As a hard to force the programmer in achieving a variety of needs, there is always need to call a variety of API, that is, to read all kinds of documents. We have to do is figure out these parameters can then call the API. Speaking very simple, is actually not the case. Now popular machine learning, essentially in a variety of transfer package and parameter adjustment . But this transfer package is slightly more complex: numpy, pandas, matplotlib, sklearn , pytorch, tensorflow ....... It can be said quickly learn to call the API for programmers is an important capability.
I Here are some specific cases of API calls:
- For some very common API, for example, like the python's string manipulation functions, we do not need to refer to the official documents, direct Baidu search for some blog, to meet their own needs. Of course, I personally think that if you are able to better read the official document, because official documents are the most accurate.
- If some less common API, the information can be found on Baidu little, this time on the need to refer to the official documents.
- The first two cases is relatively easy to solve, sometimes, some API does not provide documentation of! ! ! Or documentation written pulpy, simply can not read. This time, we need to from the code start with. The most primitive way, directly inside a package directory, look at each source file. There is also a better way to view the source code API through IDE, here to vscode for example, the mouse to select the function name, press ctrl and click you can see the corresponding source code. Subsequently, by reading the source to learn how to use the API.