Skip to content

Commit e352077

Browse files
yanislavdgeorgepaw
yanislavd
authored and
georgepaw
committed
Fix API changes [tf2]
Summary: TF2.5 Only Reviewers: grahamh, #tensorflow, #framework_ip_review_-_any_oss_or_third-party_code_use_has_been_approved, jamiep, georgep Reviewed By: #tensorflow, #framework_ip_review_-_any_oss_or_third-party_code_use_has_been_approved, georgep Subscribers: jamiep Maniphest Tasks: T58548 Differential Revision: https://phabricator.sourcevertex.net/D63453
1 parent 4eb4957 commit e352077

File tree

1 file changed

+60
-45
lines changed

1 file changed

+60
-45
lines changed

‎tensorflow/compiler/plugin/poplar/docs/api-changes.rst‎

Lines changed: 60 additions & 45 deletions
Original file line numberDiff line numberDiff line change
@@ -41,11 +41,11 @@ Removal of deprecated APIs
4141
:py:class:`~tensorflow.python.ipu.horovod.popdist_strategy.PopDistStrategy`.
4242
- The `save_interval_report` flag provided to `TF_POPLAR_FLAGS` has been
4343
removed following deprecation in Poplar SDK version 2.4. From Poplar SDK
44-
version 2.5, LIBPVA should be used.
44+
version 2.5, libpva should be used.
4545
- `tensorflow.python.ipu.ops.summary_ops` has been removed as both
4646
`ipu_compile_summary` and `get_ipu_reports` were deprecated in Poplar
4747
SDK version 2.4. From Poplar SDK 2.5 onwards, the
48-
[PopVision] (https://docs.graphcore.ai/projects/tf-model-parallelism/en/latest/profiler.html)
48+
`PopVision <https://docs.graphcore.ai/en/latest/software.html#profiling-and-debugging>`_
4949
suite of tools should be used instead. Additionally, the `compile_summary`
5050
parameter of `IPURunConfig` has been removed.
5151

