jtuyls commented on a change in pull request #14: URL: https://github.com/apache/tvm-rfcs/pull/14#discussion_r683240728
########## File path: rfcs/0012-pipeline-executor.md ########## @@ -0,0 +1,367 @@ +<!--- 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. --> +- Feature Name: (fill me in with a unique identifier, `my_awesome_feature`) +- Start Date: (fill me in with today's date, YYYY-MM-DD) +- RFC PR: [apache/tvm-rfcs#0014](https://github.com/apache/tvm-rfcs/pull/0014) +- GitHub Issue: [apache/tvm#8596](https://github.com/apache/tvm/issues/8596) + +## 1. Summary + + +This proposal introduces Pipeline Executor: A runtime executor that by scheduling +splitted subgraph of relay graph in pipeline to implement task level parallism to +reduce compute latency. Review comment: On point 2, with latency in ML we typically mean the time it takes to process one unit of data, which for image classification is one image if the batch size is 1. In the example given, I would argue that the batch size is still 1 instead of 2 as the data inputs are just being queued and executed separately and not being processed together. Therefore, the latency is not reduced but the throughput is improved. -- 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]
