hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
3554f41b2dca528a32e160ceb2814add6fe9eb8e | Fix tflite-to-tosa convolution intermediate qtype
The intermediate quantized type can only be inferred from
the bias type if it increases the bit depth as at minimum
we need `i32` intermediate types.
PiperOrigin-RevId: 563477896 | Robert Suderman | suderman@google.com | 1,694,108,456,000 |
0caae80adf169e0106946b2d92faad362854f220 | Update build rules to use Starlark-based repository rule | Terry Heo | terryheo@google.com | 1,694,037,089,000 |
d03ec72ddfa3bf3f3931fe7c9ff31a91ed0de52d | Check in generated pyi files for some py_extension targets.
PiperOrigin-RevId: 563478607 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,108,576,000 |
ca11d8d89366943ac6881badbf22fb64d976a38f | Fix the delattr method of TF modules
PiperOrigin-RevId: 563482518 | Jiawei Xia | jiaweix@google.com | 1,694,109,265,000 |
08b3321c48bd71cc0e69be8d1b66a5b3285c1b84 | Make TFG GraphDef import/export visible to the inference converter V2
PiperOrigin-RevId: 563487618 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,110,152,000 |
daf3c0454a74611dcf83e71a0be6c83e841f17f4 | multiple host session should NOT combine the peak bw in PerfEnv.
which have one entry per memory space.
PiperOrigin-RevId: 563487677 | Jie Sun | jiesun@google.com | 1,694,110,165,000 |
318875ded0e7062dba06eff2c3519f54cb542591 | Optionally support partitioning for constant initializers. This is particularly needed for TPUEmbeddingV2 class which can shard variables across multiple cores.
PiperOrigin-RevId: 563505377 | Isha Arkatkar | ishark@google.com | 1,694,113,647,000 |
5ff6f74c8db9d93e6870d7b48f1817f5613fa6be | Allow disabling ACD per tf.function
PiperOrigin-RevId: 563508006 | Faizan Muhammad | fmuham@google.com | 1,694,114,231,000 |
a32e1d2e4b678faf5801b3569ff02006e3277685 | Check in generated pyi files for some py_extension targets.
PiperOrigin-RevId: 563514688 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,115,640,000 |
450f4c3c1c3be0c70b9b22a67a202f529f508cf4 | Check in generated pyi files for some py_extension targets.
PiperOrigin-RevId: 563514772 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,115,657,000 |
56345842f3d45f8f2507b4f4ba10383c326be1f0 | PR #5488: Add information to existing error.
Imported from GitHub PR https://github.com/openxla/xla/pull/5488
Copybara import of the project:
--
06dc621de14b233823f78ca95ef103b8d08aae11 by Frederic Bastien <fbastien@nvidia.com>:
Add information to existing error.
Merging this change closes #5488
PiperOrigin-RevId... | Frédéric Bastien | frederic.bastien@gmail.com | 1,694,116,946,000 |
22a2143b2e1006af085bd65740c5299ff6294f03 | Use FunctionType for printing ConcreteFunction signature
PiperOrigin-RevId: 563529140 | Faizan Muhammad | fmuham@google.com | 1,694,118,606,000 |
0281aa1e5665836ab0baa59360dc392fc398a659 | Use tf2xla implementation of StridedSlice
PiperOrigin-RevId: 563529234 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,118,624,000 |
5d4fe8a712cbe1cce1d8af663ab5ac54fcd42baf | Implement mhlo dot_general unpacking.
PiperOrigin-RevId: 563530895 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,118,971,000 |
6d115e7b900d1629980320d56b58c453e959cb61 | Delete perftools::gputools namespace alias now that all uses are gone.
