1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
|
{
lib,
stdenv,
buildPythonPackage,
fetchFromGitHub,
# build-system
setuptools,
# dependencies
cloudevents,
fastapi,
grpc-interceptor,
grpcio,
grpcio-tools,
httpx,
kubernetes,
numpy,
orjson,
pandas,
prometheus-client,
protobuf,
psutil,
pydantic,
python-dateutil,
pyyaml,
six,
tabulate,
timing-asgi,
uvicorn,
# optional-dependencies
# storage
kserve-storage,
# logging
asgi-logger,
# ray
ray,
# llm
vllm,
# tests
avro,
grpcio-testing,
jinja2,
pytest-asyncio,
pytest-cov-stub,
pytest-httpx,
pytest-xdist,
pytestCheckHook,
tomlkit,
}:
buildPythonPackage rec {
pname = "kserve";
version = "0.16.0";
pyproject = true;
src = fetchFromGitHub {
owner = "kserve";
repo = "kserve";
tag = "v${version}";
hash = "sha256-f6ILZMLxfckEpy7wSgCqUx89JWSnn0DbQiqRSHcQHms=";
};
# Fix vllm 0.12.0 compatibility
# Patch submitted upstream: https://github.com/kserve/kserve/pull/4882
postPatch = ''
substituteInPlace kserve/protocol/rest/openai/types/__init__.py \
--replace-fail \
"from vllm.entrypoints.openai.protocol import EmbeddingRequest, EmbeddingResponse as Embedding, EmbeddingResponseData, EmbeddingCompletionRequest" \
"from vllm.entrypoints.pooling.embed.protocol import EmbeddingRequest, EmbeddingResponse as Embedding, EmbeddingResponseData, EmbeddingCompletionRequest" \
--replace-fail \
"from vllm.entrypoints.openai.protocol import RerankRequest, RerankResponse as Rerank" \
"from vllm.entrypoints.pooling.score.protocol import RerankRequest, RerankResponse as Rerank"
'';
sourceRoot = "${src.name}/python/kserve";
pythonRelaxDeps = [
"fastapi"
"httpx"
"numpy"
"prometheus-client"
"protobuf"
"uvicorn"
"psutil"
];
build-system = [
setuptools
];
dependencies = [
cloudevents
fastapi
grpc-interceptor
grpcio
grpcio-tools
httpx
kubernetes
numpy
orjson
pandas
prometheus-client
protobuf
psutil
pydantic
python-dateutil
pyyaml
six
tabulate
timing-asgi
uvicorn
]
++ uvicorn.optional-dependencies.standard;
optional-dependencies = {
storage = [
kserve-storage
];
logging = [
asgi-logger
];
ray = [
ray
]
++ ray.optional-dependencies.serve;
llm = [
vllm
];
};
nativeCheckInputs = [
avro
grpcio-testing
jinja2
pytest-asyncio
pytest-cov-stub
pytest-httpx
pytest-xdist
pytestCheckHook
tomlkit
]
++ lib.concatAttrValues optional-dependencies;
pythonImportsCheck = [ "kserve" ];
disabledTestPaths = [
# Looks for a config file at the root of the repository
"test/test_inference_service_client.py"
# AssertionError
"test/test_server.py::TestTFHttpServerLoadAndUnLoad::test_unload"
# Race condition when called concurrently between two instances of the same model (i.e. in nixpkgs-review)
"test/test_dataplane.py::TestDataPlane::test_model_metadata[TEST_RAY_SERVE_MODEL]"
]
++ lib.optionals stdenv.hostPlatform.isDarwin [
# RuntimeError: Failed to start GCS
"test/test_dataplane.py::TestDataPlane::test_explain"
"test/test_dataplane.py::TestDataPlane::test_infer"
"test/test_dataplane.py::TestDataPlane::test_model_metadata"
"test/test_dataplane.py::TestDataPlane::test_server_readiness"
"test/test_server.py::TestRayServer::test_explain"
"test/test_server.py::TestRayServer::test_health_handler"
"test/test_server.py::TestRayServer::test_infer"
"test/test_server.py::TestRayServer::test_list_handler"
"test/test_server.py::TestRayServer::test_liveness_handler"
"test/test_server.py::TestRayServer::test_predict"
# Permission Error
"test/test_server.py::TestMutiProcessServer::test_rest_server_multiprocess"
];
disabledTests = [
# Started failing since vllm was updated to 0.13.0
# pydantic_core._pydantic_core.ValidationError: 1 validation error for RerankResponse
# usage.prompt_tokens
# Field required [type=missing, input_value={'total_tokens': 100}, input_type=dict]
# For further information visit https://errors.pydantic.dev/2.11/v/missing
"test_create_rerank"
"test_create_embedding"
# AssertionError: assert CompletionReq...lm_xargs=None) == CompletionReq...lm_xargs=None)
"test_convert_params"
# Flaky: ray.exceptions.ActorDiedError: The actor died unexpectedly before finishing this task.
"test_explain"
"test_infer"
"test_predict"
# Require network access
"test_infer_graph_endpoint"
"test_infer_path_based_routing"
# Tries to access `/tmp` (hardcoded)
"test_local_path_with_out_dir_exist"
]
++ lib.optionals stdenv.hostPlatform.isDarwin [
"test_local_path_with_out_dir_not_exist"
];
__darwinAllowLocalNetworking = true;
meta = {
description = "Standardized Serverless ML Inference Platform on Kubernetes";
homepage = "https://github.com/kserve/kserve/tree/master/python/kserve";
changelog = "https://github.com/kserve/kserve/releases/tag/${src.tag}";
license = lib.licenses.asl20;
maintainers = with lib.maintainers; [ GaetanLepage ];
};
}
|
