|
1 |
| -from distilabel.llms.oneai import OneAI |
2 |
| -from typing import Any, Dict |
3 |
| -from unittest import mock |
4 |
| -from unittest.mock import AsyncMock, MagicMock, Mock, patch |
5 |
| - |
6 |
| -import nest_asyncio |
7 |
| -import pytest |
8 |
| -from .utils import DummyUserDetail |
9 |
| - |
10 |
| - |
11 |
| -@patch("openai.AsyncOpenAI") |
12 |
| -class TestOneAI: |
13 |
| - model_id: str = "gpt-4" |
14 |
| - |
15 |
| - def test_oneai_llm(self, _: MagicMock) -> None: |
16 |
| - llm = OneAI(model=self.model_id, api_key="api.key") # type: ignore |
17 |
| - |
18 |
| - assert isinstance(llm, OneAI) |
19 |
| - assert llm.model_name == self.model_id |
20 |
| - |
21 |
| - def test_oneai_llm_env_vars(self, _: MagicMock) -> None: |
22 |
| - with mock.patch.dict(os.environ, clear=True): |
23 |
| - os.environ["01AI_API_KEY"] = "another.api.key" |
24 |
| - os.environ["01AI_BASE_URL"] = "https://example.com" |
25 |
| - |
26 |
| - llm = OneAI(model=self.model_id) |
27 |
| - |
28 |
| - assert isinstance(llm, OneAI) |
29 |
| - assert llm.model_name == self.model_id |
30 |
| - assert llm.base_url == "https://example.com" |
31 |
| - assert llm.api_key.get_secret_value() == "another.api.key" # type: ignore |
32 |
| - |
33 |
| - @pytest.mark.asyncio |
34 |
| - async def test_agenerate(self, mock_openai: MagicMock) -> None: |
35 |
| - llm = OneAI(model=self.model_id, api_key="api.key") # type: ignore |
36 |
| - llm._aclient = mock_openai |
37 |
| - |
38 |
| - mocked_completion = Mock( |
39 |
| - choices=[Mock(message=Mock(content=" Aenean hendrerit aliquam velit. ..."))] |
40 |
| - ) |
41 |
| - llm._aclient.chat.completions.create = AsyncMock(return_value=mocked_completion) |
42 |
| - |
43 |
| - await llm.agenerate( |
44 |
| - input=[ |
45 |
| - {"role": "system", "content": ""}, |
46 |
| - { |
47 |
| - "role": "user", |
48 |
| - "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
49 |
| - }, |
50 |
| - ] |
51 |
| - ) |
52 |
| - |
53 |
| - @pytest.mark.asyncio |
54 |
| - async def test_agenerate_structured(self, mock_openai: MagicMock) -> None: |
55 |
| - llm = OneAI( |
56 |
| - model=self.model_id, |
57 |
| - api_key="api.key", |
58 |
| - structured_output={ |
59 |
| - "schema": DummyUserDetail, |
60 |
| - "mode": "tool_call", |
61 |
| - "max_retries": 1, |
62 |
| - }, |
63 |
| - ) # type: ignore |
64 |
| - llm._aclient = mock_openai |
65 |
| - |
66 |
| - sample_user = DummyUserDetail(name="John Doe", age=30) |
67 |
| - |
68 |
| - llm._aclient.chat.completions.create = AsyncMock(return_value=sample_user) |
69 |
| - |
70 |
| - generation = await llm.agenerate( |
71 |
| - input=[ |
72 |
| - {"role": "system", "content": ""}, |
73 |
| - { |
74 |
| - "role": "user", |
75 |
| - "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
76 |
| - }, |
77 |
| - ] |
78 |
| - ) |
79 |
| - assert generation[0] == sample_user.model_dump_json() |
80 |
| - |
81 |
| - @pytest.mark.skipif( |
82 |
| - sys.version_info < (3, 9), reason="`mistralai` requires Python 3.9 or higher" |
83 |
| - ) |
84 |
| - @pytest.mark.asyncio |
85 |
| - async def test_generate(self, mock_openai: MagicMock) -> None: |
86 |
| - llm = OneAI(model=self.model_id, api_key="api.key") # type: ignore |
87 |
| - llm._aclient = mock_openai |
88 |
| - |
89 |
| - mocked_completion = Mock( |
90 |
| - choices=[Mock(message=Mock(content=" Aenean hendrerit aliquam velit. ..."))] |
91 |
| - ) |
92 |
