You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/08/09 21:40:24 UTC
[GitHub] [tvm] ashutosh-arm opened a new pull request, #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
ashutosh-arm opened a new pull request, #12353:
URL: https://github.com/apache/tvm/pull/12353
Pass that fuses nn.pad and qnn.conv2d for CMSIS-NN target.
cc @manupa-arm @Mousius @lhutton1 @grant-arm
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942615935
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
Review Comment:
ACK
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
+ dtype,
Review Comment:
ACK
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
+ dtype,
+ dtype,
Review Comment:
ACK
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] lhutton1 commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
lhutton1 commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r948992261
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,340 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+from .utils import CheckForPadsWithinCompositeFunc
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Negative tests for pads preceding Conv2D that cannot be fused."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
Review Comment:
Nit: possible to check the whole error message? i.e. with the pad values?
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -136,15 +145,72 @@ def check_qnn_conv2d(pattern):
):
is_depthwise = True
- return (
+ ret = (
+ conv2d.attrs.out_dtype == "int32"
+ and conv2d_input.checked_type.dtype == "int8"
+ and conv2d_weight.checked_type.dtype == "int8"
+ and pattern.checked_type.dtype == "int8"
+ and bias_dtype == "int32"
+ and all([zp == 0 for zp in kernel_zp])
+ and (not is_depthwise or bias_add is not None)
+ )
+ return ret
+
+ def check_qnn_conv2d_pad(pattern):
Review Comment:
Since some of the logic here is shared between `check_qnn_conv2d` and `check_qnn_conv2d_pad`, could we create a generic function that returns the appropriate check function? e.g. `create_check_qnn_conv2d(with_pad=False)` and then in the pattern table we can have something like:
```
("cmsis-nn.qnn_conv2d", qnn_conv2d_pattern(with_pad=True), create_check_conv2d(True)),
("cmsis-nn.qnn_conv2d", qnn_conv2d_pattern(with_pad=False), create_check_conv2d(False)),
```
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
+
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ int pad_diff_w = pad_right - pad_left;
+ int pad_diff_h = pad_bottom - pad_top;
Review Comment:
Nice :)
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,208 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file src/relay/backend/contrib/cmsisnn/fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+inline IntImm ToIntImm(int32_t value) { return IntImm(DataType::Int(32), value); }
+
+/*!
+ * \brief From padding attributes of nn.pad and qnn.conv2d, calculates effective padding along H
+ * and W dimensions.
+ */
+Array<IntImm> GetEffectiveConv2DPadding(Expr conv2d, Expr pad) {
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ auto* conv2d_call = conv2d.as<CallNode>();
+ auto* conv2d_attrs = conv2d_call->attrs.as<Conv2DAttrs>();
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ auto* pad_call = pad.as<CallNode>();
+ Array<Array<Integer>> pad_width = pad_call->attrs.as<PadAttrs>()->pad_width;
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ return {ToIntImm(pad_top), ToIntImm(pad_left), ToIntImm(pad_bottom), ToIntImm(pad_right)};
+}
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
Review Comment:
Nit: newline after "!"
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] lhutton1 commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
lhutton1 commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r951552476
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -136,15 +145,72 @@ def check_qnn_conv2d(pattern):
):
is_depthwise = True
- return (
+ ret = (
+ conv2d.attrs.out_dtype == "int32"
+ and conv2d_input.checked_type.dtype == "int8"
+ and conv2d_weight.checked_type.dtype == "int8"
+ and pattern.checked_type.dtype == "int8"
+ and bias_dtype == "int32"
+ and all([zp == 0 for zp in kernel_zp])
+ and (not is_depthwise or bias_add is not None)
+ )
+ return ret
+
+ def check_qnn_conv2d_pad(pattern):
Review Comment:
Sounds good!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944012860
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,279 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides method to test number of pads present inside the function being visited."""
+
+ def __init__(self):
+ super().__init__()
+ self.num_pads_ = 0
+
+ def visit_call(self, call):
+ super().visit_call(call)
+ if (
+ isinstance(call, tvm.relay.Call)
+ and isinstance(call.op, tvm.ir.op.Op)
+ and call.op.name == "nn.pad"
+ ):
+ self.num_pads_ += 1
+
+ def check_num_pads(self):
+ assert self.num_pads_ == 0, "CMSIS-NN composite function should not have pads"
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Negative tests for pads preceding Conv2D that cannot be fused."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
+
+ with pytest.raises(tvm.TVMError, match=error_regex):
+ mod = CMSISNNFusePads()(mod)
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 1), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 1), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 2), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_pad_conv2d_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pads and conv2d fusion."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
+
+
+def test_without_preceding_pad():
+ """Tests the pass FusePads when padding is not present before qnn.conv2d."""
