hash
stringlengths
40
40
msg
stringlengths
1
131k
author
stringlengths
1
33
email
stringlengths
0
57
date
int64
1,447B
1,698B
b0a9536d587e341b088b9a52d007c8c0633037b2
[XLA:GPU] [NFC] Remove gpu_device_info as a redundant indirection layer PiperOrigin-RevId: 569523071
George Karpenkov
cheshire@google.com
1,696,007,534,000
e2a703188ce58e0448bc5ddae2c4f3fd325dbba1
Refactors the strategy & cost creation within EnumerateAll2DPartition[Reshape] into their own dedicated functions. PiperOrigin-RevId: 569529051
A. Unique TensorFlower
gardener@tensorflow.org
1,696,008,919,000
cee7495bd2d351a4a1c217380ab6cd8f71a8e43a
Add gather test coverage for some bug. PiperOrigin-RevId: 569532078
Tongfei Guo
tongfei@google.com
1,696,009,585,000
a3ff731fd9e75ac8b035de1b402844ba57f9999d
Merge pull request #58255 from CaveMike:dev_n1to1_b PiperOrigin-RevId: 569536196
TensorFlower Gardener
gardener@tensorflow.org
1,696,010,769,000
fed78494b256794b95f03cf459fbb5d8f7428eb1
[xla:gpu] Deallocate and disable gpu graphs that cause graph update error PiperOrigin-RevId: 569538929
Anlun Xu
anlunx@google.com
1,696,010,993,000
dab293abf1b258358660173965bc955c3884e5c0
Enable cross compilation for GPU PJRT compiler/client. PiperOrigin-RevId: 569541694
Shixin Li
shixinli@google.com
1,696,011,532,000
cedad1c00b75d563a8c6a0792309287ad26c07ce
Move tpu import from python/__init__.py to python/modules_with_exports.py. PiperOrigin-RevId: 569547232
Fiona Lang
flang@google.com
1,696,012,800,000
ad63ca43a2dcc28ee937a2eb8d20da6ddd77087b
Rename the parameters of the IsDivisible function in auto_sharding_util to be more intuitive. PiperOrigin-RevId: 569548788
A. Unique TensorFlower
gardener@tensorflow.org
1,696,013,192,000
8f12131a42a2573cd4824e41a4058f31e18b13b8
#tf-data Reduce the starting `max_outstanding_requests` to 16 before any element is produced to avoid OOM when jobs start with many tasks. PiperOrigin-RevId: 569550401
Wilsin Gosti
wilsin@google.com
1,696,013,589,000
fa4873704576a8522f46e3bde4a79296a9afa00c
#tf-data-service Optimize filepath for snapshot writes. PiperOrigin-RevId: 569564835
Matt Callanan
mpcallanan@google.com
1,696,017,140,000
15e8ebf798d02f3ea489f467bb94b3ec3befedcd
Add a presubmit check to ensure no imports are added to python/__init__.py. PiperOrigin-RevId: 569567547
Fiona Lang
flang@google.com
1,696,017,683,000
78da7d4ba33fdfb709bc5c6f2ae7e2d0d2b70fac
Rollback of PR #58255 PR #58255: Fix ReLUN1To1 fused activation for OpenCL and OpenGL Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/58255 Fix GPU (OpenCL and OpenGL) compile errors on Android. Fix ReLUN1To1 fused activation for OpenCL and OpenGL. Copybara import of the project: -- aa2a1ced77...
A. Unique TensorFlower
gardener@tensorflow.org
1,696,018,232,000
c0ec7cea40eba9b381c12cbe502ddfc539fabb8d
Rewrites EnumerateAll2DPartition[Reshape] using a recursive implementation that supports an arbitrary number of device mesh shapes. PiperOrigin-RevId: 569570620
A. Unique TensorFlower
gardener@tensorflow.org
1,696,018,288,000
dd4ee66b5c075ae1c2f9b53b568aeed8c6f2c290
Remove redundant sponge link and non-functional fusion link from generate_index_html.sh PiperOrigin-RevId: 569570997
David Dunleavy
ddunleavy@google.com
1,696,018,370,000
3b0a7418fad00d63857cc563ad56f67aa57747b7
Fix conversion of ops that use options declared in BuiltinOption2. PiperOrigin-RevId: 569575379
Quentin Khan
qkhan@google.com
1,696,019,436,000
9fe8a78bfe981a2308b7e5fc81f5c74041c7dae7
fix test
Kanvi Khanna
kanvi.khanna@intel.com
1,696,020,099,000
af3421baf53759a776c375228e5e8e72a5103eeb
Reenable flaky test flag for 2.15 branch cut and release. This will be reverted at HEAD post branch cut. PiperOrigin-RevId: 569587811
Michael Hudgins
michaelhudgins@google.com
1,696,022,498,000
bb181e5019d03eef43741e68e63045dc450ed70d
#tf-data-service Increase default client timeout to 5 minutes. Otherwise, if the dispatcher or client restarts, this can lead to false client cleanup. PiperOrigin-RevId: 569592616
Yang Chen
yangchen@google.com
1,696,023,780,000
c47518f0239a74dbed900a20d8cd6b89f4b2ef70
PR #5855: [NVIDIA:GPU] Add XLA compilations flags per this discussion in this thread https:/… Imported from GitHub PR https://github.com/openxla/xla/pull/5855 …/mail.google.com/chat/\\#chat/space/AAAAQuUms_4/6XkLzS2sEUA: As of now there are tons of works in XLA sharding. To have the tests maintainable, we were think...
