpuDeviceInfo::name with compute_capability and serialize it in proto as well.
PiperOrigin-RevId: 560954035 | Christian Sigg | csigg@google.com | 1,693,297,831,000 |
065b09eac8e42e59e04db425355c6a2eaaf567a8 | compat: Update forward compatibility horizon to 2023-08-29
PiperOrigin-RevId: 560960506 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,299,768,000 |
2b7576949f130d39cbecc5e351287f6d3aa49007 | Update GraphDef version to 1603.
PiperOrigin-RevId: 560960598 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,299,789,000 |
257db693972d87bf5ff71b013f7585c8a5125e47 | Add missing headers for calibration_reader.h/.cc.
PiperOrigin-RevId: 560962713 | Dan Suh | dansuh@google.com | 1,693,300,374,000 |
323e5b42fbd83b34e4f2ea094865dfc46b9ac09b | Merge remote-tracking branch 'origin/master' into toupstream/int16_8_rsqrt | Saoirse Stewart | saoirse.stewart@arm.com | 1,692,782,420,000 |
7d5d6eae811969571c6866bc5e1eda8edc11022f | Merge remote-tracking branch 'origin/master' into toupstream/int6x8_32_accum | Saoirse Stewart | saoirse.stewart@arm.com | 1,693,302,190,000 |
fe4ba75638b5cb1b6c8d9ade7d17001a0feef015 | Update Transpose_conv_test | Saoirse Stewart | saoirse.stewart@arm.com | 1,693,302,619,000 |
81759529910dd9270c0c51836b74f04de1277f56 | Update versions | Saoirse Stewart | saoirse.stewart@arm.com | 1,693,303,100,000 |
fc4d12e3782c0b2440a54e90648a73e67da9f1af | Pass calibration options to CalibrationStatisticsCollector.
PiperOrigin-RevId: 560971107 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,302,993,000 |
7a619e53db699fe8340fd1800d9b200ed7b9d932 | Internal Code Change
PiperOrigin-RevId: 560972715 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,303,469,000 |
09f2f3991c6dfea5abb0e8e255345b835d6beb64 | Update version | Saoirse Stewart | saoirse.stewart@arm.com | 1,693,304,050,000 |
306a58ad46058e995babf691524d54ab4cafb347 | Fix typo | Yuriy Chernyshov | thegeorg@yandex-team.com | 1,693,305,411,000 |
3dd63788c8878f14ac6a47626a7b6a81df92357a | Integrate LLVM at llvm/llvm-project@c1b5674fbb76
Updates LLVM usage to match
[c1b5674fbb76](https://github.com/llvm/llvm-project/commit/c1b5674fbb76)
PiperOrigin-RevId: 560981107 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,306,159,000 |
7480158fcd1e5cafa36ff2fa5638168de2972249 | Add comment in c/c_api_types.h pointing to core/c/c_api_types.h for docs,
similar to the comment in c/c_api.h that points to core/c/c_api.h.
PiperOrigin-RevId: 560995155 | Fergus Henderson | fergus@google.com | 1,693,310,716,000 |
f9a07f93c9a361a3b93b385290d8768b59cde648 | Define macro TFLITE_CONDITIONAL_NAMESPACE, for use in code using cc_library_with_tflite.
PiperOrigin-RevId: 560995178 | Fergus Henderson | fergus@google.com | 1,693,310,722,000 |
9fec6572e0f4ed2fb4eb98ef5d3eb131c7b5285f | Avoid duplicating documentation between c/common.h and core/c/common.h.
Instead, add comment in c/common.h pointing to core/c/common.h for docs,
similar to the comment in c/c_api.h that points to core/c/c_api.h.