PiperOrigin-RevId: 563545627 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,121,989,000 |
cf3cabba7fa96566036413113292c9eaa963ba6e | Allow create_mesh and create_distributed_mesh to take dict for mesh_dims
PiperOrigin-RevId: 563548183 | Yu Feng | feyu@google.com | 1,694,122,569,000 |
eb1cbc22b2205036c3474996ff65daaec9909c06 | Fix erroneous @tsl//tensorflow/tsl paths
PiperOrigin-RevId: 563550115 | Jake Harmon | jakeharmon@google.com | 1,694,123,015,000 |
e19f50687abaf19671ec0be4f25568fb72dd5b8d | Merge pull request #61583 from halseycamilla:master
PiperOrigin-RevId: 563552138 | TensorFlower Gardener | gardener@tensorflow.org | 1,694,123,711,000 |
24b54dcd5120334252d39bda47189755651d96df | Update requirements for TF-related projects to 2.15 versions. | Marc Fisher II | fisherii@google.com | 1,692,985,217,000 |
6eac7a68a281c1564b0346ad506490b30a9e0c08 | Update requirements. | Marc Fisher II | fisherii@google.com | 1,693,582,182,000 |
07051d554f60c632833d886d4a8cd83b488d04ef | Integrate StableHLO at openxla/stablehlo@77a59815
PiperOrigin-RevId: 563562910 | Kevin Gleason | gleasonk@google.com | 1,694,125,866,000 |
2576094a3acaee1dcca2e9b748da1a988fca8f76 | Fix byte_size adjustment for ALL_REDUCE collectives.
PiperOrigin-RevId: 563562951 | Clive Verghese | cliveverghese@google.com | 1,694,125,877,000 |
22a77643a64a84e2df279dcac6560b24bc5c16d5 | Updated ml_dtypes to latest commit.
This is required by XLA to allow casts to int128.
PiperOrigin-RevId: 563565776 | Antonio Sanchez | cantonios@google.com | 1,694,126,487,000 |
d1fab65a87d5842d2283fa278561343cd1cd6ec9 | Sort the imports before writing the public API to ensure determinism.
PiperOrigin-RevId: 563569088 | Brian Wieder | bwieder@google.com | 1,694,127,278,000 |
d79a841d0041a2f54334df1218c945980e1a766c | Create unit tests with simple model to verify UniformQuantizedConvolution/Dot numerical correctness. Fix padding bug for Convolution in lowering passes.
PiperOrigin-RevId: 563569683 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,127,422,000 |
1f56e5821326ce586a9fe0061159526b63d8f8b1 | Allow S4/U4 for BitcastConvert and XlaCallModule.
PiperOrigin-RevId: 563570251 | Dinghua Li | dinghua@google.com | 1,694,127,580,000 |
cae357c57441ecf111b736cb2f276e1e9f5bcd1c | Fix TSL path locations after TSL/XLA Vendoring
PiperOrigin-RevId: 563575970 | Jake Harmon | jakeharmon@google.com | 1,694,128,907,000 |
44e33dd1886d80d875e11ed037cdb0191e5cd1a4 | Strip BAZEL_CURRENT_REPOSITORY from the options passed to nvcc on Windows.
BAZEL_CURRENT_REPOSITORY is given a value like
```
BAZEL_CURRENT_REPOSITORY=\"xla\"
```
but nvcc is unable to parse the quotes in that argument. It's a value produced by Bazel and probably not used, so just skip it.
PiperOrigin-RevId: 563577393 | Peter Hawkins | phawkins@google.com | 1,694,129,233,000 |
5aa02a6765473bc8c11b793afbef02e2a789568a | Support stablehlo.batch_norm_training conversion
Spec: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#batch_norm_training
Lowering the op with some TF ops (tf.Mean, tf.SquaredDifference)
which will be eventually lowered to TFL ops.
Tested with tensorflow/compiler/mlir/lite/stablehlo/tests:unfuse_mhlo_ba... | Terry Heo | terryheo@google.com | 1,694,130,581,000 |
6657dcc51c50e0a8c64ab84bfccb8c52ab9e9024 | Remove unused test dependencies
PiperOrigin-RevId: 563596107 | Weiyi Wang | weiyiw@google.com | 1,694,134,459,000 |
2e5c82c4de47d3570448bf12cd4acb6a0a68ff49 | Add int4 C++ types in TensorFlow.