+ dtype = "int8"
+ ifm_shape = (1, 56, 56, 64)
+ ofm_shape = (1, 56, 56, 64)
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(64, 3, 3, 64),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ local_input,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=64,
+ kernel_size=(3, 3),
+ padding=(1, 1, 1, 1),
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ relu = relay.nn.relu(requantize)
+ local_func = relay.Function(relay.analysis.free_vars(relu), relu)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function(relay.analysis.free_vars(call_extern_func), call_extern_func)
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
Review Comment:
Done
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944013801
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -729,7 +830,7 @@ def test_invalid_parameters(
in_dtype,
kernel_dtype,
in_dtype,
- False,
+ is_depthwise=False,
)
model, params = make_model(
shape=ifm_shape,
Review Comment:
Added two tests: test_invalid_pad_conv2d_fusion_int8 and test_pad_conv2d_fusion_int8. For the invalid padding cases, I had to introduce another pattern checker named check_qnn_conv2d_pad() in cmsisnn.py without which it defaults to including pads in the pattern.
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
Review Comment:
Ack
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r951497535
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,208 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file src/relay/backend/contrib/cmsisnn/fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+inline IntImm ToIntImm(int32_t value) { return IntImm(DataType::Int(32), value); }
+
+/*!
+ * \brief From padding attributes of nn.pad and qnn.conv2d, calculates effective padding along H
+ * and W dimensions.
+ */
+Array<IntImm> GetEffectiveConv2DPadding(Expr conv2d, Expr pad) {
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ auto* conv2d_call = conv2d.as<CallNode>();
+ auto* conv2d_attrs = conv2d_call->attrs.as<Conv2DAttrs>();
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ auto* pad_call = pad.as<CallNode>();
+ Array<Array<Integer>> pad_width = pad_call->attrs.as<PadAttrs>()->pad_width;
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ return {ToIntImm(pad_top), ToIntImm(pad_left), ToIntImm(pad_bottom), ToIntImm(pad_right)};
+}
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
Review Comment:
ACK
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] SebastianBoblest commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
SebastianBoblest commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942613452
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -122,6 +130,25 @@ def check_qnn_conv2d(pattern):
conv2d_input = conv2d.args[0]
conv2d_weight = conv2d.args[1]
+ # check if sum of paddings from pad() and conv2d() satisfies CMSIS-NN constraints
+ can_pad_be_fused = True
+ if isinstance(conv2d_input, tvm.relay.expr.Call) and str(conv2d_input.op.name) == "nn.pad":
+ data_layout = conv2d.attrs.data_layout
+ conv2d_padding = conv2d.attrs.padding # (top, left, bottom, right)
+ pad = conv2d_input
+ pad_width = pad.attrs.pad_width
+ pad_width_w = pad_width[data_layout.find("W")] # (left, right)
+ pad_width_h = pad_width[data_layout.find("H")] # (top, bottom)
+ # calculate effective padding post pad fusion
+ pad_top = pad_width_h[0] + conv2d_padding[0]
+ pad_bottom = pad_width_h[1] + conv2d_padding[2]
+ pad_left = pad_width_w[0] + conv2d_padding[1]
+ pad_right = pad_width_w[1] + conv2d_padding[3]
+ # check if difference in the side paddings is 1 along each dimension
+ pad_w_diff = int(pad_right - pad_left)
Review Comment:
> They are IntImm. When they are used in conditional checks, it returns IntImm in place of boolean which is what we'd expect normally. We have seen this in the past too:
>
> https://github.com/apache/tvm/blob/22ba659438a317ca59c8201430c662f86e2550fd/python/tvm/relay/op/contrib/cmsisnn.py#L134
Thanks for the explanation!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944012978
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,279 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides method to test number of pads present inside the function being visited."""
+
+ def __init__(self):
+ super().__init__()
+ self.num_pads_ = 0
+
+ def visit_call(self, call):
+ super().visit_call(call)
+ if (
+ isinstance(call, tvm.relay.Call)
+ and isinstance(call.op, tvm.ir.op.Op)
+ and call.op.name == "nn.pad"
+ ):
+ self.num_pads_ += 1
+
+ def check_num_pads(self):
+ assert self.num_pads_ == 0, "CMSIS-NN composite function should not have pads"
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Negative tests for pads preceding Conv2D that cannot be fused."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
+
+ with pytest.raises(tvm.TVMError, match=error_regex):
+ mod = CMSISNNFusePads()(mod)
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 1), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 1), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 2), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_pad_conv2d_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pads and conv2d fusion."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
+
+
+def test_without_preceding_pad():
+ """Tests the pass FusePads when padding is not present before qnn.conv2d."""
