Lunderberg commented on code in PR #15806: URL: https://github.com/apache/tvm/pull/15806#discussion_r1334768115
########## python/tvm/relax/transform/remove_redundant_reshape.py: ########## @@ -0,0 +1,73 @@ +# 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, missing-function-docstring, abstract-method + +from tvm.ir.module import IRModule +from tvm.ir.transform import PassContext +from tvm.relax import Expr, Function +from tvm.relax.dpl import is_op, rewrite_call, wildcard +from . import function_pass + + +@function_pass(opt_level=0) +class RemoveRedundantReshape: + """ + Transformation pass to remove redundant reshape operator + """ + + def __init__(self): + input = wildcard() + shape1 = wildcard() + pattern_redundant_reshape = is_op("relax.reshape")(input, shape1) + self.pattern = pattern_redundant_reshape + + def transform_function(self, func: Expr, mod: IRModule, ctx: PassContext) -> IRModule: + """ + Tarnsformation function to remove redundant reshape + where tensors before and after reshape are of same dimentions. + + Parameters + -------------- + func: Expr + The relax function to be optimized + + mod: IRModule + The IR module + + ctx: PassContext + Relax pass context + """ + + updated_func = func + for _, func in mod.functions.items(): + # Skip non-relax functions + if not isinstance(func, Function): + continue + # Skip primitive functions + if "Primitive" in func.attrs.keys() and func.attrs["Primitive"] != 0: + continue + + def rewriter(expr, matches): + args = matches[self.pattern] Review Comment: Can we first check that `args.args[0].struct_info.shape is not None`? Otherwise, this rewrite rule would throw an error when the argument is `R.TensorStructInfo()` (tensor of unknown dimension) or `R.TensorStructInfo(ndim=N)` (tensor whose dimensionality is known, but whose full shape is not known). ########## python/tvm/relax/transform/remove_redundant_reshape.py: ########## @@ -0,0 +1,73 @@ +# 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, missing-function-docstring, abstract-method + +from tvm.ir.module import IRModule +from tvm.ir.transform import PassContext +from tvm.relax import Expr, Function +from tvm.relax.dpl import is_op, rewrite_call, wildcard +from . import function_pass + + +@function_pass(opt_level=0) +class RemoveRedundantReshape: + """ + Transformation pass to remove redundant reshape operator + """ + + def __init__(self): + input = wildcard() + shape1 = wildcard() + pattern_redundant_reshape = is_op("relax.reshape")(input, shape1) + self.pattern = pattern_redundant_reshape Review Comment: This pattern only checks for reshapes whose argument already matches the new shape. Should we also check for reshapes whose argument is itself a reshape? That is, `reshape(reshape(arg, shape1), shape2)` can always be written as `reshape(arg, shape2)`. ########## python/tvm/relax/transform/remove_redundant_reshape.py: ########## @@ -0,0 +1,73 @@ +# 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, missing-function-docstring, abstract-method + +from tvm.ir.module import IRModule +from tvm.ir.transform import PassContext +from tvm.relax import Expr, Function +from tvm.relax.dpl import is_op, rewrite_call, wildcard +from . import function_pass + + +@function_pass(opt_level=0) +class RemoveRedundantReshape: + """ + Transformation pass to remove redundant reshape operator + """ + + def __init__(self): + input = wildcard() + shape1 = wildcard() + pattern_redundant_reshape = is_op("relax.reshape")(input, shape1) + self.pattern = pattern_redundant_reshape + + def transform_function(self, func: Expr, mod: IRModule, ctx: PassContext) -> IRModule: + """ + Tarnsformation function to remove redundant reshape + where tensors before and after reshape are of same dimentions. + + Parameters + -------------- + func: Expr + The relax function to be optimized + + mod: IRModule + The IR module + + ctx: PassContext + Relax pass context + """ + + updated_func = func + for _, func in mod.functions.items(): + # Skip non-relax functions + if not isinstance(func, Function): + continue + # Skip primitive functions + if "Primitive" in func.attrs.keys() and func.attrs["Primitive"] != 0: + continue + + def rewriter(expr, matches): + args = matches[self.pattern] + if list(args.args[0].struct_info.shape) == list(args.args[1]): Review Comment: Instead of using `list(a) == list(b)`, can we use `tvm.ir.structural_equal(a,b)`? This would ensure that symbolic shapes with `PrimExpr` expressions are correctly handled as well as static shapes. ########## tests/python/relax/test_remove_redundant_reshape.py: ########## @@ -0,0 +1,68 @@ +# 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. + +""" +Test relax transform - Eliminate redundant reshape operations +""" +import tvm.testing +from tvm import relax +from tvm.relax.transform import DeadCodeElimination +from tvm.relax.transform import RemoveRedundantReshape +from tvm.script import ir as I, relax as R + + +def _run_pass_compare_output(Before, Expected): + fused_mod = RemoveRedundantReshape()(Before) + if not relax.analysis.well_formed(fused_mod): + print("IRModule is not well-formed") + + fused_mod = DeadCodeElimination()(fused_mod) + if not relax.analysis.well_formed(fused_mod): + print("IRModule is not well-formed") Review Comment: Should these cases throw an exception instead of just printing to terminal? -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