Vladislav
vkozlov@nvidia.com
1,696,024,865,000
76123afecc091e201cf53bc2606a3f5a5280c1d8
Remove the long-deprecated type gtl::optional PiperOrigin-RevId: 569602891
Dmitri Gribenko
dmitrig@google.com
1,696,026,407,000
41c80fa0144db695c39a3e255f2263ee4eea0d04
Use dladdr1() to get the full path of current module PiperOrigin-RevId: 569614887
Haibo Huang
hhb@google.com
1,696,029,696,000
94bde4690ca299ec754a3056830f5413a30cc006
Delete dtensor LazyLoader instances now that python/__init__.py has been emptied. PiperOrigin-RevId: 569618035
Fiona Lang
flang@google.com
1,696,030,694,000
b476f9d220c8e6b922da3d39b226248df07620ca
Move some common methods to PjrtCApiTestBase. PiperOrigin-RevId: 569623763
Jieying Luo
jieying@google.com
1,696,032,367,000
6334682d8fcd1413eab2464e35696be48638a7a7
lite: add passes to insert transpose to & from [b, 0, 1, f] for transpose conv PiperOrigin-RevId: 569665655
Zichuan Wei
zichuanwei@google.com
1,696,049,644,000
28828b566c3e77bcda72b1865d12f6ae0fbf29e3
compat: Update forward compatibility horizon to 2023-09-30 PiperOrigin-RevId: 569693143
A. Unique TensorFlower
gardener@tensorflow.org
1,696,064,525,000
e337d91c7d3396528b1bf3fefa707f284958b5cf
Update GraphDef version to 1635. PiperOrigin-RevId: 569693153
A. Unique TensorFlower
gardener@tensorflow.org
1,696,064,528,000
e24e476386ad381ca9c08420aad7e697932eb1d2
Fix
Eli Kobrin
kobrineli@ispras.ru
1,696,080,407,000
6872b8b1d4c0f52e90a38cd90dee49316b6b87f1
Consolidates the dot handler's mesh dimension enumeration logic into a single method. PiperOrigin-RevId: 569771316
A. Unique TensorFlower
gardener@tensorflow.org
1,696,113,114,000
022d32e6d9a0c666113d078460af91e260ce3c9f
Move `:gpu_client` into `if_cuda_or_rocm` rather than duplicating it within `if_cuda` and `if_rocm` This prevents a build error when bazel matches both cuda and rocm PiperOrigin-RevId: 569787230
David Dunleavy
ddunleavy@google.com
1,696,123,826,000
0c79e428577c49f91cba0d0fefd34be8fea35c85
Added new threat model. PiperOrigin-RevId: 569813726
Krzysztof Kotowicz
koto@google.com
1,696,140,130,000
397aabb53fa4cba39aed5a32c38c1492db208bb3
compat: Update forward compatibility horizon to 2023-10-01 PiperOrigin-RevId: 569832476
A. Unique TensorFlower
gardener@tensorflow.org
1,696,150,930,000
825dcfb0253f214b94904cc9a3e4e79570bb4d29
Update GraphDef version to 1636. PiperOrigin-RevId: 569832478
A. Unique TensorFlower
gardener@tensorflow.org
1,696,150,931,000
ad39c958186235fa84e6574ec74ab0ae333d9bd7
Consolidates the dimension checking logic into a single method. PiperOrigin-RevId: 569865640
A. Unique TensorFlower
gardener@tensorflow.org
1,696,170,762,000
59ba591a05928e2da3a4535beb4c2d3d76d4ef0d
Pass along the shape param from tf.compat.v1.get_variable to its underlying variable creator. At this point, we have already checked that either shape is compatible w/ initial_value, or that it is not specified (None). Therefore it should always be safe to pass it. Having it can help subsequent nested variable creators...