This is a prerequisite to adding proper dtype support. Exposes
`tsl::int4` and `tsl::uint4` (with corresponding `tensorflow::` versions)
into the platform list of types.
PiperOrigin-RevId: 563607886 | Antonio Sanchez | cantonios@google.com | 1,694,138,403,000 |
b1ec9af0ec003df3a1b6e12271c84d60e421aaf9 | [XLA:runtime] Fix compilation problem under MSVC.
~Argument() names the destructor of the superclass, and MSVC rightly complains about it in the event that (*self)[i] does not return an Argument, which it does in some cases.
PiperOrigin-RevId: 563615070 | Peter Hawkins | phawkins@google.com | 1,694,141,446,000 |
260c006f3a1b08b1f7d3efa418ab7c0ca184c928 | Remove custom u4/s4 definition, use common ml_dtypes one.
We want a consistent definition for all int4 usages. Currently,
JAX/TF use the underlying type defined in `ml_dtypes`.
PiperOrigin-RevId: 563633360 | Antonio Sanchez | cantonios@google.com | 1,694,147,778,000 |
9e45e99ebcf9f8e033c806ac3bb990f20440f6ce | Merge pull request #61699 from DrMarcII:master
PiperOrigin-RevId: 563639265 | TensorFlower Gardener | gardener@tensorflow.org | 1,694,150,025,000 |
0f0d0202e81f6565f47e48464c696c6aebf083c2 | Add XlaCallModuleOp based function lifting
This change introduces new function lifting mode in the TPU quantizer to lift
the StableHLO subgraph with TF::XlaCallModuleOp.
PiperOrigin-RevId: 563640156 | Jaesung Chung | jaesung@google.com | 1,694,149,911,000 |
d8fa26de7f08386307a54fe141e04e70225255a6 | [XLA:GPU] Fix compilation failures under MSVC.
Two problems:
* 1L is not int64 sized on Windows, leading to a template argument deduction problem.
* MSVC limits string literals to 16K characters. However, you can use string literal concatenation to concatenate shorter literals into a long literal. Do that instead.
Pi... | Peter Hawkins | phawkins@google.com | 1,694,150,183,000 |
1131998241ba0d5a114701f89cb9ba0a49be13dc | Merge pull request #61767 from boomanaiden154:gemmlowp-fix
PiperOrigin-RevId: 563641616 | TensorFlower Gardener | gardener@tensorflow.org | 1,694,151,163,000 |
00751f0ba2a4f7c99c8887bda469fe004e1d6d31 | Internal Code Change
PiperOrigin-RevId: 563648249 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,152,141,000 |
a042beb1dceee9c24f22bd6d556bd1d28a4bcb81 | compat: Update forward compatibility horizon to 2023-09-08
PiperOrigin-RevId: 563686005 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,163,753,000 |
453a85371a3b7ee93babd20ad4793a57e679b73b | Update GraphDef version to 1613.
PiperOrigin-RevId: 563686010 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,163,754,000 |
8cf6ece80115da66eb7286a04c9722d6398c26e5 | Fusion analysis: Remove remaining dependencies on fusion_.
PiperOrigin-RevId: 563691979 | Johannes Reifferscheid | jreiffers@google.com | 1,694,165,714,000 |
2eaadca804328bbba543ee5ef889fb35e1d61375 | [XLA:GPU][NFC] Triton fusions: skip reductions with multiple reduced dimensions.
PiperOrigin-RevId: 563705867 | Ilia Sergachev | sergachev@google.com | 1,694,170,268,000 |
d83597075d282759860f643531d3a4c2dbf95449 | [XLA:GPU] Minor stylistic fixes to standalone pass pipelines.
PiperOrigin-RevId: 563712154 | Benjamin Chetioui | bchetioui@google.com | 1,694,172,261,000 |
c8cee32441d4c4459e6b3bdbd8f2aac96e6f23c7 | Triton emitter: remove some unnecessary helpers/templates.
