[GitHub] [tvm] apeskov commented on a change in pull request #7299: Introduce Apple BNNS backend

[GitBox <gi...@apache.org>]

apeskov commented on a change in pull request #7299:
URL: https://github.com/apache/tvm/pull/7299#discussion_r559692993



##########
File path: python/tvm/relay/op/contrib/bnns.py
##########
@@ -0,0 +1,247 @@
+# 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.
+# pylint: disable=invalid-name, unused-argument
+"""BNNS library supported operators.
+Is a part of Accelerate framework on macOS/iOS platforms. Apple provide several APIs
+to handle tensor processing. Particularly:
+ * BNNS (basic neural )
+ * vDSP (1D and 2D tensor processing)
+ * BLAS (gemm provide)
+
+# There are two ways to registering a function for an op to indicate if it is
+# supported by DNNL.
+
+# - The first and simplest way is to use the helper so that
+# users only need to provide the operator name and a boolean value to indicate if
+# it is supported. For example:
+#
+#     .. code-block:: python
+#
+#       add = _register_external_op_helper("add")
+#       add = _register_external_op_helper("add", True)
+#       add = _register_external_op_helper("add", False)
+#
+# - The other way is to implement the function by themselves to
+# check the attributes of the op and decide if it should be offloaded to DNNL.
+"""
+import math
+import tvm.ir
+from ...dataflow_pattern import wildcard, is_op, is_expr, is_constant
+from .register import register_pattern_table, get_pattern_table
+
+from tvm.relay import transform
+from tvm.relay.expr import const
+from tvm.relay.build_module import bind_params_by_name
+
+def partition_for_bnns(mod, params=None):
+    """Partition the graph greedily offloading supported
+    operators to BNNS.
+
+    Parameters
+    ----------
+    mod : Module
+        The module to run passes on.
+    params : Optional[Dict[str, NDArray]]
+        Constant input parameters.
+
+    Returns
+    -------
+    ret : annotated and partitioned module.
+    """
+    if params:
+        mod["main"] = bind_params_by_name(mod["main"], params)
+
+    seq = tvm.transform.Sequential(
+        [
+            transform.InferType(),
+            transform.FoldConstant(),
+            transform.FoldScaleAxis(),
+            transform.DynamicToStatic(),
+            transform.AlterOpLayout(),
+            # TODO(apeskov): WA. AlterOpLayout call lead to constants shape transformation
+            #   Some expand_dims op may appears after constants. It breaks BNNS fusing.
+            #   So we have to call FoldConstant right before bnns composite passes.
+            transform.FoldConstant(),
+            transform.MergeComposite(get_pattern_table("bnns")),
+            transform.AnnotateTarget("bnns"),
+            #   If you no need in per layer performance statistic you can
+            #   uncomment next line
+            # transform.MergeCompilerRegions(),
+            transform.PartitionGraph(),
+        ]
+    )
+
+    return seq(mod)
+
+
+def _register_external_op_helper(op_name, supported=True):
+    """The helper function to indicate that a given operator can be supported
+    by BNNS.
+
+    Paramters
+    ---------
+    op_name : Str
+        The name of operator that will be registered.
+
+    Returns
+    -------
+    f : callable
+        A function that returns if the operator is supported by BNNS.
+    """
+
+    @tvm.ir.register_op_attr(op_name, "target.bnns")
+    def _func_wrapper(expr):
+        return supported
+
+    return _func_wrapper
+
+_register_external_op_helper("nn.batch_matmul")
+
+
+# TODO [apeskov]:
+#   1. enlarge list of supported types on
+#   2. clarify meaning of "" value
+def dtype_is_supported(dtype):
+    return dtype == "float32" or dtype == ""
+
+
+@tvm.ir.register_op_attr("nn.conv2d", "target.bnns")
+def conv2d_check(expr):
+    """Check if the conv2d can be executed in BNNS"""
+    attrs, args = expr.attrs, expr.args
+    data_typ = args[0].checked_type
+    if len(data_typ.shape) != 4 or data_typ.dtype != "float32":
+        return False
+    if not isinstance(args[1], tvm.relay.expr.Constant):
+        return False
+    kernel_typ = args[1].checked_type
+    if len(kernel_typ.shape) != 4 or kernel_typ.dtype != "float32":
+        return False
+    if attrs.data_layout != "NCHW":
+        return False
+    if not dtype_is_supported(attrs.out_dtype):
+        return False
+    return True
+
+
+def bias_check(expr):
+    """Check is bias added through the correct dimension"""
+    attrs, args = expr.attrs, expr.args
+    if not isinstance(args[1], tvm.relay.expr.Constant):
+        return False
+    if expr.op.name == "nn.bias_add":
+        return attrs.axis == 1
+    elif expr.op.name == "add":
+        b_shape = args[1].checked_type.shape
+        if len(b_shape) == 4:
+            return bool(b_shape[0] == 1 and b_shape[2] == 1 and b_shape[3] == 1)
+        elif len(b_shape) == 3:
+            return bool(b_shape[1] == 1 and b_shape[2] == 1)

Review comment:
       `b_shape[0] == 1`  is an instance of `tvm.tir.expr.IntImm`. External consumer of this function expect bool value and further "bool like" processing lead to exception. 
   
   This explicit cast was a shorter way to avoid this.




----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org