| - llm._aclient.chat.completions.create = AsyncMock(return_value=mocked_completion) |
93 |
| - |
94 |
| - nest_asyncio.apply() |
95 |
| - |
96 |
| - llm.generate( |
97 |
| - inputs=[ |
98 |
| - [ |
99 |
| - {"role": "system", "content": ""}, |
100 |
| - { |
101 |
| - "role": "user", |
102 |
| - "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
103 |
| - }, |
104 |
| - ] |
105 |
| - ] |
106 |
| - ) |
107 |
| - |
108 |
| - with pytest.raises(ValueError): |
109 |
| - llm.generate( |
110 |
| - inputs=[ |
111 |
| - [ |
112 |
| - {"role": "system", "content": ""}, |
113 |
| - { |
114 |
| - "role": "user", |
115 |
| - "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
116 |
| - }, |
117 |
| - ] |
118 |
| - ], |
119 |
| - response_format="unkown_format", |
120 |
| - ) |
121 |
| - |
122 |
| - @pytest.mark.parametrize( |
123 |
| - "structured_output, dump", |
124 |
| - [ |
125 |
| - ( |
126 |
| - None, |
127 |
| - { |
128 |
| - "model": "gpt-4", |
129 |
| - "generation_kwargs": {}, |
130 |
| - "max_retries": 6, |
131 |
| - "base_url": "https://api.openai.com/v1", |
132 |
| - "timeout": 120, |
133 |
| - "structured_output": None, |
134 |
| - "type_info": { |
135 |
| - "module": "distilabel.llms.oneai", |
136 |
| - "name": "OneAI", |
137 |
| - }, |
138 |
| - }, |
139 |
| - ), |
140 |
| - ( |
141 |
| - { |
142 |
| - "schema": DummyUserDetail.model_json_schema(), |
143 |
| - "mode": "tool_call", |
144 |
| - "max_retries": 1, |
145 |
| - }, |
146 |
| - { |
147 |
| - "model": "gpt-4", |
148 |
| - "generation_kwargs": {}, |
149 |
| - "max_retries": 6, |
150 |
| - "base_url": "https://api.openai.com/v1", |
151 |
| - "timeout": 120, |
152 |
| - "structured_output": { |
153 |
| - "schema": DummyUserDetail.model_json_schema(), |
154 |
| - "mode": "tool_call", |
155 |
| - "max_retries": 1, |
156 |
| - }, |
157 |
| - "type_info": { |
158 |
| - "module": "distilabel.llms.oneai", |
159 |
| - "name": "OneAI", |
160 |
| - }, |
161 |
| - }, |
162 |
| - ), |
163 |
| - ], |
164 |
| - ) |
165 |
| - def test_serialization( |
166 |
| - self, _: MagicMock, structured_output: Dict[str, Any], dump: Dict[str, Any] |
167 |
| - ) -> None: |
168 |
| - llm = OneAI(model=self.model_id, structured_output=structured_output) |
169 |
| - |
170 |
| - assert llm.dump() == dump |
171 |
| - assert isinstance(OneAI.from_dict(dump), OneAI) |
| 1 | +# tests/test_oneai.py |
| 2 | + |
| 3 | +import os |
| 4 | +import sys |
| 5 | +from typing import Any, Dict |
| 6 | +from unittest import mock |
| 7 | +from unittest.mock import AsyncMock, MagicMock, Mock, patch |
| 8 | + |
| 9 | +import nest_asyncio |
| 10 | +import pytest |
| 11 | + |
| 12 | +from distilabel.llms.oneai import OneAI |
| 13 | +from .utils import DummyUserDetail |
| 14 | + |
| 15 | + |
| 16 | +@patch("openai.AsyncOpenAI") |
| 17 | +class TestOneAI: |
| 18 | + model_id: str = "yi-large" |
| 19 | + |
| 20 | + def test_oneai_llm(self, _: MagicMock) -> None: |
| 21 | + llm = OneAI(model=self.model_id, api_key="api.key") # type: ignore |
| 22 | + assert isinstance(llm, OneAI) |
| 23 | + assert llm.model_name == self.model_id |
| 24 | + |
| 25 | + def test_oneai_llm_env_vars(self, _: MagicMock) -> None: |
| 26 | + with mock.patch.dict(os.environ, clear=True): |
| 27 | + os.environ["01AI_API_KEY"] = "another.api.key" |
| 28 | + os.environ["01AI_BASE_URL"] = "https://api.01.ai/v1/chat/completions" |
| 29 | + llm = OneAI(model=self.model_id) |