PiperOrigin-RevId: 563735974 | Johannes Reifferscheid | jreiffers@google.com | 1,694,179,942,000 |
3cba63ada1e1ef45bee5567f596d9a358aa1b826 | [XLA:GPU] Give HloFusionAnalysis an entry point for custom boundary functions
This can be used to supply it with a graph cut instead of a fusion computation
PiperOrigin-RevId: 563739866 | Benjamin Kramer | kramerb@google.com | 1,694,181,209,000 |
c83bdfa50298ea98838055f34de9ae0128a8f9a3 | Add `Dilate` op for the implementation of StableHLO `pad` and `reduce_window`.
PiperOrigin-RevId: 563746345 | Quentin Khan | qkhan@google.com | 1,694,183,192,000 |
e73eb8b434e4bbe4fd1184a562c0342ea823a9a4 | [Triton] Add a patch to fix MSVC compilation.
PiperOrigin-RevId: 563753463 | Peter Hawkins | phawkins@google.com | 1,694,185,348,000 |
75d5511233b900b5fc6c3e221979c5d1ee8251a6 | [XLA:GPU][NFC] Improve iterator safety of gemm_rewriter_triton.cc
Check iterators before dereference if the validity is not trivial.
(I think there are invariants which ensure that they are valid currently, but this could break accidentally in a future change.)
Change an algorithm to use iterators more safely.
PiperO... | Tamás Danyluk | tdanyluk@google.com | 1,694,188,829,000 |
fa13ad9a5504e314ebba995f456c670e1548c8dd | Fix the name of the memories env var
PiperOrigin-RevId: 563773691 | Yash Katariya | yashkatariya@google.com | 1,694,190,761,000 |
a93d4a231a87565520cd52308f3951e0c975f98b | Move tpu import from python/__init__.py to python/modules_with_exports.py.
PiperOrigin-RevId: 563776249 | Fiona Lang | flang@google.com | 1,694,191,383,000 |
ec5107388dad0c1168df242c916e702fa9e14427 | Remove allow_multiple_exports param from tf_export.
PiperOrigin-RevId: 563784409 | Marc Fisher | fisherii@google.com | 1,694,193,165,000 |
426680e3d13105c7e573f21701230093ef9763c0 | [XLA:Python] Remove get_default_device_assignment().
This method is no longer used by JAX, and its existence is misleading (what it returns does not correspond to any particular default device assignment that JAX uses). Remove it.
PiperOrigin-RevId: 563784500 | Peter Hawkins | phawkins@google.com | 1,694,193,182,000 |
759a629a85a92264a6f2cae6860b1102c3078699 | Add no outside compilation markers verification pass to the bridge
PiperOrigin-RevId: 563787003 | Mason Chang | masonchang@google.com | 1,694,193,704,000 |
5001e89b452858d6c50c5c4cd0bf41e395668c18 | Disable XNNPACK_ENABLE_ARM_I8MM for aarch64 pip build
The current toolchain doesn't support it.
PiperOrigin-RevId: 563796297 | Terry Heo | terryheo@google.com | 1,694,195,617,000 |
04db8d85ecf3d916114cdfa45a8f652ac4a2e1c7 | [XLA:Python] Remove unused tpu_driver code.
PiperOrigin-RevId: 563822172 | Peter Hawkins | phawkins@google.com | 1,694,201,211,000 |
1a6c8304e874c3188940da477427e410894c4c71 | Enable tensorflow-io-gcs-filesystem as a runtime dependency for Apple Silicon machines
This does not need to be disabled anymore now that TensorFlow IO has Apple Silicon wheels available.
See https://pypi.org/project/tensorflow-io-gcs-filesystem/0.34.0/#files
PiperOrigin-RevId: 563826666 | Nitin Srinivasan | srnitin@google.com | 1,694,202,309,000 |
8c3a8e2cbfe09db2baef0dcbefcd5e160bf2da4d | Increase maximum number of resource ops iterations.
PiperOrigin-RevId: 563829150 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,202,904,000 |
dbb135daa199a83a7f621846c6ec6f3a257f4f82 | lite: Update filepath to stats_calculator.cc
Update CMake build rules to adopt new TSL path
PiperOrigin-RevId: 563829912 | Terry Heo | terryheo@google.com | 1,694,203,096,000 |
3dae6ab60a830f4d02c27eabdea91ef5a24cfe98 | Add Int4 types to TensorFlow.