A. Unique TensorFlower
gardener@tensorflow.org
1,696,218,236,000
397615b1550c5d73ea9cf0f963de3dfae090b73b
Update GraphDef version to 1637. PiperOrigin-RevId: 569987303
A. Unique TensorFlower
gardener@tensorflow.org
1,696,237,325,000
038a5416cab5e6daf1729cb78d2d4ca3f54db7d7
compat: Update forward compatibility horizon to 2023-10-02 PiperOrigin-RevId: 569987324
A. Unique TensorFlower
gardener@tensorflow.org
1,696,237,330,000
63ab401f8d6f70221de50178b39d754c1801a464
Use template parameter rather than non-template parameter for KernelType. PiperOrigin-RevId: 570020629
Fergus Henderson
fergus@google.com
1,696,245,435,000
3955407448705f347ae4387b7b304b0d140d4139
Merge pull request #62010 from cjflan:cf/2.15-release-bump PiperOrigin-RevId: 570050293
TensorFlower Gardener
gardener@tensorflow.org
1,696,254,464,000
9736f56f7d6416b849ca1ce8c6ab1a4a8667df5c
[XLA:GPU] Make loaded scalars 1-element tensors in the Triton emitter. PiperOrigin-RevId: 570055575
Ilia Sergachev
sergachev@google.com
1,696,255,577,000
7ae63482b9cdf51b773a0b52eb2aa6e097aed612
Generalizes the Dot Handler's mesh dims to be of arbitrary size (though still limited to 2 for now) PiperOrigin-RevId: 570060379
A. Unique TensorFlower
gardener@tensorflow.org
1,696,256,954,000
65fd4e5b5d4c127ea46d90aaf38ebe32e57e80a8
[stream_executor] NFC: Forward all users of dnn_proto to tsl/protobuf library Remove indirection via stream_executor:dnn_proto target. PiperOrigin-RevId: 570071946
Eugene Zhulenev
ezhulenev@google.com
1,696,260,037,000
97c6f22c91db5f5cca04462eafc0fd69590c50b2
Refactor batches_ into a function to make future changes easier batches_ will be replaced with high and low priority versions. It will be easier to update all the call sites if they all call the same function to get batches_. PiperOrigin-RevId: 570075341
Marissa Ikonomidis
marissaw@google.com
1,696,260,930,000
702de55dbbc726d49d52bbceac42307aef64bff1
Use non-member equality operators for TensorShape and add a non-member operator== for PartialTensorShape. PiperOrigin-RevId: 570081029
A. Unique TensorFlower
gardener@tensorflow.org
1,696,262,278,000
95edcf9e835ba67939fe582846211ae1d669df80
Add modifications for ARM 64 ci
Michael Hudgins
michaelhudgins@google.com
1,696,263,158,000
759fda54f65f59354f68fc324387f363779b572a
Update bazelrc
Michael Hudgins
michaelhudgins@google.com
1,696,263,449,000
3cc8d2d34f92b8aa0736ae8edef37e85088cd30b
Internal changes only. PiperOrigin-RevId: 570085339
A. Unique TensorFlower
gardener@tensorflow.org
1,696,263,191,000
18831bfbf7801bb9b8dfac2b6a6a5833e9cf24c4
Introduces a HandlerBase class containing functionality -- like Split() -- common to both DotHandler and ConvHandler. PiperOrigin-RevId: 570089566
A. Unique TensorFlower
gardener@tensorflow.org
1,696,264,149,000
2b26764c0d13baca0c2bd938ebe7c9f7f0c5ea71
Tag TF-TPU wheel with manylinux_2_27_x86_64 PiperOrigin-RevId: 570096613
Kanglan Tang
kanglan@google.com
1,696,265,691,000
1034d8438b0695d4452e92bbe25e9018c4ee46cf
Deprecate batches_ and replace with high_priority_batches_ In a future cl, low_priority_batches_ will be used. Batches will default to high priority unless otherwise specified. PiperOrigin-RevId: 570099579
Marissa Ikonomidis
marissaw@google.com
1,696,266,294,000
6bd9b9bb83a61a74cb8d356390d83dd83bd1572c
[XLA: Python] Update PJRT plugin configuration mapping with bool type PiperOrigin-RevId: 570100527
Samuel Agyakwa
sagyakwa@google.com
1,696,266,476,000