+ dtype = "int8"
+ ifm_shape = (1, 56, 56, 64)
+ ofm_shape = (1, 56, 56, 64)
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(64, 3, 3, 64),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ local_input,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=64,
+ kernel_size=(3, 3),
+ padding=(1, 1, 1, 1),
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ relu = relay.nn.relu(requantize)
+ local_func = relay.Function(relay.analysis.free_vars(relu), relu)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function(relay.analysis.free_vars(call_extern_func), call_extern_func)
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
Review Comment:
New test name: test_pad_conv2d_fusion_noncmsisnn_target
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r951486012
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -136,15 +145,72 @@ def check_qnn_conv2d(pattern):
):
is_depthwise = True
- return (
+ ret = (
+ conv2d.attrs.out_dtype == "int32"
+ and conv2d_input.checked_type.dtype == "int8"
+ and conv2d_weight.checked_type.dtype == "int8"
+ and pattern.checked_type.dtype == "int8"
+ and bias_dtype == "int32"
+ and all([zp == 0 for zp in kernel_zp])
+ and (not is_depthwise or bias_add is not None)
+ )
+ return ret
+
+ def check_qnn_conv2d_pad(pattern):
Review Comment:
It looks like pattern_table API does not support this. What I can do is this: Create a sub function that will get called from the other as both of them have many common checks.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] SebastianBoblest commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
SebastianBoblest commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942434869
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -122,6 +130,25 @@ def check_qnn_conv2d(pattern):
conv2d_input = conv2d.args[0]
conv2d_weight = conv2d.args[1]
+ # check if sum of paddings from pad() and conv2d() satisfies CMSIS-NN constraints
+ can_pad_be_fused = True
+ if isinstance(conv2d_input, tvm.relay.expr.Call) and str(conv2d_input.op.name) == "nn.pad":
+ data_layout = conv2d.attrs.data_layout
+ conv2d_padding = conv2d.attrs.padding # (top, left, bottom, right)
+ pad = conv2d_input
+ pad_width = pad.attrs.pad_width
+ pad_width_w = pad_width[data_layout.find("W")] # (left, right)
+ pad_width_h = pad_width[data_layout.find("H")] # (top, bottom)
+ # calculate effective padding post pad fusion
+ pad_top = pad_width_h[0] + conv2d_padding[0]
+ pad_bottom = pad_width_h[1] + conv2d_padding[2]
+ pad_left = pad_width_w[0] + conv2d_padding[1]
+ pad_right = pad_width_w[1] + conv2d_padding[3]
+ # check if difference in the side paddings is 1 along each dimension
+ pad_w_diff = int(pad_right - pad_left)
Review Comment:
Why is this cast to int necessary? What are the data types of pad_right and pad_left?
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
Review Comment:
I was not familiar with the signature of get_conv2d_qnn_params, so for me it would be helpful to have
```suggestion
input_dtype=dtype,
```
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
+ dtype,
Review Comment:
```suggestion
weights_dtype=dtype,
```
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,285 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides methods to test number of constants present in a function"""
+
+ def __init__(self):
+ super().__init__()
+ self.num_pads_ = 0
+
+ def visit_call(self, call):
+ super().visit_call(call)
+ if (
+ isinstance(call, tvm.relay.Call)
+ and isinstance(call.op, tvm.ir.op.Op)
+ and call.op.name == "nn.pad"
+ ):
+ self.num_pads_ += 1
+
+ def check_num_pads(self):
+ assert self.num_pads_ == 0, "CMSIS-NN composite function should not have pads"
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pass ExternConstants when a composite function
+ is present within global function
+ """
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
+
+ with pytest.raises(tvm.TVMError, match=error_regex):
+ mod = CMSISNNFusePads()(mod)
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 1), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 1), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 2), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_pad_conv2d_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pass ExternConstants when a composite function
+ is present within global function
+ """
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
+
+
+def test_without_preceding_pad():
+ """Tests the pass ExternConstants when a composite function
Review Comment:
This test and the two preceding ones have identical doc strings. Could you maybe add one sencentence to make the doc strings more informative?
