Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support TF GNR-BASE #1415

Merged
merged 12 commits into from
Nov 28, 2023
Merged

Support TF GNR-BASE #1415

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
d536da2
add tf version number of gnr-base
zehao-intel Sep 13, 2023
755f201
remove estimator check
zehao-intel Sep 25, 2023
145f9f3
Use function_optimization in aggresive mode to register swish_f32
zehao-intel Oct 17, 2023
be42f35
Update gnr-base for WW41 build
zehao-intel Oct 22, 2023
4fa1f02
Update gnr-base version for tf_utils/util.py
zehao-intel Nov 7, 2023
b2e9239
Support int8 reshape and transpose ops
zehao-intel Nov 17, 2023
76b23e3
remove int8 reshape and transpose
zehao-intel Nov 24, 2023
8bf9129
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Nov 24, 2023
258d25b
fix estimator ut
zehao-intel Nov 24, 2023
bb8d03c
Merge branch 'zehao/gnr-base' of https://github.com/intel/neural-comp…
zehao-intel Nov 24, 2023
c54f69a
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Nov 24, 2023
ee3865f
Merge branch 'master' into zehao/gnr-base
zehao-intel Nov 26, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions examples/tensorflow/oob_models/quantization/ptq/tf_benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,6 +94,8 @@ def create_tf_config(args):
config.allow_soft_placement = True
config.intra_op_parallelism_threads = OMP_NUM_THREADS
config.inter_op_parallelism_threads = 1
# additional options
config.graph_options.rewrite_options.function_optimization = rewriter_config_pb2.RewriterConfig.AGGRESSIVE
if args.precision == 'bfloat16':
config.graph_options.rewrite_options.auto_mixed_precision_mkl = rewriter_config_pb2.RewriterConfig.ON
return config
Expand Down
20 changes: 18 additions & 2 deletions neural_compressor/adaptor/tensorflow.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,15 @@
from .query import QueryBackendCapability

tensorflow = LazyImport("tensorflow")
spr_base_verions = ("2.11.0202242", "2.11.0202250", "2.11.0202317", "2.11.0202323")
spr_base_verions = (
"2.11.0202242",
"2.11.0202250",
"2.11.0202317",
"2.11.0202323",
"2.14.0202335",
"2.14.dev202335",
"2.15.0202341",
)


@adaptor_registry
Expand Down Expand Up @@ -2146,7 +2154,15 @@ def _compare(version1, version2):
if self.version in sub_data["version"]["name"]:
return sub_data
else:
if sub_data["version"]["name"] == ["2.11.0202242", "2.11.0202250", "2.11.0202317", "2.11.0202323"]:
if sub_data["version"]["name"] == [
"2.11.0202242",
"2.11.0202250",
"2.11.0202317",
"2.11.0202323",
"2.14.0202335",
"2.14.dev202335",
"2.15.0202341",
]:
continue
sorted_list = copy.deepcopy(sub_data["version"]["name"])
sorted_list.remove("default") if "default" in sorted_list else None
Expand Down
2 changes: 1 addition & 1 deletion neural_compressor/adaptor/tensorflow.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@
---
-
version:
name: ['2.11.0202242', '2.11.0202250', '2.11.0202317', '2.11.0202323']
name: ['2.11.0202242', '2.11.0202250', '2.11.0202317', '2.11.0202323', '2.14.0202335', '2.14.dev202335', '2.15.0202341']

bf16: ["_MklLayerNorm", "Conv2D", "Conv2DBackpropFilter", "Conv2DBackpropInput", "Conv3D", "Conv3DBackpropFilterV2", "Conv3DBackpropInputV2",
"DepthwiseConv2dNative", "DepthwiseConv2dNativeBackpropFilter", "DepthwiseConv2dNativeBackpropInput", "GRUBlockCell",
Expand Down
2 changes: 1 addition & 1 deletion neural_compressor/adaptor/tf_utils/graph_rewriter/bf16/bf16_convert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,9 +31,9 @@
from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper as Helper
from neural_compressor.adaptor.tf_utils.util import TF_SPR_BASE_VERSIONS

from ..generic.dequantize_cast_optimizer import DequantizeCastOptimizer
from ..generic.graph_cse_optimizer import GraphCseOptimizer
from ..graph_base import GraphRewriterBase
from .dequantize_cast_optimizer import DequantizeCastOptimizer