@@ -76,45 +76,60 @@ release.
7676

7777
The table below lists all of the deprecated layers, and their new locations:
7878

79-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
80-
| **TensorFlow** | **IPU TensorFlow Addons** |
81-
+===========================================================================================+===========================================================================================+
82-
| tensorflow.python.ipu.keras.layers.assume_equal_across_replicas.AssumeEqualAcrossReplicas | ipu_tensorflow_addons.keras.layers.assume_equal_across_replicas.AssumeEqualAcrossReplicas |
83-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
84-
| tensorflow.python.ipu.keras.layers.ctc.CTCInferenceLayer | ipu_tensorflow_addons.keras.layers.ctc.CTCInferenceLayer |
85-
| tensorflow.python.ipu.keras.layers.ctc.CTCPredictionsLayer | ipu_tensorflow_addons.keras.layers.ctc.CTCPredictionsLayer |
86-
| tensorflow.python.ipu.keras.**losses**.ctc.CTCLoss | ipu_tensorflow_addons.keras.**layers**.ctc.CTCLoss |
87-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
88-
| tensorflow.python.ipu.keras.layers.dropout.Dropout | ipu_tensorflow_addons.keras.layers.dropout.Dropout |
89-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
90-
| tensorflow.python.ipu.keras.layers.effective_transformer.EffectiveTransformer | ipu_tensorflow_addons.keras.layers.effective_transformer.EffectiveTransformer |
91-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
92-
| tensorflow.python.ipu.keras.layers.embedding_lookup.Embedding | ipu_tensorflow_addons.keras.layers.embedding_lookup.Embedding |
93-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
94-
| tensorflow.python.ipu.keras.layers.ipu_layer.IPULayer | ipu_tensorflow_addons.keras.layers.ipu_layer.IPULayer |
95-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
96-
| tensorflow.python.ipu.keras.layers.**math**.SerialDense | ipu_tensorflow_addons.keras.layers.**dense**.SerialDense |
97-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
98-
| tensorflow.python.ipu.keras.layers.normalization.GroupNorm | ipu_tensorflow_addons.keras.layers.normalization.GroupNorm |
99-
| tensorflow.python.ipu.keras.layers.normalization.GroupNormalization | ipu_tensorflow_addons.keras.layers.normalization.GroupNormalization |
100-
| tensorflow.python.ipu.keras.layers.normalization.InstanceNorm | ipu_tensorflow_addons.keras.layers.normalization.InstanceNorm |
101-
| tensorflow.python.ipu.keras.layers.normalization.InstanceNormalization | ipu_tensorflow_addons.keras.layers.normalization.InstanceNormalization |
102-
| tensorflow.python.ipu.keras.layers.normalization.LayerNorm | ipu_tensorflow_addons.keras.layers.normalization.LayerNorm |
103-
| tensorflow.python.ipu.keras.layers.normalization.LayerNormalization | ipu_tensorflow_addons.keras.layers.normalization.LayerNormalization |
104-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
105-
| tensorflow.python.ipu.keras.layers.recomputation.RecomputationCheckpoint | ipu_tensorflow_addons.keras.layers.recomputation.RecomputationCheckpoint |
106-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
107-
| tensorflow.python.ipu.keras.layers.rnn.GRU | ipu_tensorflow_addons.keras.layers.rnn.GRU |
108-
| tensorflow.python.ipu.keras.layers.rnn.LSTM | ipu_tensorflow_addons.keras.layers.rnn.LSTM |
109-
| tensorflow.python.ipu.keras.layers.rnn.PopnnGRU | ipu_tensorflow_addons.keras.layers.rnn.PopnnGRU |
110-
| tensorflow.python.ipu.keras.layers.rnn.PopnnLSTM | ipu_tensorflow_addons.keras.layers.rnn.PopnnLSTM |
111-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
112-
| tensorflow.python.ipu.ops.rnn_ops.PopnnAUGRU | ipu_tensorflow_addons.v1.layers.rnn_ops.PopnnAUGRU |
113-
| tensorflow.python.ipu.ops.rnn_ops.PopnnDynamicGRU | ipu_tensorflow_addons.v1.layers.rnn_ops.PopnnDynamicGRU |
114-
| tensorflow.python.ipu.ops.rnn_ops.PopnnDynamicLSTM | ipu_tensorflow_addons.v1.layers.rnn_ops.PopnnDynamicLSTM |
115-
| tensorflow.python.ipu.ops.rnn_ops.PopnnGRU | ipu_tensorflow_addons.v1.layers.rnn_ops.PopnnGRU |
116-
| tensorflow.python.ipu.ops.rnn_ops.PopnnLSTM | ipu_tensorflow_addons.v1.layers.rnn_ops.PopnnLSTM |
117-
+-------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------+
79+
.. table:: Mapping from TensorFlow and IPU Addons
80+
:class: nostripe
81+
82+
+-------------------------------+---------------------------------------------------------------------+
83+
| **TensorFlow** | ``tensorflow.python.ipu.keras.layers`` |
84+
+-------------------------------+---------------------------------------------------------------------+
85+
| **IPU Addons** | ``ipu_tensorflow_addons.keras.layers`` |
86+
+-------------------------------+---------------------------------------------------------------------+
87+
| ``AssumeEqualAcrossReplicas`` |
88+
+-------------------------------+---------------------------------------------------------------------+
89+
| **TensorFlow** | ``tensorflow.python.ipu.keras.losses`` |
90+
+-------------------------------+---------------------------------------------------------------------+
91+
| **IPU Addons** | ``ipu_tensorflow_addons.keras.layers`` |
92+
+-------------------------------+---------------------------------------------------------------------+
93+
| ``CTCLoss`` |
94+
+-------------------------------+---------------------------------------------------------------------+
95+
| **TensorFlow** | ``tensorflow.python.ipu.keras.layers`` |
96+
+-------------------------------+---------------------------------------------------------------------+
97+
| **IPU Addons** | ``ipu_tensorflow_addons.keras.layers`` |
98+
+-------------------------------+---------------------------------------------------------------------+
99+
| | ``CTCInferenceLayer`` |
100+
| | ``CTCPredictionsLayer`` |
101+
| | ``Dropout`` |
102+
| | ``EffectiveTransformer`` |
103+
| | ``Embedding`` |
104+
| | ``GroupNorm`` |
105+
| | ``GroupNormalization`` |
106+
| | ``GRU`` |
107+
| | ``InstanceNorm`` |
108+
| | ``InstanceNormalization`` |
109+
| | ``LayerNorm`` |
110+
| | ``LayerNormalization`` |
111+
| | ``LSTM`` |
112+
| | ``PopnnGRU`` |
113+
| | ``PopnnLSTM`` |
114+
| | ``RecomputationCheckpoint`` |
115+
| | ``SerialDense`` |
116+
+-------------------------------+---------------------------------------------------------------------+
117+
| **TensorFlow** | ``tensorflow.python.ipu.keras.layers.ipu_layer`` |
118+
+-------------------------------+---------------------------------------------------------------------+
119+
| **IPU Addons** | ``ipu_tensorflow_addons.keras.layers.ipu_layer`` |
120+
+-------------------------------+---------------------------------------------------------------------+
121+
| ``IPULayer`` |
122+
+-------------------------------+---------------------------------------------------------------------+
123+
| **TensorFlow** | ``tensorflow.python.ipu.ops.rnn_ops`` |
124+
+-------------------------------+---------------------------------------------------------------------+
125+
| **IPU Addons** | ``ipu_tensorflow_addons.v1.layers`` |
126+
+-------------------------------+---------------------------------------------------------------------+
127+
| | ``PopnnAUGRU`` |
128+
| | ``PopnnDynamicGRU`` |
129+
| | ``PopnnDynamicLSTM`` |
130+
| | ``PopnnGRU`` |
131+
| | ``PopnnLSTM`` |
132+
+-----------------------------------------------------------------------------------------------------+
118133

119134
.. deprecated_pipeline_ga_options:
120135
@@ -123,9 +138,9 @@ Deprecated pipeline and gradient_accumulation options
123138

124139
The `experimental_normalize_gradients` parameter when setting gradient accumulation options
125140
(`Model.set_gradient_accumulation_options()`) and pipeline options (`Model.set_pipelining_options()`)
126-
has been deprecated and will be removed in future. This feature will be provided by new mean reduction
127-
methods being added via the `reduction_method` parameter
128-
(`tensorflow.python.ipu.optimizers.GradientAccumulationReductionMethod`).
141+
has been deprecated and will be removed in future. This feature is now provided
142+
by new mean reduction methods (see:py:class:`~tensorflow.python.ipu.optimizers.GradientAccumulationReductionMethod`)
143+
that are selected using the `reduction_method` parameter in `Model.set_gradient_accumulation_options()` and the `gradient_accumulation_reduction_method` parameter in `Model.set_pipelining_options()`.
129144

130145
RNN available_memory_proportion_fwd/available_memory_proportion_bwd deprecated
131146
''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
@@ -277,7 +292,7 @@ ____________________
277292
- `IPUMultiWorkerStrategy` is deprecated. Using `IPUMultiWorkerStrategy`
278293
will trigger a deprecation warning.
279294
- The flag `save_interval_report` from `TF_POPLAR_FLAGS` is now deprecated. Please
280-
use LIBPVA instead.
295+
use libpva instead.
281296

282297

283298
Release 2.3

0 commit comments

Comments
 (0)