OSSForks/tensorflow
1
0
Fork 0
mirror of https://github.com/zebrajr/tensorflow.git synced 2025-12-06 00:19:58 +01:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity
107,707 Commits 50 Branches 144 Tags 1.3 GiB
0d1805aede
Commit Graph

107707 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Mihai Maruseac
0d1805aede
Merge pull request #48488 from tensorflow/geetachavan1-patch-1 
Fixing typo for estimator
 2021-04-12 09:43:40 -07:00
geetachavan1
67379de5cd
Fixing typo for estimator 2021-04-12 09:40:07 -07:00
Mihai Maruseac
9d2013c5dd
Merge pull request #48348 from geetachavan1/cherrypicks_VI88X 
[CHerryPick:r2.5] Include additional required headers in Windows libtensorflow releases.
 2021-04-09 13:12:31 -07:00
Mihai Maruseac
f448ad7a26
Merge branch 'r2.5' into cherrypicks_VI88X 2021-04-09 13:11:44 -07:00
Mihai Maruseac
fadc35d749
Merge pull request #48283 from reedwm/cherrypick-sparse-to-dense 
2.5-rc1 cherry-pick request: Disable GPU SparseToDense
 2021-04-09 13:09:20 -07:00
Mihai Maruseac
6b3140c1ca
Merge pull request #48315 from lgeiger/r2.5-fix-callback-logs 
[r2.5 cherry-pick] Fix Keras Callbacks logs / numpy_logs sync
 2021-04-09 13:09:02 -07:00
Mihai Maruseac
43abad5bb9
Merge pull request #48335 from geetachavan1/cherrypicks_OHAKM 
[CherryPick:r2.5] [TF/XLA] Fix the CollectiveReduceV2Op lowering by specifying the number of replicas needed at compile-time
 2021-04-09 13:08:36 -07:00
Mihai Maruseac
9a38fc098d
Merge pull request #48339 from geetachavan1/cherrypicks_N3KGF 
[CherryPick:r2.5] Rollback of rollback of [TF/XLA] Make sure that XLAControlFlowContext is injected on all paths of jit_compile=True
 2021-04-09 13:08:28 -07:00
Mihai Maruseac
7095771f55
Merge pull request #48341 from geetachavan1/cherrypicks_CKQN6 
[CherryPick:r2.5][tf.data] Apply options for a dataset before it is copied to another device in `tf.data.experimental.copy_to_device()`.
 2021-04-09 13:07:40 -07:00
Mihai Maruseac
709a48480c
Merge pull request #48345 from penpornk/cherrypicks_TKRV2 
r2.5-rc1 cherry-pick request: [Intel MKL] Update oneDNN to v2.2 official release
 2021-04-09 13:07:32 -07:00
Mihai Maruseac
e65805e7e3
Merge pull request #48347 from penpornk/cherrypicks_UH1A1 
r2.5-rc1 cherry-pick request: [Intel MKL] Fix MKL auto-mixed precision grappler pass in --config=cuda
 2021-04-09 13:07:23 -07:00
Mihai Maruseac
1143efac78
Merge pull request #48384 from ckkuang/cherrypicks_P5FG5 
release 2.5-rc1 cherry-pick request: Fix MirroredStrategy eager performance regression
 2021-04-09 13:06:58 -07:00
Mihai Maruseac
ac576d2209
Merge pull request #48387 from penpornk/cherrypicks_591U2 
r2.5-rc1 cherry-pick request: Upgrade curl to 7.76.0
 2021-04-09 13:06:48 -07:00
Mihai Maruseac
d443542d1e
Merge pull request #48417 from tensorflow/libtf_win_25 
[r2.5] Adding missing Windows libtensorflow files.
 2021-04-09 13:06:29 -07:00
Mihai Maruseac
abc0deff1d
Merge pull request #48437 from tensorflow-jenkins/version-numbers-2.5.0rc1-5980 
Update version numbers for TensorFlow 2.5.0-rc1
 2021-04-09 11:38:39 -07:00
TensorFlow Release Automation
a127e77415 Update version numbers to 2.5.0-rc1 2021-04-09 11:19:58 -07:00
Thai Nguyen
e78e4d13fc
Merge pull request #48379 from geetachavan1/cherrypicks_7A5B4 
[CherryPick:r2.5] Update runtime version to 2.5.0
 2021-04-09 10:25:21 +09:00
Amit Patankar
bd29388cdd Adding missing Windows libtensorflow files. 2021-04-08 11:42:44 -07:00
Rama ketineni
6dde44561d Update curl.BUILD 
Added "include/curl/options.h"  to hdrs list
 2021-04-07 18:46:19 -07:00
Rama Ketineni
c9d2c84508 added lib/strtok.c 2021-04-07 18:46:14 -07:00
Rama Ketineni
f5f18079cc curl is upgraded to 7.76.0 from 7.69 due to required CVE fixes. 2021-04-07 18:46:08 -07:00
Chenkai Kuang
fcda86f6b8 Use merge_call for optimizer distributed_apply if strategy doesn't support merge_call free path. 
In principle even if the strategy doesn't support merge_call free path, distributed_apply can still be called w/o merge_call. However it would be slower than calling it with merge_call in pure eager mode, mainly because MS uses multiple python threads to run replica function, and those python threads are running sequentially.