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
+ dtype,
+ dtype,
+ )
+
+ invar = relay.var("input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ invar,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=input_zero_point,
+ pad_mode="constant",
+ )
+
+ model, params = make_model(
+ ifm_shape,
+ kernel_shape,
+ input_zero_point,
+ input_scale,
+ kernel_zero_point,
+ kernel_scale,
+ output_zero_point,
+ output_scale,
+ padding,
+ strides,
+ dilation,
+ groups,
+ dtype,
Review Comment:
Here, keyword argument syntax could make this a bit more readable
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -403,6 +401,109 @@ def test_conv2d_asymmetric_padding_int8(
)
+@tvm.testing.requires_cmsisnn
+@pytest.mark.parametrize("ifm_shape", [(1, 25, 25, 12), (1, 64, 100, 4)])
+@pytest.mark.parametrize(
+ "pad_width",
+ [
+ ((0, 0), (0, 1), (1, 2), (0, 0)),
+ ((0, 0), (1, 1), (1, 1), (0, 0)),
+ ((0, 0), (2, 2), (3, 4), (0, 0)),
+ ],
+)
+def test_conv2d_separate_padding_int8(
+ ifm_shape,
+ pad_width,
+):
+ """Tests QNN Conv2D where the padding is asymmetric on different sides of input"""
+ interface_api = "c"
+ use_unpacked_api = True
+ test_runner = AOT_USMP_CORSTONE300_RUNNER
+
+ ifm_shape = (1, 25, 25, 12)
+ kernel_size = (5, 5)
+ strides = (2, 2)
+ dilation = (1, 1)
+ padding = "SAME"
+ dtype = "int8"
+ enable_bias = True
+ relu_type = "NONE"
+ input_zero_point = 10
+ input_scale = 0.0128
+ kernel_scale = [0.11, 0.22]
+ out_channels = 2
+ groups = 1
+ weight_format = "HWIO"
+ kernel_h = kernel_size[0]
+ kernel_w = kernel_size[1]
+ kernel_shape = (kernel_h, kernel_w, ifm_shape[3] // groups, out_channels)
+ kernel_zero_point = 0
+ in_min, in_max = get_range_for_dtype_str(dtype)
+
+ output_scale, output_zero_point = get_conv2d_qnn_params(
+ kernel_shape,
+ input_scale,
+ input_zero_point,
+ kernel_scale,
+ kernel_zero_point,
+ dtype,
+ dtype,
+ dtype,
Review Comment:
```suggestion
output_dtype=dtype,
```
##########
tests/python/contrib/test_cmsisnn/utils.py:
##########
@@ -50,8 +50,19 @@ def visit_call(self, call):
return counter.count
-def assert_partitioned_function(orig_mod, cmsisnn_mod):
- """If kCompiler attribute is missing, this function raises assertion"""
+def assert_partitioned_function(orig_mod, cmsisnn_mod, is_num_calls_same=True):
Review Comment:
```suggestion
def assert_partitioned_function(orig_mod, cmsisnn_mod, num_call_nodes_expected_unchanged=True):
```
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,285 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides methods to test number of constants present in a function"""
Review Comment:
```suggestion
"""Provides methods to test how many constants are present in a function"""
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944013097
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
Review Comment:
I am not sure how to do that. Is there any example you can point to?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942617075
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,285 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides methods to test number of constants present in a function"""
Review Comment:
Done
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] lhutton1 commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
lhutton1 commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r943274640
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
Review Comment:
Nit: src/relay/backend/contrib/cmsisnn/fuse_pads.cc
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,279 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides method to test number of pads present inside the function being visited."""
+
+ def __init__(self):
+ super().__init__()
+ self.num_pads_ = 0
+
+ def visit_call(self, call):
+ super().visit_call(call)
+ if (
+ isinstance(call, tvm.relay.Call)
+ and isinstance(call.op, tvm.ir.op.Op)
+ and call.op.name == "nn.pad"
+ ):
+ self.num_pads_ += 1
+
+ def check_num_pads(self):
+ assert self.num_pads_ == 0, "CMSIS-NN composite function should not have pads"
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Negative tests for pads preceding Conv2D that cannot be fused."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
+
+ with pytest.raises(tvm.TVMError, match=error_regex):
+ mod = CMSISNNFusePads()(mod)
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 1), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 1), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 2), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_pad_conv2d_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pads and conv2d fusion."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
+
+
+def test_without_preceding_pad():
+ """Tests the pass FusePads when padding is not present before qnn.conv2d."""
+ dtype = "int8"
+ ifm_shape = (1, 56, 56, 64)
+ ofm_shape = (1, 56, 56, 64)
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(64, 3, 3, 64),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ local_input,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=64,
+ kernel_size=(3, 3),
+ padding=(1, 1, 1, 1),
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ relu = relay.nn.relu(requantize)
+ local_func = relay.Function(relay.analysis.free_vars(relu), relu)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function(relay.analysis.free_vars(call_extern_func), call_extern_func)
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
Review Comment:
Could we add a test to check that padding for non-cmsisnn functions is not altered?
