On this page, I’m going to collect interesting articles about ML compilers, and (maybe) review research papers on this topic.
| Article | Notes |
|---|---|
| A friendly introduction to machine learning compilers and optimizers | This is quite a good article for absolute newbies in this field. Paragraph 5 describes the main problem of all ML compilers: it’s quite difficult to say whether some set of optimizations would give a performance benefit or not for a particular model. That’s why, in a lot of cases, manually-tuned optimizations are in charge now. |
| AI Compilers Demystified | The article contains the list of the most popular and famous AI compilers with descriptions of their main features. Also you can find here a brief introduction to compilation pipeline. The picture of high-level design architecture of DL compilers shows a lot of interesting concepts that are have to be learned for deep understating of this field. |