| Commit message (Collapse) | Author | Age |
|
|
|
|
|
| |
Support peephole and num_proj as well.
PiperOrigin-RevId: 216467578
|
|
|
|
| |
PiperOrigin-RevId: 216392908
|
|
|
|
|
|
|
| |
The conversion process fails for graphs that use tf.boolean_mask(..., axis=0) --
this op calls tf.concat with an empty array as the first argument.
PiperOrigin-RevId: 214451470
|
|
|
|
| |
PiperOrigin-RevId: 213729750
|
|
|
|
| |
PiperOrigin-RevId: 213227615
|
|
|
|
| |
PiperOrigin-RevId: 211874785
|
|
|
|
| |
PiperOrigin-RevId: 211159438
|
|
|
|
| |
PiperOrigin-RevId: 210507220
|
|
|
|
| |
PiperOrigin-RevId: 210051131
|
|
|
|
| |
PiperOrigin-RevId: 207917321
|
|
|
|
| |
PiperOrigin-RevId: 206549781
|
|
|
|
| |
PiperOrigin-RevId: 206185190
|
|
|
|
| |
PiperOrigin-RevId: 206012140
|
|
|
|
| |
PiperOrigin-RevId: 205134621
|
|
|
|
| |
PiperOrigin-RevId: 205122011
|
|
|
|
| |
PiperOrigin-RevId: 205111208
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Stack is not Pack. tf.stack() yields Pack ops. Pack ops stack tensors. Stack ops manipulate the TF runtime stack.
This cl unaliases "Stack" and "Pack" ops in toco, and renames most things that refer to "Stack" ops to "Pack" to be consistent across the codebase.
In summary:
Stack is whack. 'Stack' should be 'Pack'. Hack 'Stack's into 'Pack's like a maniac. This keeps 'Stack's from wracking runtime graphs.
(We apologize for the fault in the change description. Those responsible have been...er...sacked).
PiperOrigin-RevId: 204951155
|
|
|
|
| |
PiperOrigin-RevId: 204846139
|
|
|
|
| |
PiperOrigin-RevId: 203908601
|
|
|
|
| |
PiperOrigin-RevId: 203027634
|
|
|
|
|
|
|
|
|
|
| |
implementation.
Due to an issue with negative StridedSlice indices in TensorFlow, the end indices can specify degenerate slices when negative indices are used to shrink an axis (e.g. for tf.range(4)[-1], start is -1, end is 0, and stride is 1). This fix works around the issue by ignoring stop indices entirely when an axis is shrinking, since in order to be shrunk the length is by definition 1.
Fixes Issue #19260.
PiperOrigin-RevId: 202398678
|
|
|
|
|
|
|
|
|
|
|
| |
constant value 0 and already have their shape set from the output
activations shape; instead of having it create dummy placeholders
and relying on PropagateFixedSizes to create the constant array.
Rationale: It wasn't PropagateFixedSizes's job to create constant
arrays, and that broke down in a case where the bias vectors not
being constant prevented FuseBinaryIntoPrecedingAffine from running.
PiperOrigin-RevId: 202365030
|
|
|
|
| |
PiperOrigin-RevId: 202262513
|
|
|
|
| |
PiperOrigin-RevId: 202130585
|
|
|
|
|
|
|
|
|
|
|
| |
constant value 0 and already have their shape set from the output
activations shape; instead of having it create dummy placeholders
and relying on PropagateFixedSizes to create the constant array.
Rationale: It wasn't PropagateFixedSizes's job to create constant
arrays, and that broke down in a case where the bias vectors not
being constant prevented FuseBinaryIntoPrecedingAffine from running.
PiperOrigin-RevId: 202120850
|
|
|
|
| |
PiperOrigin-RevId: 201422566
|
|
|
|
|
|
| |
properly check the necessary precondition that the input depth is 1.
PiperOrigin-RevId: 201389819
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
200617269 by A. Unique TensorFlower:
Internal change
--
200603378 by jpienaar:
The output of the merge should be the value's and not the original output port.
The output port of the IfOp is already taken into account by selecting the
merge node and the output of the merge should be the value used (which is the 0th
output of the merge node).
--
200601721 by A. Unique TensorFlower:
Basic support for tf.tile that multiplies a single axis.
--
200600686 by A. Unique TensorFlower:
Internal change.
--
PiperOrigin-RevId: 200617269
|
|
|
|
|
|
| |
similar to conv.
PiperOrigin-RevId: 200592004
|
|
|
|
| |
PiperOrigin-RevId: 199839745
|
|
|
|
| |
PiperOrigin-RevId: 199602232
|
|
|
|
| |
PiperOrigin-RevId: 198710452
|
|
|
|
| |
PiperOrigin-RevId: 196062186
|
|
|
|
| |
PiperOrigin-RevId: 195734246
|
|
|
|
| |
PiperOrigin-RevId: 195704492
|
|
|
|
| |
PiperOrigin-RevId: 195503894
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
1. Fix ResolveBatchNormalization to avoid deleting arrays that may still be
used.
2. Correctly count the number of ops using a given array, even when some ops
use the same array as more than one of their inputs.
3. In PropagateFixedSizes for Concatenation ops, when resolving a -1 wildcard
to a fixed value, we were doing so in a local 'axis' variable without actually
updating op->axis! The resulting -1 value still in op->axis tripped runtime code,
causing the concatenation to misbehave during inference.
PiperOrigin-RevId: 195454037
|
|
|
|
| |
PiperOrigin-RevId: 194997009
|
|
|
|
| |
PiperOrigin-RevId: 194446000
|
|
|
|
| |
PiperOrigin-RevId: 194200078
|
|
|
|
| |
PiperOrigin-RevId: 193977375
|
|
|
|
| |
PiperOrigin-RevId: 193659701
|
|
|
|
|
|
| |
correctly.
PiperOrigin-RevId: 193245539
|
|
|
|
| |
PiperOrigin-RevId: 192511036
|
|
|
|
|
|
| |
run time support.
PiperOrigin-RevId: 192135843
|
|
|
|
| |
PiperOrigin-RevId: 191604311
|
|
|
|
|
|
| |
RandomUniform, and constant resolution with static seeds.
PiperOrigin-RevId: 191293897
|
|
|
|
| |
PiperOrigin-RevId: 190097557
|
|
|
|
|
|
| |
implementation.
PiperOrigin-RevId: 188210975
|
|
|
|
|
|
| |
gather.
PiperOrigin-RevId: 187241089
|