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
Review Comment:
Is it possible to copy conv2d_attrs instead?
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
+
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ int pad_diff_w = pad_right - pad_left;
+ int pad_diff_h = pad_bottom - pad_top;
+ bool can_pad_be_fused =
+ ((pad_diff_w == 0 || pad_diff_w == 1) && (pad_diff_h == 0 || pad_diff_h == 1));
+ std::string error = "Difference on each side of a dimension should be either 0 or 1. ";
+ error += "Effective padding in this case: (pad_top, pad_left, pad_bottom, pad_right)=(";
+ error += std::to_string(pad_top);
+ error += ", ";
+ error += std::to_string(pad_left);
+ error += ", ";
+ error += std::to_string(pad_bottom);
+ error += ", ";
+ error += std::to_string(pad_right);
+ error += ")";
+ ICHECK(can_pad_be_fused) << error;
+
+ attrs->padding = {pad_top, pad_left, pad_bottom, pad_right};
+ *new_attrs = tvm::Attrs{attrs};
+ }
+
+ /*!
+ * \brief Identifies the sequence for qnn.conv2D and fuses the preceding nn.pad present within the
+ * CMSIS-NN partitioned function. */
+ Expr FusePadConv2d(const Expr& expr) {
+ const CallNode* clip_call = nullptr;
+ const CallNode* requantize_call = nullptr;
+ const CallNode* bias_add_call = nullptr;
Review Comment:
Curious, why do we need to maintain references to these ops and reconstruct them at the end of this function? Will a call to `VisitExpr` after the new conv2d has been created do this for us instead?
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -729,7 +830,7 @@ def test_invalid_parameters(
in_dtype,
kernel_dtype,
in_dtype,
- False,
+ is_depthwise=False,
)
model, params = make_model(
shape=ifm_shape,
Review Comment:
Would be good to check a couple of other things here:
- The pass runs as part of the pipeline as these tests could pass even when the pass is not run
- Attempting to fuse a pad operation that is unsupported falls back to offloading just conv2d (think we might have missed this case in microNPU)
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
Review Comment:
Nit: additional line here?
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
+
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ int pad_diff_w = pad_right - pad_left;
+ int pad_diff_h = pad_bottom - pad_top;
Review Comment:
Is it possible to separate this logic and register it globally, so we can use it in python in the check function as well to avoid duplication?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944012748
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
Review Comment:
Done
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
+
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ int pad_diff_w = pad_right - pad_left;
+ int pad_diff_h = pad_bottom - pad_top;
Review Comment:
Done
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r951486012
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -136,15 +145,72 @@ def check_qnn_conv2d(pattern):
):
is_depthwise = True
- return (
+ ret = (
+ conv2d.attrs.out_dtype == "int32"
+ and conv2d_input.checked_type.dtype == "int8"
+ and conv2d_weight.checked_type.dtype == "int8"
+ and pattern.checked_type.dtype == "int8"
+ and bias_dtype == "int32"
+ and all([zp == 0 for zp in kernel_zp])
+ and (not is_depthwise or bias_add is not None)
+ )
+ return ret
+
+ def check_qnn_conv2d_pad(pattern):
Review Comment:
It looks like pattern_table API does not support this. What I can do is to create a sub function that will get called from the other as most of them have common checks.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] lhutton1 commented on pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
lhutton1 commented on PR #12353:
URL: https://github.com/apache/tvm/pull/12353#issuecomment-1223772593
Thanks @ashutosh-arm, @SebastianBoblest!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942622944
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -122,6 +130,25 @@ def check_qnn_conv2d(pattern):
conv2d_input = conv2d.args[0]
conv2d_weight = conv2d.args[1]
+ # check if sum of paddings from pad() and conv2d() satisfies CMSIS-NN constraints
+ can_pad_be_fused = True
+ if isinstance(conv2d_input, tvm.relay.expr.Call) and str(conv2d_input.op.name) == "nn.pad":
+ data_layout = conv2d.attrs.data_layout
+ conv2d_padding = conv2d.attrs.padding # (top, left, bottom, right)
+ pad = conv2d_input
+ pad_width = pad.attrs.pad_width
+ pad_width_w = pad_width[data_layout.find("W")] # (left, right)
+ pad_width_h = pad_width[data_layout.find("H")] # (top, bottom)
+ # calculate effective padding post pad fusion
+ pad_top = pad_width_h[0] + conv2d_padding[0]
+ pad_bottom = pad_width_h[1] + conv2d_padding[2]
+ pad_left = pad_width_w[0] + conv2d_padding[1]
+ pad_right = pad_width_w[1] + conv2d_padding[3]
+ # check if difference in the side paddings is 1 along each dimension
+ pad_w_diff = int(pad_right - pad_left)
Review Comment:
No worries :+1:
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944014209
##########
tests/python/contrib/test_cmsisnn/test_conv2d.py:
##########
@@ -729,7 +830,7 @@ def test_invalid_parameters(
in_dtype,
kernel_dtype,
in_dtype,
- False,
+ is_depthwise=False,
)
model, params = make_model(
shape=ifm_shape,
Review Comment:
I have also added extra checks to test presence of pads in the invalid cases post partition'g.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r944485521
##########
src/relay/backend/contrib/cmsisnn/fuse_pads.cc:
##########
@@ -0,0 +1,219 @@
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+/*!