PiperOrigin-RevId: 560996814 | Fergus Henderson | fergus@google.com | 1,693,311,275,000 |
ba6148ad9382a6eb841d8377f5828026988f230b | [XLA:GPU] Dump TritonIR when dumping is requested
PiperOrigin-RevId: 560999729 | George Karpenkov | cheshire@google.com | 1,693,312,277,000 |
7957fe00ee87ca412c793fffe1e3312c4cfe35af | Set default values for `mhlo.convolution` in `legalize_hlo.cc` when not set
PiperOrigin-RevId: 561001028 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,312,746,000 |
00122c2341466bc1af8aab6e7f0cb607829a40bd | Update pthreadpool dependency
PiperOrigin-RevId: 561007153 | Zhi An Ng | zhin@google.com | 1,693,314,413,000 |
3d9ee92b478316d0f3a6ac7432300a9d3a85342d | reference int4->float32 cast
PiperOrigin-RevId: 561010877 | Alan Kelly | alankelly@google.com | 1,693,315,398,000 |
4601e74cdcd86e6d229ec1d98ff08ba8e147c092 | PR #5345: [ROCm] Fix typo from #5050
Imported from GitHub PR https://github.com/openxla/xla/pull/5345
Copybara import of the project:
--
f45a2638fe07ddcc70e1f2b8c3e7f181f9b7251e by Dragan Mladjenovic <Dragan.Mladjenovic@amd.com>:
[ROCm] Fix typ from #5050
Merging this change closes #5345
PiperOrigin-RevId: 561018... | Dragan Mladjenovic | Dragan.Mladjenovic@amd.com | 1,693,317,781,000 |
1cfb1163393994c1b68357a14934a5d0e6f5661f | [xla] Add a flag to enable the runtime optimization when acquiring NCCL
cliques.
Add xla_gpu_enable_clique_optimization, and use it in the current time to allow
skipping rendezvous when acquiring NCCL cliques.
PiperOrigin-RevId: 561039560 | Bixia Zheng | bixia@google.com | 1,693,323,602,000 |
3530d58aab7a46642b39d0bee07a9ec0d48a6f6b | [XLA:GPU][NFC] Make some pipelines free functions and move them into dedicated
files for reuse.
This is in preparation of further autotuning-related changes that need to use
these pipelines.
PiperOrigin-RevId: 561039877 | Benjamin Chetioui | bchetioui@google.com | 1,693,323,657,000 |
e3d3e94de1879db79de65df2868bb581fbc198b8 | adding missing file schannel_int.h | justkw | karen.wu@intel.com | 1,693,324,258,000 |
6158b870fe0f1538dfe733b86f36636d8cccac1e | [NFC] [XLA:GPU] Provide a ToString() method and logging for TritonFusionAnalysis
Sample output:
```
ir_emitter_triton.cc:755] TritonFusionAnalysis{
LHS: IterSpec{parameter_0.1: {0: [{stride=128, count=480, subfragments=[480]}], 1: [{stride=1, count=128, subfragments=[128]}]}}
RHS: IterSpec{parameter_1.1: {1: [{stride... | George Karpenkov | cheshire@google.com | 1,693,326,384,000 |
0c6c233ea54d6485e2e91bc7c7492154e2edfac7 | Add name to glob_lit_tests rules.
PiperOrigin-RevId: 561055294 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,326,974,000 |
95e2519046c8c14893e8a09c13b82c6ca72bd34b | Update generated pyi for checkpoint reader.
PiperOrigin-RevId: 561061847 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,328,379,000 |
1c7f8bcef378fa03100471705dc9fb77789b33b5 | Support degenerate case of mul operation.
PiperOrigin-RevId: 561062490 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,328,515,000 |
50e4cdbf1c8e1a8c765932dd40cc1742972aa6f6 | PR #5340: [ROCm] enable more ROCm tests
Imported from GitHub PR https://github.com/openxla/xla/pull/5340
We found out 4 tests that are disable on ROCm for a long time that no longer valid
1. gpu_unrolling_test
2. convolution_test_cudnn_frontend_disabled
3. batch_normalization_test
4. hlo_runner_main_build_test
@aku... | Chao | cchen104@amd.com | 1,693,329,545,000 |
cfe0163445a76ad5c9bf670abb596dbbca560844 | add patch back as Penporn brought back PR https://github.com/tensorflow/tensorflow/pull/61237 | guozhong.zhuang | guozhong.zhuang@intel.com | 1,693,331,400,000 |
ada4c5c94c6bc862702261925eea30bb19e83cec | Update generated pyi for checkpoint reader.
PiperOrigin-RevId: 561088502 | Brian Wieder | bwieder@google.com | 1,693,333,420,000 |
0ce15632f885f8811d0b3ee9e2a05854ae9f04ea | Add a method to return all TF devices on a given platform.
