electriclilies commented on a change in pull request #9357: URL: https://github.com/apache/tvm/pull/9357#discussion_r737907476
########## 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: Alright, this is OK for the first iteration -- 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]