These types are used by JAX/XLA. Expanding coverage in TF to match.
PiperOrigin-RevId: 563831916 | Antonio Sanchez | cantonios@google.com | 1,694,203,578,000 |
2d419f4e8f03cd099c99039db56dcd2f31093403 | Update Topology proto.
PiperOrigin-RevId: 563836927 | Clive Verghese | cliveverghese@google.com | 1,694,204,795,000 |
51e6f47e0a19e9917ed35a1d1ca5029b751c20fb | Remove dependency on tensorflow.bzl
This is required to prevent a circular dependency in https://github.com/tensorflow/tensorflow/pull/60161
PiperOrigin-RevId: 563836979 | Jake Harmon | jakeharmon@google.com | 1,694,204,811,000 |
248f5b67343bd5be8333a02629109c17d8271c3b | Add inference metrics pass to the V1 Compat pipeline
PiperOrigin-RevId: 563858734 | Mason Chang | masonchang@google.com | 1,694,210,022,000 |
994c19b152d63bb8f6c2b673895a61d203627268 | Integrate LLVM at llvm/llvm-project@f8a1c8b7c1d0
Updates LLVM usage to match
[f8a1c8b7c1d0](https://github.com/llvm/llvm-project/commit/f8a1c8b7c1d0)
PiperOrigin-RevId: 563871269 | Augie Fackler | augie@google.com | 1,694,213,271,000 |
70fc4e4a971affb31850d8ead3222a219abe73f6 | Fix proto bytes extension for windows.
PiperOrigin-RevId: 563892946 | Antonio Sanchez | cantonios@google.com | 1,694,219,801,000 |
93cb403496fb41fd60f5600adff6fd2e23c22643 | Add minimal XLA and MLIR changes to allow int4 checkpointing.
These utilities and MLIR types are required to enable PAX
model checkpointing involving the new TF int4 type.
PiperOrigin-RevId: 563915180 | Antonio Sanchez | cantonios@google.com | 1,694,228,811,000 |
fdb76cd1e4eb9d985703c5f376234d9c7392a5b2 | Move tpu import from python/__init__.py to python/modules_with_exports.py.
PiperOrigin-RevId: 563919380 | Fiona Lang | flang@google.com | 1,694,230,427,000 |
2b96998473b5adc52e9034313ae4dd70c3c25b3f | [tf2xla] fix while loop body dynamic shape handling
PiperOrigin-RevId: 563925797 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,233,165,000 |
93beea2c9ddd1355009692cbe74c54ded2e45dc2 | compat: Update forward compatibility horizon to 2023-09-09
PiperOrigin-RevId: 563968227 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,250,123,000 |
fe4d6d30fb74cb13b80bddc46369d031ec6db6d0 | Update GraphDef version to 1614.
PiperOrigin-RevId: 563968239 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,250,126,000 |
a9146b01c5fee315cfd7647a3bb5f75be0d76701 | [XLA:GPU] Ignore unsupported custom calls when --xla_gpu_mock_custom_calls is set
PiperOrigin-RevId: 563976196 | George Karpenkov | cheshire@google.com | 1,694,254,077,000 |
22430372eb3b29f5bf336295d214365e5f62f70c | fix hipcc path | Akash Patel | 17132214+acxz@users.noreply.github.com | 1,694,288,761,000 |
3821e98a7ecbb0bf6739465eeccf82bd539ee587 | Add int4 cast, fill, and functor ops to enable testing.
These are the minimal support ops required for initializing and running tests.
PiperOrigin-RevId: 564096777 | Antonio Sanchez | cantonios@google.com | 1,694,320,384,000 |
5f2db08255b405a5dd16b30f65f173ceaa42bab3 | Update GraphDef version to 1615.
