pyrox.dev
1{ lib
2, buildPythonPackage
3, hatch-fancy-pypi-readme
4, hatch-vcs
5, hatchling
6, pytestCheckHook
7, pythonOlder
8, pythonRelaxDepsHook
9, accelerate
10, bentoml
11, bitsandbytes
12, build
13, click
14, ctranslate2
15, datasets
16, docker
17, einops
18, ghapi
19, huggingface-hub
20, hypothesis
21, ipython
22, jupyter
23, jupytext
24, nbformat
25, notebook
26, openai
27, openllm-client
28, openllm-core
29, optimum
30, peft
31, pytest-mock
32, pytest-randomly
33, pytest-rerunfailures
34, pytest-xdist
35, safetensors
36, scipy
37, sentencepiece
38, soundfile
39, syrupy
40, tabulate
41, tiktoken
42, transformers
43, openai-triton
44, xformers
45}:
46
47buildPythonPackage rec {
48 inherit (openllm-core) src version;
49 pname = "openllm";
50 pyproject = true;
51
52 disabled = pythonOlder "3.8";
53
54 sourceRoot = "source/openllm-python";
55
56 nativeBuildInputs = [
57 hatch-fancy-pypi-readme
58 hatch-vcs
59 hatchling
60 pythonRelaxDepsHook
61 ];
62
63 pythonRemoveDeps = [
64 # remove cuda-python as it has an unfree license
65 "cuda-python"
66 ];
67
68 propagatedBuildInputs = [
69 accelerate
70 bentoml
71 bitsandbytes
72 build
73 click
74 einops
75 ghapi
76 openllm-client
77 openllm-core
78 optimum
79 safetensors
80 scipy
81 sentencepiece
82 tabulate
83 transformers
84 ] ++ bentoml.optional-dependencies.io
85 ++ tabulate.optional-dependencies.widechars
86 ++ transformers.optional-dependencies.tokenizers
87 ++ transformers.optional-dependencies.torch;
88
89 passthru.optional-dependencies = {
90 agents = [
91 # diffusers
92 soundfile
93 transformers
94 ] ++ transformers.optional-dependencies.agents;
95 awq = [
96 # autoawq
97 ];
98 baichuan = [
99 # cpm-kernels
100 ];
101 chatglm = [
102 # cpm-kernels
103 ];
104 ctranslate = [
105 ctranslate2
106 ];
107 falcon = [
108 xformers
109 ];
110 fine-tune = [
111 datasets
112 huggingface-hub
113 peft
114 # trl
115 ];
116 ggml = [
117 # ctransformers
118 ];
119 gptq = [
120 # auto-gptq
121 ]; # ++ autogptq.optional-dependencies.triton;
122 grpc = [
123 bentoml
124 ] ++ bentoml.optional-dependencies.grpc;
125 mpt = [
126 openai-triton
127 ];
128 openai = [
129 openai
130 tiktoken
131 ] ++ openai.optional-dependencies.datalib;
132 playground = [
133 ipython
134 jupyter
135 jupytext
136 nbformat
137 notebook
138 ];
139 starcoder = [
140 bitsandbytes
141 ];
142 vllm = [
143 # vllm
144 ];
145 full = with passthru.optional-dependencies; (
146 agents ++ awq ++ baichuan ++ chatglm ++ ctranslate ++ falcon ++ fine-tune ++ ggml ++ gptq ++ mpt ++ openai ++ playground ++ starcoder ++ vllm
147 );
148 all = passthru.optional-dependencies.full;
149 };
150
151 nativeCheckInputs = [
152 docker
153 hypothesis
154 pytest-mock
155 pytest-randomly
156 pytest-rerunfailures
157 pytest-xdist
158 pytestCheckHook
159 syrupy
160 ];
161
162 preCheck = ''
163 export HOME=$TMPDIR
164 # skip GPUs test on CI
165 export GITHUB_ACTIONS=1
166 # disable hypothesis' deadline
167 export CI=1
168 '';
169
170 disabledTestPaths = [
171 # require network access
172 "tests/models"
173 ];
174
175 disabledTests = [
176 # incompatible with recent TypedDict
177 # https://github.com/bentoml/OpenLLM/blob/f3fd32d596253ae34c68e2e9655f19f40e05f666/openllm-python/tests/configuration_test.py#L18-L21
178 "test_missing_default"
179 ];
180
181 pythonImportsCheck = [ "openllm" ];
182
183 meta = with lib; {
184 description = "Operating LLMs in production";
185 homepage = "https://github.com/bentoml/OpenLLM/tree/main/openllm-python";
186 changelog = "https://github.com/bentoml/OpenLLM/blob/${src.rev}/CHANGELOG.md";
187 license = licenses.asl20;
188 maintainers = with maintainers; [ happysalada natsukium ];
189 };
190}