- 10 12月, 2016 26 次提交
-
-
由 Justine Tunney 提交于
Change: 141613129
-
由 Vinu Rajashekhar 提交于
Change: 141612098
-
由 Justine Tunney 提交于
Change: 141611875
-
由 Jonathan Hseu 提交于
Change: 141611156
-
由 A. Unique TensorFlower 提交于
Change: 141605210
-
由 Ian Langmore 提交于
Must compute dimension size explicitly rather than using [-1] in order to deal with all multidimensional Tensors with dimensions after first equal to zero. Change: 141601835
-
由 Justine Tunney 提交于
Change: 141601412
-
由 Justine Tunney 提交于
Change: 141601347
-
由 Vijay Vasudevan 提交于
to do so correctly. Change: 141599116
-
由 A. Unique TensorFlower 提交于
Change: 141592580
-
由 A. Unique TensorFlower 提交于
they cause test timeouts. Change: 141590858
-
由 A. Unique TensorFlower 提交于
and TensorArray.unstack since pack and unpack are getting deprecated. Also, I switched a few references to tf.pack/tf.unpack to tf.stack and tf.unstack. Change: 141589815
-
由 A. Unique TensorFlower 提交于
Change: 141583614
-
由 Joshua V. Dillon 提交于
Change: 141581989
-
由 Pete Warden 提交于
Change: 141580230
-
由 Vijay Vasudevan 提交于
since it is known to not be correct. The construction of these fused loss ops does not work with our tf.gradients interface when taking second derivatives (e.g., for hessians). This introduces an op that produces the identity on the forward pass, but has no gradient registered for the backward pass. So the first derivative passes, but the second one doesn't. This op (PreventGradient) may prove useful in the future, so it seems okay to add. Partially addresses #5876 in that it will no longer silently do the wrong thing, though supporting the second derivative for this function is something we want to do. Change: 141578951
-
由 A. Unique TensorFlower 提交于
Change: 141572862
-
由 A. Unique TensorFlower 提交于
Change: 141569831
-
由 A. Unique TensorFlower 提交于
Change: 141569747
-
由 A. Unique TensorFlower 提交于
Change: 141567499
-
由 A. Unique TensorFlower 提交于
Also throw an error when the low-rank update is ill-specified. Change: 141566486
-
由 A. Unique TensorFlower 提交于
Change: 141564248
-
由 A. Unique TensorFlower 提交于
Change: 141563688
-
由 Jianwei Xie 提交于
Change: 141559664
-
由 Peter Hawkins 提交于
Add missing zlib dependency to LLVM build file. Various other small cleanups. Change: 141557751
-
由 Dan Ringwalt 提交于
Change: 141555939
-
- 09 12月, 2016 14 次提交
-
-
由 A. Unique TensorFlower 提交于
Change: 141521572
-
由 A. Unique TensorFlower 提交于
Change: 141521124
-
由 Shanqing Cai 提交于
This check-in implements the following commands (and aliases) in the stepper CLI: * cont (ct, c) * step (st, s) * list_sorted_nodes (lt) * print_tensor (pt) * inject_value (inject, override) To activate the stepper CLI, enter command "step" or "s" at the run-start CLI. Not implemented yet: * Documentation in g3doc/how_tos/debugger will be submitted in a later CL. * Commands for inspecting node information and graph structure in the stepper CLI, including ni, li and lo. Change: 141518037
-
由 A. Unique TensorFlower 提交于
Change: 141516835
-
由 A. Unique TensorFlower 提交于
Change: 141516591
-
由 Andrew Harp 提交于
Additionally: - change single quotes to double quotes to make path rewriting easier - guard windows lib reference with PLATFORM_WINDOWS - fixed failing kmeans test Change: 141515942
-
由 A. Unique TensorFlower 提交于
Increasing the tolerance of a test involving in random start to avoid flakiness. There might be a change of numerical accuracy in the underlying library to be investigated and might be addressed in a follow up CL. Change: 141514160
-
由 Andrew Selle 提交于
(need to change from array_ops.reverse to gen_array_ops._reverse) Change: 141512722
-
由 Justine Tunney 提交于
This excludes of course test library rules. Change: 141507442
-
由 A. Unique TensorFlower 提交于
Change: 141504802
-
由 Neal Wu 提交于
Removed the "Run a TensorFlow demo model" section from the g3docs because it is no longer applicable after moving the tensorflow/models directory Change: 141504491
-
由 Vijay Vasudevan 提交于
device setter, and its setting is respected, and the jobs get placed on the approropriate tasks Change: 141504459
-
由 A. Unique TensorFlower 提交于
Change: 141504398
-
由 Neal Wu 提交于
Moved tensorflow/models to models/tutorials and replaced all tutorial references to tensorflow/models Change: 141503531
-