+ * \file fuse_pads.cc
+ * \brief Fuses pads that precede qnn.conv2d ops inside CMSIS-NN composite functions.
+ */
+
+#include <tvm/relay/attrs/nn.h>
+#include <tvm/relay/attrs/transform.h>
+#include <tvm/relay/expr_functor.h>
+#include <tvm/relay/transform.h>
+#include <tvm/runtime/ndarray.h>
+
+#include "../../../op/make_op.h"
+#include "../../../qnn/utils.h"
+#include "../../../transforms/pattern_utils.h"
+#include "convolutions.h"
+
+namespace tvm {
+namespace relay {
+namespace contrib {
+namespace cmsisnn {
+
+/*!
+ * \brief This Mutator will find all partitioned functions meant for CMSIS-NN Conv2D.
+ * Then, it will fuse preceding pads with qnn.conv2d.
+ */
+class FusePadsMutator : public MixedModeMutator {
+ public:
+ explicit FusePadsMutator(const IRModule& mod) : mod_(mod) {}
+
+ private:
+ /*! * \brief In order to eliminate preceding nn.pad op, pad_width of nn.pad is passed onto
+ * convolution layer to update Conv2DAttrs's padding attribute. */
+ void UpdateConv2DPadding(const CallNode* conv2d_call, const Array<Array<Integer>>& pad_width,
+ const Conv2DAttrs* conv2d_attrs, Attrs* new_attrs) {
+ auto attrs = make_object<Conv2DAttrs>();
+ attrs->strides = std::move(conv2d_attrs->strides);
+ attrs->dilation = std::move(conv2d_attrs->dilation);
+ attrs->groups = conv2d_attrs->groups;
+ attrs->channels = std::move(conv2d_attrs->channels);
+ attrs->kernel_size = std::move(conv2d_attrs->kernel_size);
+ attrs->data_layout = std::move(conv2d_attrs->data_layout);
+ attrs->kernel_layout = std::move(conv2d_attrs->kernel_layout);
+ attrs->out_layout = std::move(conv2d_attrs->out_layout);
+ attrs->out_dtype = std::move(conv2d_attrs->out_dtype);
+
+ // pad_width: ((), (top, bottom), (left, right), ()) for NHWC layout
+ // conv2d_attrs->padding: (top, left, bottom, right)
+ std::string data_layout = conv2d_attrs->data_layout.c_str();
+ int pos_h = data_layout.find("H");
+ int pos_w = data_layout.find("W");
+
+ int pad_top =
+ qnn::get_const_int(conv2d_attrs->padding[0]) + qnn::get_const_int(pad_width[pos_h][0]);
+ int pad_left =
+ qnn::get_const_int(conv2d_attrs->padding[1]) + qnn::get_const_int(pad_width[pos_w][0]);
+ int pad_bottom =
+ qnn::get_const_int(conv2d_attrs->padding[2]) + qnn::get_const_int(pad_width[pos_h][1]);
+ int pad_right =
+ qnn::get_const_int(conv2d_attrs->padding[3]) + qnn::get_const_int(pad_width[pos_w][1]);
+
+ int pad_diff_w = pad_right - pad_left;
+ int pad_diff_h = pad_bottom - pad_top;
+ bool can_pad_be_fused =
+ ((pad_diff_w == 0 || pad_diff_w == 1) && (pad_diff_h == 0 || pad_diff_h == 1));
+ std::string error = "Difference on each side of a dimension should be either 0 or 1. ";
+ error += "Effective padding in this case: (pad_top, pad_left, pad_bottom, pad_right)=(";
+ error += std::to_string(pad_top);
+ error += ", ";
+ error += std::to_string(pad_left);
+ error += ", ";
+ error += std::to_string(pad_bottom);
+ error += ", ";
+ error += std::to_string(pad_right);
+ error += ")";
+ ICHECK(can_pad_be_fused) << error;
+
+ attrs->padding = {pad_top, pad_left, pad_bottom, pad_right};
+ *new_attrs = tvm::Attrs{attrs};
+ }
+
+ /*!
