colin/nix-files
1
0
Fork 1
Code Issues Pull Requests Projects Releases Wiki Activity

ollama: add new models: qwq-32b (from the qwen team), marco-o1 (also from Alibaba)

Browse Source
This commit is contained in:
Colin
2024-12-05 21:46:04 +00:00
parent 410d63f08e
commit b0059e74dc
4 changed files with 26 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
modules/services/ollama.nix
View File

@@ -35,6 +35,7 @@ let
# llama3_1-70b # generates like 1 word/sec, decent output (comparable to qwen2_5-32b)
llama3_2-3b
magicoder-7b # it generates valid, if sparse, code
marco-o1-7b # untested
# mistral-7b # it generates invalid code
# mistral-nemo-12b # it generates invalid code
mistral-small-22b # quality comparable to qwen2_5
@@ -46,6 +47,7 @@ let
# qwen2_5-32b-instruct-q2_K # lower-res version of default 32b (so, slightly faster, but generates invalid code where the full res generates valid code)
qwen2_5-32b # generates 3~5 words/sec, but notably more accurate than coder-7b
# qwen2_5-coder-7b # fast, and concise, but generates invalid code
qwq-32b # untested
# solar-pro-22b # generates invalid code
# starcoder2-15b-instruct # it gets stuck
# wizardlm2-7b # generates invalid code

10
pkgs/by-name/ollamaPackages/marco-o1-7b.nix Normal file
View File

@@ -0,0 +1,10 @@
# <https://ollama.com/library/marco-o1>
{ mkOllamaModel }: mkOllamaModel {
modelName = "marco-o1";
variant = "7b";
manifestHash = "sha256-RUs48/4zT8/ylZMBOjc09ED87mXgUVfkCjT2hOOfBDo=";
modelBlob = "234ea779a388b986a6c961440f175e75c7cd336f70614d7c5b043c09a5931ad7";
modelBlobHash = "sha256-I06neaOIuYamyWFEDxdedcfNM29wYU18WwQ8CaWTGtc=";
systemBlob = "8c772364849c7ab9cc387c82994af4e17861dabc91d9937e48ec6db602783c6a";
systemBlobHash = "sha256-jHcjZIScernMOHyCmUr04Xhh2ryR2ZN+SOxttgJ4PGo=";
}

2
pkgs/by-name/ollamaPackages/package.nix
View File

@@ -30,6 +30,7 @@ lib.recurseIntoAttrs (lib.makeScope newScope (self: with self; {
llama3_1-70b = callPackage ./llama3_1-70b.nix { };
llama3_2-3b = callPackage ./llama3_2-3b.nix { };
magicoder-7b = callPackage ./magicoder-7b.nix { };
marco-o1-7b = callPackage ./marco-o1-7b.nix { };
mistral-7b = callPackage ./mistral-7b.nix { };
mistral-nemo-12b = callPackage ./mistral-nemo-12b.nix { };
mistral-small-22b = callPackage ./mistral-small-22b.nix { };
@@ -41,6 +42,7 @@ lib.recurseIntoAttrs (lib.makeScope newScope (self: with self; {
qwen2_5-32b = callPackage ./qwen2_5-32b.nix { };
qwen2_5-32b-instruct-q2_K = callPackage ./qwen2_5-32b-instruct-q2_K.nix { };
qwen2_5-coder-7b = callPackage ./qwen2_5-coder-7b.nix { };
qwq-32b = callPackage ./qwq-32b.nix { };
solar-pro-22b = callPackage ./solar-pro-22b.nix { };
starcoder2-15b-instruct = callPackage ./starcoder2-15b-instruct.nix { };
wizardlm2-7b = callPackage ./wizardlm2-7b.nix { };

12
pkgs/by-name/ollamaPackages/qwq-32b.nix Normal file
View File

@@ -0,0 +1,12 @@
# <https://ollama.com/library/qwq>
{ mkOllamaModel }: mkOllamaModel {
modelName = "qwq";
variant = "32b";
manifestHash = "sha256-9i224DAvi8Onzjzaa0nh5HX2sfcTITDow1J0KVOdrBA=";
modelBlob = "fa711da9efad72a93edc6b16aadc2dba1345f3c16491350483104978b1e89c91";
modelBlobHash = "sha256-+nEdqe+tcqk+3GsWqtwtuhNF88FkkTUEgxBJeLHonJE=";
paramsBlob = "f02dd72bb2423204352eabc5637b44d79d17f109fdb510a7c51455892aa2d216";
paramsBlobHash = "sha256-8C3XK7JCMgQ1LqvFY3tE150X8Qn9tRCnxRRViSqi0hY=";
systemBlob = "50795fcf7f4df9be65090ce980b63e8d82f51b5cbad9492fa504aefa3ad3c9a0";
systemBlobHash = "sha256-UHlfz39N+b5lCQzpgLY+jYL1G1y62UkvpQSu+jrTyaA=";
}