b68b3b2c3824e440e3c27463ac170d3627496f0d
Make enum_value packed in PackedTestValue. Enums are packable just like other scalar types. PiperOrigin-RevId: 570104566
Larry Lansing
llansing@google.com
1,696,267,337,000
a008c44880352097996205b3cc7236c9b5e639e4
Minor simplification to CheckDims (where the tensor dim indexing is performed by the caller). PiperOrigin-RevId: 570106474
A. Unique TensorFlower
gardener@tensorflow.org
1,696,267,727,000
f66187934aa12f37393ae497f0ec01a47a2c5273
Include link.h for struct link_map PiperOrigin-RevId: 570129061
Haibo Huang
hhb@google.com
1,696,272,161,000
fa5c608e56cdb0ae3ec38bc872fa4725a26290aa
Metric XLA binary size loaded onto the GPU device. PiperOrigin-RevId: 570129604
A. Unique TensorFlower
gardener@tensorflow.org
1,696,272,286,000
aa3894147f6f3a8c304f658fd35179cefc8187a1
Allow quantized tensor as the input for mhlo.reshape. This also include change the folding logic to allow constant folding on quantized tensor. There are basically two changes proposed here. The first change in reshape method is a fix to avoid folders to crash on quantized type, so that translation like MHLO -> (X) ca...
A. Unique TensorFlower
gardener@tensorflow.org
1,696,272,897,000
8e452b87adea81158a7c76af7bfdee0ad7b038db
Update TFRT dependency to use revision http://github.com/tensorflow/runtime/commit/fb9b279b454c71995f63d33a9c78d52c4bf3ca1c. PiperOrigin-RevId: 570138454
A. Unique TensorFlower
gardener@tensorflow.org
1,696,273,913,000
66349e3a058164fe1eae97487c26e688a7fe3567
py and cc do work now that i fixed labels
Michael Hudgins
michaelhudgins@google.com
1,696,274,280,000
67f519e727605f502c711cd340df46af89c4085e
Change ARM64 docker runtime base to be CUDA. A separate non-cuda will be added later PiperOrigin-RevId: 570139525
Michael Hudgins
michaelhudgins@google.com
1,696,274,163,000
ad1ca90c246e32cabbd3de92bee25220bc2a3a1d
Add host analysis to TPU timeline. PiperOrigin-RevId: 570139622
Clive Verghese
cliveverghese@google.com
1,696,274,186,000
e6b58a8b231099b0207030fcad0e4e211ef8f116
Fix config typo in TF's bazelrc file. PiperOrigin-RevId: 570141677
Camilla Halsey
halseycamilla@google.com
1,696,274,615,000
d397ccdb37e7fed88420e95d13e5ddf7bb183c84
Move `GetMaxIdsAndUniques()` function to sparse_core_ops_utils so that non-xla ops can call it. PiperOrigin-RevId: 570152309
Hye Soo Yang
hyey@google.com
1,696,276,722,000
c791fe82581e5e1ad4a05a91281413832ff09289
Add more API set up for Mock GPU client. Also clean up previous mock GPU client API. PiperOrigin-RevId: 570153877
Tao Wang
wangtao@google.com
1,696,277,043,000
d6e25b069d9ddf4111d7def74c145d118c6f4f81
Promote c_api_opaque.h to the 'mostly stable' API. For compatibility reasons, ':c_api_experimental*' targets in 'core/c' depend on 'c_api_opaque*'s. PiperOrigin-RevId: 570159652
A. Unique TensorFlower
gardener@tensorflow.org
1,696,278,265,000
d7c159b8784e217c132e58df159c17f6e286eb85
Add xla flag for `xla_sparse_core_minibatch_max_division_level`. PiperOrigin-RevId: 570162883
Hye Soo Yang
hyey@google.com
1,696,278,888,000
bd5427c612480d96c0fba35cd2384797d35b2aa0
Creates a new method -- CreateInputSpec() -- that constructs the appropriate HloSharding objects. PiperOrigin-RevId: 570164031
A. Unique TensorFlower
gardener@tensorflow.org
1,696,279,120,000
d9af31c0a6f3750eeeb57bc01e9736396fb87861
[XLA:GPU] Add the flag for xla_gpu_mock_custom_calls PiperOrigin-RevId: 570170696
Tamás Danyluk
tdanyluk@google.com
