anwang2009 commented on a change in pull request #9357: URL: https://github.com/apache/tvm/pull/9357#discussion_r736991002
########## File path: src/relay/transforms/fold_type_transformation.cc ########## @@ -0,0 +1,163 @@ +/* + * 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/transforms/fold_type_transformation.cc + * \brief A pass for transforming relay graph function + * signatures such that when a function-level inputs is + * transformed by a subsequent cast or quantize operation, + * that operation is folded into the signature itself. + */ + +#include <tvm/relay/expr.h> +#include <tvm/relay/expr_functor.h> +#include <tvm/relay/qnn/attrs.h> +#include <tvm/relay/transform.h> + +namespace tvm { +namespace relay { + +/*! \brief This class transforms a relay module's function signature + * such that when a function-level input is transformed by a subsequent + * "cast" or "qnn.quantize" operation, that operation is folded into + * the signature itself. For example, + * + * def @main(%data: Tensor[(1, 3, 224, 224), float32]) { + * %0 = qnn.quantize(%data, 2f, 0, out_dtype="uint8"); + * add(%0, %0) + * } + * + * would be transformed to + * + * def @main(%data: Tensor[(1, 3, 224, 224), uint8]) { + * add(%0, %0) + * } + * + * Note that now it is the user's responsibility to modify their + * input pre-processing pipeline to satisfy the new signature's + * constraints. + * + * For this pass to fold a type transformation, the following conditions + * must be met: + * - The relay module must contain only a single function. Review comment: Multiple function is weird because you have stuff like ``` fn %f1(%in: Tensor(dtype=float32) { cast(%in, dtype="float16") ... } fn %f2(%in: Tensor(dtype=float32) { %0 = add(%in, %in) f1(%0) } ``` and then we need to translate the result to ``` fn %f1(%in: Tensor(dtype=float16) { ... } fn %f2(%in: Tensor(dtype=float32) { %0 = add(%in, %in) %1 = cast(%0) f1(%1) } ``` which is just a lot more steps. Most quantized or to-mixed-precision passed models are single function so unless there is a greater need for multi function transformations I'm hoping to leave this as a simple transformation in the first iteration. ########## File path: src/relay/transforms/fold_type_transformation.cc ########## @@ -0,0 +1,163 @@ +/* + * 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/transforms/fold_type_transformation.cc + * \brief A pass for transforming relay graph function + * signatures such that when a function-level inputs is + * transformed by a subsequent cast or quantize operation, + * that operation is folded into the signature itself. + */ + +#include <tvm/relay/expr.h> +#include <tvm/relay/expr_functor.h> +#include <tvm/relay/qnn/attrs.h> +#include <tvm/relay/transform.h> + +namespace tvm { +namespace relay { + +/*! \brief This class transforms a relay module's function signature + * such that when a function-level input is transformed by a subsequent + * "cast" or "qnn.quantize" operation, that operation is folded into + * the signature itself. For example, + * + * def @main(%data: Tensor[(1, 3, 224, 224), float32]) { + * %0 = qnn.quantize(%data, 2f, 0, out_dtype="uint8"); + * add(%0, %0) + * } + * + * would be transformed to + * + * def @main(%data: Tensor[(1, 3, 224, 224), uint8]) { + * add(%0, %0) + * } + * + * Note that now it is the user's responsibility to modify their + * input pre-processing pipeline to satisfy the new signature's + * constraints. + * + * For this pass to fold a type transformation, the following conditions + * must be met: + * - The relay module must contain only a single function. + * - The type of each function-level input is transformed only once + * per program. + * - The type transformation operation must be either a "cast" + * or "qnn.quantize". + */ +class FoldTypeTransformationRewriter : public MixedModeMutator { + protected: + Expr Rewrite_(const CallNode* pre_call_node, const Expr& post) final { + const CallNode* post_call_node = post.as<CallNode>(); + CHECK(post_call_node) << "Expected a CallNode, but got " << post; + + Expr cur_op = pre_call_node->op; + for (auto arg : pre_call_node->args) { + auto