PiperOrigin-RevId: 561089543 | Changhui Lin | changhuilin@google.com | 1,693,333,595,000 |
9e5fec042ec0311a3977493d6bea2522e1354556 | [XNNPack] Add support of transient indirection buffer for RESIZE_BILINEAR op
PiperOrigin-RevId: 561091268 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,333,969,000 |
116ff4755c4245beb27ca1a62fb363c7cabf8425 | power metric specific proto
PiperOrigin-RevId: 561091977 | Jie Sun | jiesun@google.com | 1,693,334,137,000 |
2073cad41ec3f5228a09f40bfba61641657c99c8 | lite: stablehlo: add serialization of stablehlo ops: dot_general, reduce_window, sort, while
PiperOrigin-RevId: 561092607 | Zichuan Wei | zichuanwei@google.com | 1,693,334,274,000 |
ca12a50af36f6a9d1d5f7072f6cc9a74a67f55b0 | Create a new pass to fuse FC into BatchMatmul and only enable it when unfolding batch matmul is set to false. Previously, the `FuseTransposeFCLhsToBatchMatmul` and `FuseTransposeFCRhsToBatchMatmul` is running inside optimize pass, which will introduce BatchMatmul op even if user doesn't want to run it on their backend.... | Haoliang Zhang | haoliang@google.com | 1,693,336,770,000 |
5cfd647364306d2a84c8a344895eec96096ed93f | #tf-data Turn `VLOG(1)` into `LOG(INFO)` to let user know that `tf.data` is saving symbolic checkpoints. This info line is printed once for each checkpoint.
PiperOrigin-RevId: 561108445 | Wilsin Gosti | wilsin@google.com | 1,693,337,948,000 |
84b54f07cb51125d92e66e67daa5342f36a65db1 | Add FunctionType to public API
PiperOrigin-RevId: 561111836 | Faizan Muhammad | fmuham@google.com | 1,693,338,766,000 |
98150bb26a4db1d170433004aaa2e5f9f06b79a0 | Add traceme for both async start and end in TF executor
PiperOrigin-RevId: 561114720 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,339,389,000 |
50dd11b049b3cc3a0e3364c5e9ebb1003520af74 | [XLA:GPU] Fix broken build under gcc
JAX CI started failing with:
```
external/xla/xla/service/gpu/fusion_pipeline.cc:85:10: error: use of deleted function 'xla::HloPassPipeline::HloPassPipeline(const xla::HloPassPipeline&)'
85 | return fusion;
| ^~~~~~
In file included from external/xla/xla/servi... | Peter Hawkins | phawkins@google.com | 1,693,341,225,000 |
b2e2a83e32eeb7fa0232f6fc808496e763e7e30e | Fix wrong error message in `legalize_hlo.cc`
PiperOrigin-RevId: 561126894 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,341,970,000 |
8d3bdb4c32e39768b66e82cefe61a8a0c549e506 | Update experimental_convert_saved_model_v1
Update experimental_convert_saved_model_v1 typo to experimental_convert_saved_model_v1_to_mlir | Sachin Prasad | sachinprasad@google.com | 1,693,343,474,000 |
7df4bd5b06f94104e7e42fecc6887b5f7441b1b0 | Mlir module deserialization for AoT Compilation
PiperOrigin-RevId: 561134253 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,343,461,000 |
26612c8623c828c786dcebaa3151c5ea6d9ae0ec | Fix typo and relative path, Same as https://github.com/tensorflow/tensorflow/pull/59974
PiperOrigin-RevId: 561134835 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,343,577,000 |
95abdf8dfaa1c2e71de6a2e1474220846051fcf9 | Rollback of: Update TopkRewriter to handle tensors with rank > 2
Broke argsort on gpu.
PiperOrigin-RevId: 561136138 | Kevin Chen | kevinch@google.com | 1,693,343,829,000 |
86fedfbf19fa22c76e93ffcbf4b16251e4b9da35 | Set default values for `mhlo.convolution` in `legalize_hlo.cc` when not set
PiperOrigin-RevId: 561137986 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,344,240,000 |
29bb1223719f0f3573cacb8b50bc944a03b424d7 | Remove unused PopulateExecutionProfile method.
PiperOrigin-RevId: 561141352 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,344,987,000 |
f167a3532d6c51ab4224b9a3811e1026533020d4 | Add the FP8 ops. | Kulin Seth | kulin_seth@apple.com | 1,693,345,663,000 |
840633cb23ae111075999514484fd5416f0cef83 | [PJRT:C] Remove xla::Shape getter and related logic from PJRT C API
This changes removes `PJRT_Buffer_OnDeviceTrimmedShape` and related
functionality, since passing around xla::Shapes is expensive and often
includes more information than is necessary or even meaningful, and we
now have more specific getters that shoul... | Skye Wanderman-Milne | skyewm@google.com | 1,693,346,551,000 |
cb598c672fc31c6c48245dabfaff786b0cb51f35 | Remove proto_splitter_public, add platform logic to internal_api read.
PiperOrigin-RevId: 561148975 | Adam Cogdell | adamcogdell@google.com | 1,693,346,671,000 |
9b9fd1be990c3079008782bd49260f9fc290d275 | Remove unused Stream methods.