1,696,280,340,000
54b55af7a69d212d88cc43d2a4f73fe69fd3d569
Added a new reference type. PiperOrigin-RevId: 570173600
A. Unique TensorFlower
gardener@tensorflow.org
1,696,280,837,000
4184aa8e64b65fcec7cde71678b66f6c96bdb9ad
Enable UniqueV2Op in MLIR PiperOrigin-RevId: 570174493
A. Unique TensorFlower
gardener@tensorflow.org
1,696,281,018,000
61f4f2f9a0e4549309f112f7479c3b2d14a2d6b4
Fix error message in function attribute parser. PiperOrigin-RevId: 570174801
Haoyu Zhang
haoyuzhang@google.com
1,696,281,074,000
b439a7f42ca8ea9e8c24be2945382918e126bd0c
Allow quantized tensor as the input for mhlo.transpose. PiperOrigin-RevId: 570183739
A. Unique TensorFlower
gardener@tensorflow.org
1,696,282,937,000
79ac64f6a7bca73ecbdee6507439f3234cb1a6c2
Integrate StableHLO at openxla/stablehlo@6bf3f5d PiperOrigin-RevId: 570185242
Sandeep Dasgupta
sdasgup@google.com
1,696,283,224,000
d18f72626b42c787eeead92b64278412e7e32dce
Update TFRT dependency to use revision http://github.com/tensorflow/runtime/commit/09614836c83b716cfc7f6ed63208a0a122020799. PiperOrigin-RevId: 570188293
A. Unique TensorFlower
gardener@tensorflow.org
1,696,283,772,000
44aaffee483e90a1e60a9e263952a8dec40ceb98
Creates a single entry point to handle dimension checking & input spec creation & strategy generation. PiperOrigin-RevId: 570194157
A. Unique TensorFlower
gardener@tensorflow.org
1,696,284,869,000
2a64b462701e35cae9f7e8c1bfad3363245c5267
[XLA:SPMD] Bug fix that empty tuple sharding shouldn't be a shard group. PiperOrigin-RevId: 570195864
Tongfei Guo
tongfei@google.com
1,696,285,221,000
e6cb578e18e43a04702644a4fefb2169bc2e4b09
pin ml-dtypes version
jameshollyer
j.c.hollyer@gmail.com
1,696,285,793,000
c37628c576dca6a45bb0b13ca85204829b5893c4
[pjrt] Use a single implementation for Vec128 transpose We can generalize the transpose logic a bit further to make it so that it can handle rectangular Vec128 transposes. PiperOrigin-RevId: 570201065
David Majnemer
majnemer@google.com
1,696,286,407,000
59d97aece0420ab1576ff9adc469e04279e68766
Add support for detecting the AllGather->DynamicSlice pattern. PiperOrigin-RevId: 570203891
A. Unique TensorFlower
gardener@tensorflow.org
1,696,287,050,000
b39de08b5ee1c7795da483e6b57f5215832efed6
Adds an experimental kwargs in TPUEmbeddingServing for restore. PiperOrigin-RevId: 570207893
A. Unique TensorFlower
gardener@tensorflow.org
1,696,287,991,000
fa580a07c95756615439c0d9386321c226df3580
Add flag to allow printing more debug information in buffer allocation PiperOrigin-RevId: 570210990
Sergey Kozub
sergeykozub@google.com
1,696,288,785,000
ba466831602c68ec066fb9257ebdfe294773f6cc
[XLA:SPMD] Support sharding propagation/partitioning for variadic Scatter. 1. Make sharding prop work for variadic Scatter. 2. If variadic Scatter's operands do not share the same sharding, find a common sharding and do reshard. 3. Some code refactoring. 4. Tests. PiperOrigin-RevId: 570218446
Tongfei Guo
tongfei@google.com
1,696,290,668,000
1e2ba432f94c54ffeaea535d91ff11a1f76b4eae
Remove rc3 from python 3.12 install now that stable release is out PiperOrigin-RevId: 570226890
Yash Katariya
yashkatariya@google.com
1,696,292,920,000
d1ba59bcc03112ba5351b8f50db206b40d63d643
[xla:gpu] Deallocate all graphs with same device ordinal if a graph fallbacks PiperOrigin-RevId: 570233968
Anlun Xu
anlunx@google.com
1,696,294,934,000
80bfa43d06907ac9ce1df928eb74983f1ad10bac