PiperOrigin-RevId: 367305737
Change-Id: Ie2316fd62131c60eff605cf9f51e044c09ec9abc
 2021-04-07 16:44:24 -07:00
Thai Nguyen
30b4378272 [CherryPick:r2.5] Update runtime version to 2.5.0 2021-04-07 11:01:19 -07:00
Amit Patankar
08ad4f402a Include additional required headers in Windows libtensorflow releases. 
PiperOrigin-RevId: 366494907
Change-Id: I3abc6f432549275abb8a57ba0f2a5a5ed27250ff
 2021-04-06 15:25:43 -07:00
Mahmoud Abuzaina
1d09bdd474 Adding #ifdef INTEL_MKL to avoid MAC build issue 2021-04-06 15:09:21 -07:00
Mahmoud Abuzaina
bf3bba114c Enabled MKL auto mixed precision in the single binary build 2021-04-06 15:09:21 -07:00
Jiri Simsa
aa08128321 [tf.data] Apply options for a dataset before it is copied to another device in tf.data.experimental.copy_to_device(). 
PiperOrigin-RevId: 366293759
Change-Id: I2f38afd28f448cbd5a731a4a0261d74b3cc4ad0c
 2021-04-06 12:55:15 -07:00
Penporn Koanantakool
99ef304f77 Fix the mirror package path. 2021-04-06 12:54:45 -07:00
Xiaoming (Jason) Cui
d361ef99b2 [INTEL MKL] update oneDNN version in TensorFlow to v2.2 official release 2021-04-06 12:52:14 -07:00
George Karpenkov
4e2da93a96 Rollback of rollback of [TF/XLA] Make sure that XLAControlFlowContext is injected on all paths of jit_compile=True 
PiperOrigin-RevId: 366389161
Change-Id: If4aca391be05315ccc875ac43acdcbc70bc64c67
 2021-04-06 12:44:15 -07:00
George Karpenkov
c4d24e402f [TF/XLA] Fix the CollectiveReduceV2Op lowering by specifying the number of replicas needed at compile-time 
PiperOrigin-RevId: 365890700
Change-Id: I8c213c4ed469d7307d0a8be99a1a32bc6e06a1da
 2021-04-06 10:27:56 -07:00
Lukas Geiger
6bb50433ee Only convert logs if batch hooks do not support TF logs 
This is a small performance optimization that prevents conversion if not
necessary.
 2021-04-05 22:15:54 +02:00
Lukas Geiger
cbbd1cd765 Fix Keras Callbacks logs sync 2021-04-05 22:15:46 +02:00
Mihai Maruseac
fb37439d64
Merge pull request #48284 from tensorflow/geetachavan1-patch-1 
Update Estimator Version to 2.5.0-rc0
 2021-04-02 14:32:54 -07:00
Sanjoy Das
15c594ae6d Disable SparseToDense (it fails an internal test) 
PiperOrigin-RevId: 366099814
Change-Id: I13b5f43aae4155b7d61439671c560930eed4cc29
 2021-04-02 11:01:36 -07:00
geetachavan1
ecf0809b8f
Update setup.py 2021-04-02 10:48:28 -07:00
Goldie Gadde
a8b6d5ff93
Merge pull request #48222 from tensorflow/mm-fix-fileystem-on-r2.5 
Fix GCS filesystem
 2021-03-31 21:24:10 -07:00
Mihai Maruseac
b9e31e669e Fix typo/logic bug in modular plugins. 
We want to enable plugins only if the environment variable exists!