+ * \brief Identifies the sequence for qnn.conv2D and fuses the preceding nn.pad present within the
+ * CMSIS-NN partitioned function. */
+ Expr FusePadConv2d(const Expr& expr) {
+ const CallNode* clip_call = nullptr;
+ const CallNode* requantize_call = nullptr;
+ const CallNode* bias_add_call = nullptr;
Review Comment:
@lhutton1 Could you PTAL? I've tried compressing the recreation of the new Conv2D now.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] lhutton1 merged pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
lhutton1 merged PR #12353:
URL: https://github.com/apache/tvm/pull/12353
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942621766
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,285 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+class CheckForPadsWithinCompositeFunc(tvm.relay.ExprVisitor):
+ """Provides methods to test number of constants present in a function"""
+
+ def __init__(self):
+ super().__init__()
+ self.num_pads_ = 0
+
+ def visit_call(self, call):
+ super().visit_call(call)
+ if (
+ isinstance(call, tvm.relay.Call)
+ and isinstance(call.op, tvm.ir.op.Op)
+ and call.op.name == "nn.pad"
+ ):
+ self.num_pads_ += 1
+
+ def check_num_pads(self):
+ assert self.num_pads_ == 0, "CMSIS-NN composite function should not have pads"
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pass ExternConstants when a composite function
+ is present within global function
+ """
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
+
+ with pytest.raises(tvm.TVMError, match=error_regex):
+ mod = CMSISNNFusePads()(mod)
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 1), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 1), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 2), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_pad_conv2d_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Tests the pass ExternConstants when a composite function
+ is present within global function
+ """
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ mod = CMSISNNFusePads()(mod)
+ pad_verifier = CheckForPadsWithinCompositeFunc()
+ pad_verifier.visit_function(mod[extern_var])
+ pad_verifier.check_num_pads()
+
+
+def test_without_preceding_pad():
+ """Tests the pass ExternConstants when a composite function
Review Comment:
Added them. Thanks for picking up on that.
##########
tests/python/contrib/test_cmsisnn/utils.py:
##########
@@ -50,8 +50,19 @@ def visit_call(self, call):
return counter.count
-def assert_partitioned_function(orig_mod, cmsisnn_mod):
- """If kCompiler attribute is missing, this function raises assertion"""
+def assert_partitioned_function(orig_mod, cmsisnn_mod, is_num_calls_same=True):
Review Comment:
Makes sense.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r942611235
##########
python/tvm/relay/op/contrib/cmsisnn.py:
##########
@@ -122,6 +130,25 @@ def check_qnn_conv2d(pattern):
conv2d_input = conv2d.args[0]
conv2d_weight = conv2d.args[1]
+ # check if sum of paddings from pad() and conv2d() satisfies CMSIS-NN constraints
+ can_pad_be_fused = True
+ if isinstance(conv2d_input, tvm.relay.expr.Call) and str(conv2d_input.op.name) == "nn.pad":
+ data_layout = conv2d.attrs.data_layout
+ conv2d_padding = conv2d.attrs.padding # (top, left, bottom, right)
+ pad = conv2d_input
+ pad_width = pad.attrs.pad_width
+ pad_width_w = pad_width[data_layout.find("W")] # (left, right)
+ pad_width_h = pad_width[data_layout.find("H")] # (top, bottom)
+ # calculate effective padding post pad fusion
+ pad_top = pad_width_h[0] + conv2d_padding[0]
+ pad_bottom = pad_width_h[1] + conv2d_padding[2]
+ pad_left = pad_width_w[0] + conv2d_padding[1]
+ pad_right = pad_width_w[1] + conv2d_padding[3]
+ # check if difference in the side paddings is 1 along each dimension
+ pad_w_diff = int(pad_right - pad_left)
Review Comment:
They are IntImm. When they are used in conditional checks, it returns IntImm in place of boolean which is what we'd expect normally. We have seen this in the past too: https://github.com/apache/tvm/blob/22ba659438a317ca59c8201430c662f86e2550fd/python/tvm/relay/op/contrib/cmsisnn.py#L134
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r951508497
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,340 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+from .utils import CheckForPadsWithinCompositeFunc
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Negative tests for pads preceding Conv2D that cannot be fused."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
Review Comment:
I intentionally ignored that bit as it requires re-derivation of combined padding in the test.
In case we make the combined padding golden (make the absolute values as test params), we could extend the error message. A part of it is being done through `InferType()` just before it as `ofm_shape` was added as a golden value to the test.
