vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

nix: remove nixConfig from flake.nix

Browse source 
Why? I've demoed `nix run ggerganov/llama.cpp` to three people now, and
the prompts caused by this setting have been a fly in the ointment every
time, where I have to hand-wave away what's being asked... and because
it happens on every single command run, it's a major pain in the ass.

It's fine to leave it in the flake for those who read it, but the Nix
ecosystem isn't yet ready to have flakes with nixConfig, in my
judgement.
This commit is contained in:
Philip Taron 2024-01-16 09:23:53 -08:00
parent 862f5e41ab
commit a41c736c2f
No known key found for this signature in database
1 changed files with 35 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

57
flake.nix
View file

@ -6,28 +6,41 @@
flake-parts.url = "github:hercules-ci/flake-parts"; flake-parts.url = "github:hercules-ci/flake-parts";
}; };
# Optional binary cache # There's an optional binary cache available. The details are below, but they're commented out.
nixConfig = { #
extra-substituters = [ # Why? The terrible experience of being prompted to accept them on every single Nix command run.
# Populated by the CI in ggerganov/llama.cpp # Plus, there are warnings shown about not being a trusted user on a default Nix install
"https://llama-cpp.cachix.org" # if you *do* say yes to the prompts.
#
# A development cache for nixpkgs imported with `config.cudaSupport = true`. # This experience makes having `nixConfig` in a flake a persistent UX problem.
# Populated by https://hercules-ci.com/github/SomeoneSerge/nixpkgs-cuda-ci. #
# This lets one skip building e.g. the CUDA-enabled openmpi. # To make use of the binary cache, please add the relevant settings to your `nix.conf`.
# TODO: Replace once nix-community obtains an official one. # It's located at `/etc/nix/nix.conf` on non-NixOS systems. On NixOS, adjust the `nix.settings`
"https://cuda-maintainers.cachix.org" # option in your NixOS configuration to add `extra-substituters` and `extra-trusted-public-keys`,
]; # as shown below.
#
# Verify these are the same keys as published on # ```
# - https://app.cachix.org/cache/llama-cpp # nixConfig = {
# - https://app.cachix.org/cache/cuda-maintainers # extra-substituters = [
extra-trusted-public-keys = [ # # Populated by the CI in ggerganov/llama.cpp
"llama-cpp.cachix.org-1:H75X+w83wUKTIPSO1KWy9ADUrzThyGs8P5tmAbkWhQc=" # "https://llama-cpp.cachix.org"
"cuda-maintainers.cachix.org-1:0dq3bujKpuEPMCX6U4WylrUDZ9JyUG0VpVZa7CNfq5E=" #
]; # # A development cache for nixpkgs imported with `config.cudaSupport = true`.
}; # # Populated by https://hercules-ci.com/github/SomeoneSerge/nixpkgs-cuda-ci.
# # This lets one skip building e.g. the CUDA-enabled openmpi.
# # TODO: Replace once nix-community obtains an official one.
# "https://cuda-maintainers.cachix.org"
# ];
#
# # Verify these are the same keys as published on
# # - https://app.cachix.org/cache/llama-cpp
# # - https://app.cachix.org/cache/cuda-maintainers
# extra-trusted-public-keys = [
# "llama-cpp.cachix.org-1:H75X+w83wUKTIPSO1KWy9ADUrzThyGs8P5tmAbkWhQc="
# "cuda-maintainers.cachix.org-1:0dq3bujKpuEPMCX6U4WylrUDZ9JyUG0VpVZa7CNfq5E="
# ];
# };
# ```
# For inspection, use `nix flake show github:ggerganov/llama.cpp` or the nix repl: # For inspection, use `nix flake show github:ggerganov/llama.cpp` or the nix repl:
# #