PiperOrigin-RevId: 564133178 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,336,588,000 |
3774cc70ae533a4459dafbb30c47ddc407f3ca20 | compat: Update forward compatibility horizon to 2023-09-10
PiperOrigin-RevId: 564133181 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,336,589,000 |
4c377d72b9aed143967b135f0de6c5ee5342fe8f | Explicit forbid dynamic shape aliasing due to potential runtime-compiler dynamic shape size calculation mismatching
PiperOrigin-RevId: 564183652 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,364,928,000 |
71c059585d5f713f2f9baeba39db524b42665d43 | Refactor quantization options
- Remove Experimental in the message name
- Add NO_QUANTIZE quantization scheme
- Change freeze_all_variables to bool type
- Allow users to specify both op type and node name in the quantization unit
PiperOrigin-RevId: 564231195 | Thai Nguyen | thaink@google.com | 1,694,392,496,000 |
b92caec1b984b5ad725a623d01e4c3aec9f37896 | Extend conversion function from GoogleEdgeTpuSettings in protobuf format to GoogleEdgeTpuSettings in flatbuffer format.
PiperOrigin-RevId: 564301432 | Matt Kreileder | mattalexander@google.com | 1,694,420,476,000 |
2787022672cd5b79d7cced4f043714bc95122d54 | Rename whitelist to allowlist
The change has been made in Bazel in 2020, so it should be safe to rename in TF as well.
PiperOrigin-RevId: 564301792 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,420,582,000 |
8440c32a535656ca27dab167e41823cad291d9bc | When we use LOG(INFO), the messages appear on screen in the OSS build - this is usually not what we want.
PiperOrigin-RevId: 564305557 | Sergey Kozub | sergeykozub@google.com | 1,694,421,814,000 |
ce7f76225b8b786f38b016cee8a7b9164e9b2754 | compat: Update forward compatibility horizon to 2023-09-11
PiperOrigin-RevId: 564309075 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,422,926,000 |
b25ff8451015b1abfef0ca5aa8c2edd2edf95e92 | Update GraphDef version to 1616.
PiperOrigin-RevId: 564309076 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,422,926,000 |
70f8a1f2e466f8d618328fbedf7e81b980aaec60 | Only allocate one xnn_value per delegated tensor
This reduces massively the memory allocated by XNNPACK for highly partitioned subgraphs
PiperOrigin-RevId: 564313169 | Alan Kelly | alankelly@google.com | 1,694,424,061,000 |
2aed80580d0402edf76c8c72982e44d842bafe4e | NFC: Use std::optional instead of -1 in triton matmul emitter.
This makes it harder to accidentally use -1.
PiperOrigin-RevId: 564334287 | Johannes Reifferscheid | jreiffers@google.com | 1,694,430,495,000 |
6ccad5790e54214daa2e93e92b3497d7f3496c2b | Define macro TFLITE_CONDITIONAL_NAMESPACE, for use in code using cc_library_with_tflite.
[Rolling forward]
PiperOrigin-RevId: 564345870 | Fergus Henderson | fergus@google.com | 1,694,434,396,000 |
f598d78d26ec9fce7fe7ade4c81dd0fc2d0a6b5f | Integrate LLVM at llvm/llvm-project@e592c2dcf5b7
Updates LLVM usage to match
[e592c2dcf5b7](https://github.com/llvm/llvm-project/commit/e592c2dcf5b7)
PiperOrigin-RevId: 564346651 | Benjamin Kramer | kramerb@google.com | 1,694,434,670,000 |
3124c67abf76cbe3a9a2bebdc70d6d529c5e1c56 | [XLA:GPU] Fix ir_emitter_triton_large_test's TritonSoftmaxTests to account for
different flag default.
PiperOrigin-RevId: 564374361 | Benjamin Chetioui | bchetioui@google.com | 1,694,441,904,000 |
214dea849f5f2db60c64952339c4229d76968c34 | [XLA] [NFC] Limit the number of edges between *same* nodes.
Sometimes, there's an exorbitant number of edges between two same notes A and B
(e.g. passing parameters through). After ~64 it becomes unreadable, so better
to skip others.