Support int4 in hlo_utils.cc functions. In particular, support int4 in CreateDenseAttrFromLiteral and CopyDenseElementsDataToXlaFormat. PiperOrigin-RevId: 570238807
Reed Wanderman-Milne
reedwm@google.com
1,696,296,696,000
832ec6f6edf3814382e6d6433ef3a3b02c12e41d
Fix to restore from SC checkpoints PiperOrigin-RevId: 570241756
A. Unique TensorFlower
gardener@tensorflow.org
1,696,297,725,000
929ed5d15494349577ed70b65f289272f53e4673
[StableHLO to MHLO] Update element-type check for quantized types for BroadcastInDimOp. Also, includes adding tests for BroadcastInDimOp/RehsapeOP/TransposeOp. The element type checks for the last two ops are updated as part of cl/568657411 and cl/570133154. PiperOrigin-RevId: 570242706
Sandeep Dasgupta
sdasgup@google.com
1,696,298,061,000
a965566d7b34a6ba41627a64ca1d1e659c038098
Add QuantizationUnitLoc and enforce selective quantization during lifting The QuantizationUnitLoc is added right after importing to MLIR, meaning it is the first pass to run. PiperOrigin-RevId: 570248862
Thai Nguyen
thaink@google.com
1,696,300,350,000
fa46c8f848649e769c307d0cd4d68f397800e398
Enable NO_QUANTIZE and selective quantization in TF Quantizer PiperOrigin-RevId: 570257817
Thai Nguyen
thaink@google.com
1,696,303,446,000
970e3d0f79af7be6c5afda0a3d6e96b21393ee40
fix build
Kanvi Khanna
kanvi.khanna@intel.com
1,696,304,125,000
4696642aac7916fde443f1a7e8499bf81f0736b3
Refactor stream implementation so that it does not create a thread for each streaming request. The callback are immediately invoked at the producer side. And the producer can customize the threading using ThreadPoolInterface. PiperOrigin-RevId: 570293325
Kuangyuan Chen
chky@google.com
1,696,316,483,000
6a1f455f0097971e8799c62418c23ab2bf5c2f64
Internal Code Change PiperOrigin-RevId: 570296006
A. Unique TensorFlower
gardener@tensorflow.org
1,696,317,199,000
1d6d6ee9cb43656f7111648044daaaac4558418c
NFC: replace uses of TF_DISALLOW_COPY_AND_ASSIGN/SE_DISALLOW_COPY_AND_ASSIGN macros with implementation. PiperOrigin-RevId: 570296712
Christian Sigg
csigg@google.com
1,696,317,440,000
444f92b2aafd80f70400b2ba4152490fb7d43e6e
Merge pull request #61997 from tensorflow:sushreebarsa-patch-5 PiperOrigin-RevId: 570311080
TensorFlower Gardener
gardener@tensorflow.org
1,696,322,095,000
0fbad8e3b21a08fafe76f0d414740fc756e4dc22
Merge pull request #61970 from tensorflow:sushreebarsa-patch-2 PiperOrigin-RevId: 570311614
TensorFlower Gardener
gardener@tensorflow.org
1,696,322,426,000
3122d1ae58ca9ec4cb9b4d78ea16b135b0f6440b
Merge pull request #61952 from tensorflow:sushreebarsa-patch-1 PiperOrigin-RevId: 570312042
TensorFlower Gardener
gardener@tensorflow.org
1,696,322,825,000
043c8fc649a4929a2ca76b92ecb9ab08fc6cebc2
Update GraphDef version to 1638. PiperOrigin-RevId: 570316698
A. Unique TensorFlower
gardener@tensorflow.org
1,696,323,713,000
56603b2d2063b887e07f4038645c81972ae3afc7
compat: Update forward compatibility horizon to 2023-10-03 PiperOrigin-RevId: 570316707
A. Unique TensorFlower
gardener@tensorflow.org
1,696,323,716,000
b492435349495fbab5e2ed6095f03e81ff9be17a
Internal Code Change PiperOrigin-RevId: 570326337
A. Unique TensorFlower
gardener@tensorflow.org
1,696,326,802,000
e9b95607a91db07ad5483a3b84b164fbf30cb87d
add gif patch
Eli Kobrin
kobrineli@ispras.ru
1,696,335,081,000
34bd3493bbc213ed56e59448969b6358ffa1b7f8
fix patch
Eli Kobrin
kobrineli@ispras.ru
1,696,335,289,000
4d871f75eb339e259263c41e38de40f928a30205
fix
Eli Kobrin
kobrineli@ispras.ru
1,696,335,345,000