atomicfrontiercode / keyboards Goto Github PK

What do I have to change to use a different keyboard layout, e.g. In Germany we don't use QWERTY, but QWERTZ and also have some special characters.

Watched the video before looking at the repo. Out of curiosity, how can we actually use the optimal keyboard layout? Currently, I'm using a Mac. Are there any discussion pages, wikis, or extra documentation we can read?

Hi,

it seems like on the keyboard you forgot the period

When the code git clone

https://github.com/AtomicFrontierCode/keyboards.git
cd keyboards
julia -L 'keyboardSA.jl'

is put into the MacOS julia terminal, the response

ERROR: syntax: extra token "clone" after end of expression
Stacktrace:
[1] top-level scope
@ none:1

I am not experienced with coding or the Julia language and have absolutely no clue how to fix this.

Moving understood keys, such as 1-9, can effectively lower efficiency due to real-world factors, rather than a cpu bias, is there a way this code can be updated to say, lock the top row and continue randperm-ing the other keys?
all major keyboards have this layout reguardless of alpha keying, and i think it would be neat to apply this real-world.
This could be in a config file elsewhere as a t-f or just in the code for people who want to change that.
Thanks

I'm pretty dead set on the punctuation keys being where they are. I personally want to keep them there. Can I lock those (or other) keys to stay in place throughout the annealing process?

Not familiar with Julia got the error when trying to run it from the console with julia keyboardSA.jl:

- Run `import Pkg; Pkg.add("Plots")` to install the Plots package.
Stacktrace:
 [1] macro expansion
   @ .\loading.jl:1630 [inlined]
 [2] macro expansion
   @ .\lock.jl:267 [inlined]
 [3] require(into::Module, mod::Symbol)
   @ Base .\loading.jl:1611```

Can you help how can I fix this?

I realised an ideal dataset for this would using ones for training LLMs. E.g. https://kili-technology.com/large-language-models-llms/9-open-sourced-datasets-for-training-large-language-models

A cool project would be to find the optimal layout for different spoken languages, in addition to coding languages which are also included in the LLM datasets.

It would be interesting to create an open source 'base layout' for writing, coding, and writing+coding, which then allow for 'fine tuning' using a person's code and/or writing.

My personal keyboard layout :

https://raw.githubusercontent.com/wenuam/wm_key_layout_cfilorux/main/key/tuxedocomputers/keyboard-layouts/Sirius%2016%20(Gen1)/TUXEDO%20Sirius%2016%20-%20Gen1%20-%20WM%20ISO_v3.svg

It's alphabetical ("pyramid" orientation) and programmer focused. Accents are through "dead keys".

Regards.

the variable in the julia file bookPath = "MyBook.txt" is not the same as the filename myBook.txt.

It's good for reproducibility, and helps people not run into issues like #4.

All people need to do is run pkg> instantiate to set up the package environment.

In the video you spoke about the 80/20 of massive gains, minimal changes.

I don't see any variables that appear to tweak the algo to favor fewer changes. Is that in here? If not, do you still have that draft of the code?

After around 930 lines of the terminal displaying temperature 0.0, the program crashes with the error:

ERROR: LoadError: MethodError: no method matching Int64(::String)

Closest candidates are:
  (::Type{T})(::AbstractChar) where T<:Union{Int32, Int64}
   @ Base char.jl:51
  (::Type{T})(::AbstractChar) where T<:Union{AbstractChar, Number}
   @ Base char.jl:50
  (::Type{T})(::BigInt) where T<:Union{Int128, Int16, Int32, Int64, Int8}
   @ Base gmp.jl:367
  ...

Stacktrace:
 [1] drawKeyboard(myGenome::Vector{String}, id::String)
   @ Main C:\<DIR>\keyboards\keyboardSA.jl:488
 [2] runSA()
   @ Main C:\<DIR>\keyboards\keyboardSA.jl:789
 [3] top-level scope
   @ C:\<DIR>\keyboards\keyboardSA.jl:798
in expression starting at C:\<DIR>\keyboards\keyboardSA.jl:798

The last png saved was 23836.png (Possible Int16 overflow after that?), and my last iteration score was 0.01, 23836, -37.87043, -37.88911

I'm looking to create a programming language specific symbolic layout. I wouldn't expect the layout to have as many keys as the current layouts. Could you point me in a direction for what would need to be modified to make that work?

My assumption is that I'll be training against a few codebases in the language I'm planning on targeting.