PiperOrigin-RevId: 561150086 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,346,930,000 |
bfee166bc68730e352b467c2e29637bc9f2d60a4 | Integrate StableHLO at openxla/stablehlo@7111faf
PiperOrigin-RevId: 561161226 | Sandeep Dasgupta | sdasgup@google.com | 1,693,349,566,000 |
2106dee0c7691c8513804b86486022e1c2527041 | Temporarily disable LaxBackedNumpyTests.testSortKeyValue_(2,3,4)_float32_bool_1 to pass TF nightly build
PiperOrigin-RevId: 561163175 | Tianrun Li | tianrun@google.com | 1,693,350,055,000 |
923c5311eea7b825a11ef3d545c9e371aaf9cee2 | [PJRT:C] Small follow-ups to PJRT_Buffer_Memory
* Rearrange PJRT_Api struct before major version bump
* Return unimplemented instead of nullptr
PiperOrigin-RevId: 561165318 | Skye Wanderman-Milne | skyewm@google.com | 1,693,350,515,000 |
a99d8cda8d96bf6a2dc7eb1bdfd4ef4116eda61d | Implementation for flip (left to right) + tests.
PiperOrigin-RevId: 561168251 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,351,221,000 |
2c60d3b79e72bdc6cf8ebf5b9799a0e04c29b6f0 | [JAX] Implement importing external dlpack-aware Python arrays.
See https://dmlc.github.io/dlpack/latest/python_spec.html.
This is the import path. The export path was implemented in
https://github.com/openxla/xla/commit/0b3cbfe4bc7cd68dc20924bd878cdfb2faa1a169.
This allows for creating jax.Arrays from external GPU a... | Skye Wanderman-Milne | skyewm@google.com | 1,693,352,314,000 |
f1a3492a6b0a0b30d815329befe1b3bc50b87c62 | Move proto_splitter save tests to separate file.
PiperOrigin-RevId: 561178670 | Adam Cogdell | adamcogdell@google.com | 1,693,353,891,000 |
e78cc1da3ebf7adbf8121afe56cef6a970cde2f9 | DataRef implementation to wrap tflite tensors. This lets us interop cleanly with the tflite memory system within out functions.
PiperOrigin-RevId: 561180210 | Luke Boyer | lukeboyer@google.com | 1,693,354,296,000 |
2f77608735cd26647ecf6f97ccc68a952df8ee24 | Temporarily disable testTxtWithEinsum.
PiperOrigin-RevId: 561183306 | Haoliang Zhang | haoliang@google.com | 1,693,355,194,000 |
8c7a63c76a23d1458446c0329b3942d9082b0633 | No public description
PiperOrigin-RevId: 561183522 | Fiona Lang | flang@google.com | 1,693,355,271,000 |
ce0c7426519ef46c80fd69b8cac971fbd58457d1 | Implementation for flip (up to down) + tests.
PiperOrigin-RevId: 561184976 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,355,729,000 |
763bb57686ec81636bf8f06c716b1ded8abe0942 | [tf.data] Move `warm_start` option from `OptimizationOptions` to top level `Options`.
PiperOrigin-RevId: 561188743 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,357,063,000 |
d9dd711248bb6f5e16d3a7f7f27bf54806b427ad | Merge pull request #61704 from Intel-tensorflow:karen/curl-upgrade
PiperOrigin-RevId: 561192100 | TensorFlower Gardener | gardener@tensorflow.org | 1,693,358,453,000 |
8c0d21c16d599a53b3f29e3b513c588fda77aaf1 | Integrate LLVM at llvm/llvm-project@96e83d3705c7
Updates LLVM usage to match
[96e83d3705c7](https://github.com/llvm/llvm-project/commit/96e83d3705c7)
PiperOrigin-RevId: 561194584 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,359,103,000 |
d4673148ea888f16e460bd75db74e655bebc8d4c | Add LowerQuantizedPass back
PiperOrigin-RevId: 561195486 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,359,430,000 |
315e01759f70a4db7ac3a8ba260b0fdb7db458c2 | Re-enable test, update the test to check for FCs insteads of BatchMatmul.
PiperOrigin-RevId: 561205731 | Haoliang Zhang | haoliang@google.com | 1,693,363,709,000 |
70d65c5177c089e1704ebcb6a7abb8b029ac64cc | Validate functions and flags naming.
PiperOrigin-RevId: 561206751 | A. Unique TensorFlower | gardener@tensorflow.org | 1,693,364,183,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.