maybe_var_node = arg.as<VarNode>(); + if (maybe_var_node) { + auto var = Downcast<Var>(arg); + auto it = input_transform_map_.find(var); + if (it != input_transform_map_.end()) { + // Checks that the function-level input var hasn't been an arg + // to a CallNode yet. + CHECK(!it->second) << "Function input with name '" << var->name_hint() + << "' is fed into more than one call; " + << "aborting transformation"; + Review comment: Ah, the message isn't about the function being called twice, it's about the param being directly fed into different operations, like ``` in / \ cast add ``` This is a case I didn't want to handle in the initial iteration because it's gnarly and uncommon. I just clarified the pass-level comment to this effect. ########## File path: src/relay/transforms/fold_type_transformation.cc ########## @@ -0,0 +1,163 @@ +/* + * 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/transforms/fold_type_transformation.cc + * \brief A pass for transforming relay graph function + * signatures such that when a function-level inputs is + * transformed by a subsequent cast or quantize operation, + * that operation is folded into the signature itself. + */ + +#include <tvm/relay/expr.h> +#include <tvm/relay/expr_functor.h> +#include <tvm/relay/qnn/attrs.h> +#include <tvm/relay/transform.h> + +namespace tvm { +namespace relay { + +/*! \brief This class transforms a relay module's function signature + * such that when a function-level input is transformed by a subsequent + * "cast" or "qnn.quantize" operation, that operation is folded into + * the signature itself. For example, + * + * def @main(%data: Tensor[(1, 3, 224, 224), float32]) { + * %0 = qnn.quantize(%data, 2f, 0, out_dtype="uint8"); + * add(%0, %0) + * } + * + * would be transformed to + * + * def @main(%data: Tensor[(1, 3, 224, 224), uint8]) { + * add(%0, %0) + * } + * + * Note that now it is the user's responsibility to modify their + * input pre-processing pipeline to satisfy the new signature's + * constraints. + * + * For this pass to fold a type transformation, the following conditions + * must be met: + * - The relay module must contain only a single function. + * - The type of each function-level input is transformed only once + * per program. + * - The type transformation operation must be either a "cast" + * or "qnn.quantize". + */ +class FoldTypeTransformationRewriter : public MixedModeMutator { + protected: + Expr Rewrite_(const CallNode* pre_call_node, const Expr& post) final { + const CallNode* post_call_node = post.as<CallNode>(); + CHECK(post_call_node) << "Expected a CallNode, but got " << post; + + Expr cur_op = pre_call_node->op; + for (auto arg : pre_call_node->args) { + auto maybe_var_node = arg.as<VarNode>(); + if (maybe_var_node) { + auto var = Downcast<Var>(arg); + auto it = input_transform_map_.find(var); + if (it != input_transform_map_.end()) { + // Checks that the function-level input var hasn't been an arg + // to a CallNode yet. + CHECK(!it->second) << "Function input with name '" << var->name_hint() + << "' is fed into more than one call; " + << "aborting transformation"; + + it->second = pre_call_node; + + // Get the type to transform the function signature to + DataType out_dtype; + if (cur_op == cast_op_) { + auto attrs = pre_call_node->attrs.as<CastAttrs>(); + out_dtype = attrs->dtype; + } else if (cur_op == quantize_op_) { + auto attrs = pre_call_node->attrs.as<qnn::QuantizeAttrs>(); + out_dtype = attrs->out_dtype; + } else { + CHECK(false) << "FoldTypeTransformation will only fold cast and " + << "quantize type transformations"; + } + + // Mutate the var node type + VarNode* var_node = const_cast<VarNode*>(maybe_var_node); + const TensorTypeNode* anno = var_node->type_annotation.as<TensorTypeNode>(); + auto mut_anno = const_cast<TensorTypeNode*>(anno); + auto shape = anno->shape; + mut_anno->dtype = out_dtype; + + return GetRef<Expr>(var_node); Review comment: It's required to remove the cast/quantize call, otherwise if the call is returned those operators are persisted in the graph. -- 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]