DT_FLOAT32 = attr_value_pb2.AttrValue(type=dtypes.float32.as_datatype_enum)
DT_BFLOAT16 = attr_value_pb2.AttrValue(type=dtypes.bfloat16.as_datatype_enum)
Expand Down
0 ...iter/generic/dequantize_cast_optimizer.py → ...ewriter/bf16/dequantize_cast_optimizer.py
View file
Open in desktop
File renamed without changes.
3 changes: 2 additions & 1 deletion neural_compressor/adaptor/tf_utils/graph_rewriter/generic/convert_add_to_biasadd.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@

from neural_compressor.adaptor.tf_utils.graph_util import GraphAnalyzer
from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper as Helper
from neural_compressor.adaptor.tf_utils.util import TF_SPR_BASE_VERSIONS
from neural_compressor.utils.utility import dump_elapsed_time

from ..graph_base import GraphRewriterBase
Expand All @@ -38,7 +39,7 @@ def do_transformation(self):

import tensorflow as tf

if tf.version.VERSION not in ("2.11.0202242", "2.11.0202250", "2.11.0202317", "2.11.0202323"):
if tf.version.VERSION not in TF_SPR_BASE_VERSIONS:
target_nodes = g.query_fusion_pattern_nodes([["MatMul", "Conv2D"], ["Add", "AddV2"]])
else:
target_nodes = g.query_fusion_pattern_nodes([["MatMul"], ["Add", "AddV2"]])
Expand Down
10 changes: 9 additions & 1 deletion neural_compressor/adaptor/tf_utils/util.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,15 @@

from .graph_util import GraphAnalyzer, GraphRewriterHelper

TF_SPR_BASE_VERSIONS = ("2.11.0202242", "2.11.0202250", "2.11.0202317", "2.11.0202323")
TF_SPR_BASE_VERSIONS = (
"2.11.0202242",
"2.11.0202250",
"2.11.0202317",
"2.11.0202323",
"2.14.0202335",
"2.14.dev202335",
"2.15.0202341",
)


def version1_lt_version2(version1, version2):
Expand Down
3 changes: 2 additions & 1 deletion neural_compressor/model/tensorflow_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,7 +81,8 @@ def get_model_type(model):
return "graph"
elif isinstance(model, tf.compat.v1.GraphDef):
return "graph_def"
elif isinstance(model, tf.compat.v1.estimator.Estimator):
# tf.estimator is removed after tf2.14.0
elif version1_lt_version2(tf.__version__, "2.14.0") and isinstance(model, tf.compat.v1.estimator.Estimator):
return "estimator"
elif isinstance(model, str):
model = os.path.abspath(os.path.expanduser(model))
Expand Down
2 changes: 1 addition & 1 deletion test/adaptor/tensorflow_adaptor/test_tensorflow_graph_dequantize_cast_optimizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@
import yaml
from tensorflow.python.framework import dtypes

from neural_compressor.adaptor.tf_utils.graph_rewriter.generic.dequantize_cast_optimizer import DequantizeCastOptimizer
from neural_compressor.adaptor.tf_utils.graph_rewriter.bf16.dequantize_cast_optimizer import DequantizeCastOptimizer
from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper as Helper
from neural_compressor.adaptor.tf_utils.util import disable_random

Expand Down
2 changes: 1 addition & 1 deletion test/tfnewapi/test_tensorflow_graph_dequantize_cast_optimizer_newapi.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@
import yaml
from tensorflow.python.framework import dtypes

from neural_compressor.adaptor.tf_utils.graph_rewriter.generic.dequantize_cast_optimizer import DequantizeCastOptimizer
from neural_compressor.adaptor.tf_utils.graph_rewriter.bf16.dequantize_cast_optimizer import DequantizeCastOptimizer
from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper as Helper
from neural_compressor.adaptor.tf_utils.util import disable_random

Expand Down
Loading