CC @yongtang, @kvignesh1420, @vnvo2409.

PiperOrigin-RevId: 366056903
Change-Id: Ia6f0918c97cbd6e064e19a02ce3a7d46f48b2799
 2021-03-31 14:58:02 -07:00
Mihai Maruseac
158505edc3 Switch TF filesystems to keep backwards compatibility. 
Recent changes to enable modular TF can result in breakages or require majority of users to take quick steps to fix their broken code. Instead, we should enable the modular filsystems only if requested by an environment variable and use the legacy behavior in all cases.

In a future TF release, the cloud filesystems will come from different pip packages but we have to design the path towards that to ensure that no one in the TF ecosystem is broken by this. As we stand now, without this change most TF users will need to take action in a P0 basis and this is not desirable.

CC @yongtang, @kvignesh1420, @vnvo2409. Note that I also changed the name of the envvar to match the logic and you'll have to do the same for SIG IO code. Let's sync on SIG IO meeting to discuss further

PiperOrigin-RevId: 365940168
Change-Id: Id03b793613e521999af47fd6017b70ff659bd23b
 2021-03-31 14:56:14 -07:00
Mihai Maruseac
96dfa5c37c
Merge pull request #48107 from tensorflow/mihaimaruseac-patch-1 
Fix typo in setup.py
 2021-03-26 07:19:50 -07:00
Mihai Maruseac
5f7fd89ac8
Fix typo in setup.py 2021-03-26 07:19:19 -07:00
Mihai Maruseac
f8b5b9b1fc
Merge pull request #48093 from tensorflow/mihaimaruseac-patch-1 
Pin estimator
 2021-03-25 18:03:20 -07:00
Mihai Maruseac
b84dac5ea8
Update setup.py 2021-03-25 18:02:39 -07:00
Mihai Maruseac
b42047d765
Merge pull request #48091 from tensorflow-jenkins/version-numbers-2.5.0rc0-30114 
Update version numbers for TensorFlow 2.5.0-rc0
 2021-03-25 17:58:08 -07:00
TensorFlow Release Automation
1d4885ba37 Update version numbers to 2.5.0-rc0 2021-03-25 17:54:51 -07:00
TensorFlower Gardener
6af429790a Merge pull request #48082 from njeffrie:f1_depthwise 
PiperOrigin-RevId: 365150386
Change-Id: I0cca1c34d2e600cea51ef45b37e5cf808891541a
 2021-03-25 16:41:05 -07:00
Mark Daoust
fc7529576a Move TFRecord examples to the exported class. 
PiperOrigin-RevId: 365149622
Change-Id: Ib7dfba397b5c2a9757cfd14d8a2537f38dadf35d
 2021-03-25 16:36:10 -07:00
Wenhao Jia
054d641e30 Minor build fix. 
PiperOrigin-RevId: 365148648
Change-Id: Ib0bd821dfaa7943d04e6da397bfd9a4942129536
 2021-03-25 16:30:48 -07:00
Sachin Joglekar
b8e4404f60 Simplify FP16 delegation: accept FP16 DEQUANT nodes only if the entire graph will be delegated, else leave the FP16 DEQUANTs (which are no-ops after first inference) on CPU. Also modify unit tests to reflect this. 
PiperOrigin-RevId: 365143125
Change-Id: Id663b84e581db06d0e904a4d59e52a54e8824057
 2021-03-25 16:07:39 -07:00
Allen Lavoie
2ed50f444f Prefer namedtuple properties to wrapper properties in tf.Module tracking logic 
Avoids trying to simplify the logic in the process, since apparently we still need the workarounds.

PiperOrigin-RevId: 365142662
Change-Id: I63f57b0480e67c8c26cd1b196fffc1eca8b0e9d8
 2021-03-25 16:03:40 -07:00