Create performance tips docs section #615
Conversation
Co-Authored-By: oxinabox <oxinabox@ucc.asn.au>
Co-Authored-By: oxinabox <oxinabox@ucc.asn.au>
|
CI failure is an unrelated network issue with downloading some test data. |
| Flux works great with all kinds of number types. | ||
| But often you do not need to be working with say `Float64` (let alone `BigFloat`). | ||
| Switching to `Float32` can give you a significant speed up, | ||
| not because the operations are faster, but because the memory usage is halved. |
There was a problem hiding this comment.
Seems relevant to mention Float32 on GPU here. Also, operations do tend to be faster since you can fit more numbers in a SIMD lane at a given size.
There was a problem hiding this comment.
Feel free to use the github suggestion feature, or to PR after it is merged.
I know little of GPU so someone else is better to write it.
There was a problem hiding this comment.
You could just remove the "not because the operations are faster, but because the memory usage is halved." part?
| For example: | ||
|
|
||
| ``` | ||
| my_tanh(x) = Float64(tanh(x)) |
There was a problem hiding this comment.
This seems a bit artificial, perhaps something like tanh(x) + 2.0 or 5.0 * tanh(x) etc.
There was a problem hiding this comment.
I wanted something very obvious, for the first example, the one below is less obvious.
Co-Authored-By: oxinabox <oxinabox@ucc.asn.au>
Co-Authored-By: oxinabox <oxinabox@ucc.asn.au>
Co-Authored-By: oxinabox <oxinabox@ucc.asn.au>
|
Merge? |
Co-Authored-By: oxinabox <oxinabox@ucc.asn.au>
|
Great stuff and sorely needed, thanks a lot @oxinabox and reviewers. |
| Not only should your activation functions be [type-stable](https://docs.julialang.org/en/v1/manual/performance-tips/#Write-%22type-stable%22-functions-1), | ||
| they should also preserve the type of their inputs. | ||
|
|
||
| A very artificial example using an activatioon function like |
| And you use less memory. | ||
|
|
||
|
|
||
| ## Make sure your custom activation functions preserve the type of their inputs |
| you will see a large slow-down | ||
|
|
||
| This can occur sneakily, because you can cause type-promotion by interacting with a numeric literals. | ||
| E.g. the following will have run into the same problem as above: |
There was a problem hiding this comment.
| E.g. the following will have run into the same problem as above: | |
| E.g. the following will run into the same problem as above: |
| leaky_tanh(x) = 0.01x + tanh(x) | ||
| ``` | ||
|
|
||
| While one could change your activation function (e.g. to use `0.01f0x`) to avoid this when ever your inputs change, |
There was a problem hiding this comment.
| While one could change your activation function (e.g. to use `0.01f0x`) to avoid this when ever your inputs change, | |
| While one could change the activation function (e.g. to use `0.01f0x`) to avoid this when ever your inputs change, |
|
@KristofferC ah noo. |
In part in response to #613,
but more generally I think good to have.