PiperOrigin-RevId: 564379502 | George Karpenkov | cheshire@google.com | 1,694,443,070,000 |
5c00a18e0eff008d74b53725039de53488cec434 | Return failure if num_results is negative in tfrt_fallback.executeop
PiperOrigin-RevId: 564415721 | Kuangyuan Chen | chky@google.com | 1,694,450,821,000 |
2ed84d32c46dffba6523f23a4d2dd9d7a54095cf | change triton autotuning AutotunerUtil::AddResult non-ok status to a non-blocking LOG(WARNING)
PiperOrigin-RevId: 564420754 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,451,824,000 |
d1b68813f436b6a22ff5fe4c9d81a727af523b05 | [XLA] Use max size of buffer in heap simulator and buffer assignment and make size DCHECKs into CHECKs
PiperOrigin-RevId: 564421004 | Berkin Ilbeyi | berkin@google.com | 1,694,451,877,000 |
6e2a47769bc7cf1b7a5ab9aebb33a1cb8228f298 | Add pyi file for pywrap_tfe
PiperOrigin-RevId: 564428072 | A. Unique TensorFlower | gardener@tensorflow.org | 1,694,453,075,000 |
778dd23c8baa309b0f5f38bd353eff04a09bc882 | Prepare rename GenericFunction to PolymorphicFunction
PiperOrigin-RevId: 564434713 | Faizan Muhammad | fmuham@google.com | 1,694,454,230,000 |
907fb5bddf7eafea7166748109a537fdf7195d8c | NFC: Split launch grid and validation out of MatMul emitter.
Also
- Unify naming to ${scope}_${thing}_dim_idx.
- Make logic for split_lhs_nc a bit easier to follow (at least in my opinion), by
decoupling it from the batch_size computation.
PiperOrigin-RevId: 564435916 | Johannes Reifferscheid | jreiffers@google.com | 1,694,454,455,000 |
cdfcea08745cf045ce347de2759ae038f938f53e | Refactor the step id generation so that it can be reused by other library.
PiperOrigin-RevId: 564441419 | Kuangyuan Chen | chky@google.com | 1,694,455,463,000 |
73e3999026c46ef61c77b9e9051c7a5eddec88c2 | Load tensorflow_framework with RTLD_GLOBAL
This way libtpu can use symbols from it.
PiperOrigin-RevId: 564454296 | Haibo Huang | hhb@google.com | 1,694,457,830,000 |
ebdbaa3f63e0469cb8723ed24c6edd5c9106576b | [SPMD] Create a placeholder sharding of type UNKNOWN with the lowest precedence that can be overwriten by any other shardings. (2/N)
UNKNOWN type is not supported in subgroup sharding.
PiperOrigin-RevId: 564456004 | Tongfei Guo | tongfei@google.com | 1,694,458,181,000 |
c94ea4e47ec18b8bb9fbfd8329d5d709c1a5e634 | [tflite] Add Eager API for activation functions.
Leaky Relu, Elu, Relu, Relu1, Relu0To1, Relu6, Tanh, Sigmoid, and HardSwish
PiperOrigin-RevId: 564472058 | Grant Jensen | grantjensen@google.com | 1,694,461,488,000 |
98eb94254fdc83f0fb375b77ebd9b597d2625ee5 | [tflite] Implement Eager API for quantized multiply
PiperOrigin-RevId: 564472190 | Grant Jensen | grantjensen@google.com | 1,694,461,517,000 |
66edf03932541b8141d3842d25460266a0f16132 | Handle negative `segment_ids` when auto-vectorizing `tf.math.unsorted_segment_sum`
According to
https://www.tensorflow.org/api_docs/python/tf/math/unsorted_segment_sum:
> If the given segment ID i is negative, the value is dropped and will not be
> added to the sum of the segment.
PiperOrigin-RevId: 564485570 | Philip Pham | phillypham@google.com | 1,694,464,322,000 |
fc6b8ba7419184db96c491677ef8181c0aba0cbf | Add missing S/UINT4 xla-to-TF type entries.
PiperOrigin-RevId: 564494782 | Antonio Sanchez | cantonios@google.com | 1,694,466,188,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.