Another reason for not extending it was that the message specifically originates from the `CMSISNNFusePads()`. Plus this is the only pass that is being invoked.
wdyt?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] lhutton1 commented on a diff in pull request #12353: [CMSIS-NN] Pad fusion with QNN Conv2D
Posted by GitBox <gi...@apache.org>.
lhutton1 commented on code in PR #12353:
URL: https://github.com/apache/tvm/pull/12353#discussion_r951551646
##########
tests/python/contrib/test_cmsisnn/test_fuse_pads.py:
##########
@@ -0,0 +1,340 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+"""CMSIS-NN integration tests: fuse_pads pass"""
+import numpy as np
+import pytest
+import tvm
+import tvm.testing
+from tvm import relay
+from .utils import CheckForPadsWithinCompositeFunc
+
+tvm._ffi._init_api("relay.ext.cmsisnn.transform", __name__)
+
+
+def set_external_func_attr(func, compiler, ext_symbol):
+ func = func.with_attr("Primitive", tvm.tir.IntImm("int32", 1))
+ func = func.with_attr("Compiler", compiler)
+ func = func.with_attr("global_symbol", ext_symbol)
+ return func
+
+
+def set_composite_func_attr(func, name):
+ func = func.with_attr("Composite", name)
+ return func
+
+
+@pytest.mark.parametrize(
+ "ifm_shape, pad_width, conv2d_padding, ofm_shape",
+ [
+ [(1, 25, 25, 12), ((0, 0), (0, 2), (1, 2), (0, 0)), (1, 1, 1, 1), (1, 26, 28, 2)],
+ [(1, 64, 100, 4), ((0, 0), (1, 3), (1, 1), (0, 0)), (0, 0, 0, 0), (1, 64, 100, 2)],
+ [(1, 55, 55, 3), ((0, 0), (2, 1), (3, 5), (0, 0)), (0, 0, 1, 1), (1, 57, 59, 2)],
+ ],
+)
+def test_invalid_padding_for_fusion(ifm_shape, pad_width, conv2d_padding, ofm_shape):
+ """Negative tests for pads preceding Conv2D that cannot be fused."""
+ dtype = "int8"
+ kernel_size = (3, 3)
+ ofm_channels = 2
+ local_input = relay.var("local_input", shape=ifm_shape, dtype=dtype)
+ pad = relay.nn.pad(
+ local_input,
+ pad_width=pad_width, # ((), (top, bottom), (left, right), ())
+ pad_value=10,
+ pad_mode="constant",
+ )
+ rng = np.random.default_rng(12321)
+ local_weight = tvm.nd.array(
+ rng.integers(
+ np.iinfo(dtype).min,
+ high=np.iinfo(dtype).max,
+ size=(ofm_channels, kernel_size[0], kernel_size[1], ifm_shape[3]),
+ dtype=dtype,
+ )
+ )
+ local_weight = relay.const(local_weight, dtype)
+ conv2d = relay.qnn.op.conv2d(
+ pad,
+ local_weight,
+ relay.const(1, "int32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "float32"),
+ data_layout="NHWC",
+ kernel_layout="OHWI",
+ channels=ofm_channels,
+ kernel_size=(3, 3),
+ padding=conv2d_padding,
+ out_dtype="int32",
+ )
+ requantize = relay.qnn.op.requantize(
+ conv2d,
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ relay.const(1, "float32"),
+ relay.const(1, "int32"),
+ axis=0,
+ out_dtype=dtype,
+ )
+ local_func = relay.Function(relay.analysis.free_vars(requantize), requantize)
+ local_func = set_composite_func_attr(local_func, "cmsis-nn.qnn_conv2d")
+
+ mod = tvm.IRModule()
+ ext_input = relay.var("ext_input", shape=ifm_shape, dtype=dtype)
+ call_local_func = relay.Call(local_func, [ext_input])
+ extern_func = relay.Function(relay.analysis.free_vars(call_local_func), call_local_func)
+ extern_var = relay.GlobalVar("external_function")
+ extern_func = set_external_func_attr(extern_func, "cmsis-nn", extern_var.name_hint)
+ mod[extern_var] = extern_func
+
+ main_input = relay.var("main_input", shape=ifm_shape, dtype=dtype)
+ call_extern_func = relay.Call(extern_var, [main_input])
+ main_func = relay.Function([main_input], call_extern_func, relay.TensorType(ofm_shape, dtype))
+ main_var = relay.GlobalVar("main")
+ mod[main_var] = main_func
+
+ mod = relay.transform.InferType()(mod)
+
+ error_regex = r"Difference on each side of a dimension should be either 0 or 1"
Review Comment:
No worries, lets leave as is
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org