Artificial intelligence is widely used in various industries, and the corresponding developer community has become rich and diverse. They often come from different majors and conduct development practices in different fields and scenarios, which also results in high learning costs for AI developers. Besides learning professional skills, they also need to understand industrial needs and application scenarios. To this end, Machine Heart launched the “GROWTH Plan for AI Developers”, and cooperated with leading AI enterprises to jointly customize themed open courses to help developers complete the whole process from learning to using in a short period of time.
The second “AI Developer Growth Plan” series of open classes will start on April 23 with the theme “MindSpore”.
On March 28, Huawei’s much-anticipated deep learning framework MindSpore was finally opened to the public.
As an “all scene AI framework”, MindSpore is a key part of Huawei’s AI solution, similar to popular deep learning frameworks such as TensorFlow, PyTorch and PaddlePaddle, aiming to significantly lower the threshold for AI application development and enable intelligence to go anywhere.
MindSpore is a unified training and reasoning framework that supports end, edge, and cloud independence/collaboration. Huawei hopes to achieve one-time operator development, consistent development and debugging experience through this complete software stack, so that developers can implement smooth migration capabilities across all devices in a single development.
As the first open source version 0.1.0-alpha of MindSpore, it mainly consists of automatic differentiation, automatic parallelism, data processing and other features. In order to enable developers to learn MindSpore in detail, Machine Heart and Huawei Ceng Institute jointly set up an online open course “Easy to Learn MindSpore”. The course lasts for 3 weeks and consists of 6 courses, with the following links:
Theoretical disassembly: Huawei senior engineers will explain MindSpore’s overall architecture and features of each module in detail to help beginners master relevant knowledge points step by step.
Code demonstration: along with knowledge points, the instructor will demonstrate code operation to help developers quickly get started.
After-school practice: We will set after-school practice assignments periodically and invite developers to complete and submit the results independently. Excellent assignments will be awarded with the book “Deep Learning and MindSpore Practice”. Specific assignment topics, completion methods and evaluation criteria can be found in the notice of subsequent courses.
Full q&A: set up live QA session and learning exchange group, developers in the process of learning and practice have to ask questions.
The specific learning plan is as follows:
Lecture 1 of open lecture
Topic: MindSpore Distributed Automatic Parallel training
Time: 20:00, April 23, Topic explanation + online Q&A + code Demo
Speaker: Ziyan Gong, MindSpore Front-end Development engineer, graduated from university of British Columbia.
Overview: This course will introduce the overall architecture of MindSpore, the automatic parallel and distributed deployment features involved in MindExpression module in MindSpore, and demonstrate the auto-Parallel feature of MindSpore through a practical Demo
Code Demo: ResNET-50 network as an example to demonstrate how to achieve automatic parallel training.
How to join
“MindSpore” series courses are free of charge. Add syncedai6, a small assistant in the heart of the machine, and note “open source”. You can join the learning exchange group, watch live learning theories, do homework and practice together.