| 30 | + assert isinstance(llm, OneAI) |
| 31 | + assert llm.model_name == self.model_id |
| 32 | + assert llm.base_url == "https://api.01.ai/v1/chat/completions" |
| 33 | + assert llm.api_key.get_secret_value() == "another.api.key" # type: ignore |
| 34 | + |
| 35 | + @pytest.mark.asyncio |
| 36 | + async def test_agenerate(self, mock_openai: MagicMock) -> None: |
| 37 | + llm = OneAI(model=self.model_id, api_key="api.key") # type: ignore |
| 38 | + llm._aclient = mock_openai |
| 39 | + mocked_completion = Mock( |
| 40 | + choices=[Mock(message=Mock(content=" Aenean hendrerit aliquam velit. ..."))] |
| 41 | + ) |
| 42 | + llm._aclient.chat.completions.create = AsyncMock(return_value=mocked_completion) |
| 43 | + await llm.agenerate( |
| 44 | + input=[ |
| 45 | + {"role": "system", "content": ""}, |
| 46 | + { |
| 47 | + "role": "user", |
| 48 | + "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
| 49 | + }, |
| 50 | + ] |
| 51 | + ) |
| 52 | + |
| 53 | + @pytest.mark.asyncio |
| 54 | + async def test_agenerate_structured(self, mock_openai: MagicMock) -> None: |
| 55 | + llm = OneAI( |
| 56 | + model=self.model_id, |
| 57 | + api_key="api.key", |
| 58 | + structured_output={ |
| 59 | + "schema": DummyUserDetail, |
| 60 | + "mode": "tool_call", |
| 61 | + "max_retries": 1, |
| 62 | + }, |
| 63 | + ) # type: ignore |
| 64 | + llm._aclient = mock_openai |
| 65 | + sample_user = DummyUserDetail(name="John Doe", age=30) |
| 66 | + llm._aclient.chat.completions.create = AsyncMock(return_value=sample_user) |
| 67 | + generation = await llm.agenerate( |
| 68 | + input=[ |
| 69 | + {"role": "system", "content": ""}, |
| 70 | + { |
| 71 | + "role": "user", |
| 72 | + "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
| 73 | + }, |
| 74 | + ] |
| 75 | + ) |
| 76 | + assert generation[0] == sample_user.model_dump_json() |
| 77 | + |
| 78 | + @pytest.mark.skipif( |
| 79 | + sys.version_info < (3, 9), reason="`01ai` requires Python 3.9 or higher" |
| 80 | + ) |
| 81 | + @pytest.mark.asyncio |
| 82 | + async def test_generate(self, mock_openai: MagicMock) -> None: |
| 83 | + llm = OneAI(model=self.model_id, api_key="api.key") # type: ignore |
| 84 | + llm._aclient = mock_openai |
| 85 | + mocked_completion = Mock( |
| 86 | + choices=[Mock(message=Mock(content=" Aenean hendrerit aliquam velit. ..."))] |
| 87 | + ) |
| 88 | + llm._aclient.chat.completions.create = AsyncMock(return_value=mocked_completion) |
| 89 | + nest_asyncio.apply() |
| 90 | + llm.generate( |
| 91 | + inputs=[ |
| 92 | + [ |
| 93 | + {"role": "system", "content": ""}, |
| 94 | + { |
| 95 | + "role": "user", |
| 96 | + "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
| 97 | + }, |
| 98 | + ] |
| 99 | + ] |
| 100 | + ) |
| 101 | + with pytest.raises(ValueError): |
| 102 | + llm.generate( |
| 103 | + inputs=[ |
| 104 | + [ |
| 105 | + {"role": "system", "content": ""}, |
| 106 | + { |
| 107 | + "role": "user", |
| 108 | + "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", |
| 109 | + }, |
| 110 | + ] |
| 111 | + ], |
| 112 | + response_format="unknown_format", |
| 113 | + ) |
| 114 | + |
| 115 | + @pytest.mark.parametrize( |
| 116 | + "structured_output, dump", |
| 117 | + [ |
| 118 | + ( |
| 119 | + None, |
| 120 | + { |
| 121 | + "model": "yi-large", |
| 122 | + "generation_kwargs": {}, |
| 123 | + "max_retries": 6, |
| 124 | + "base_url": "https://api.01.ai/v1/chat/completions", |
| 125 | + "timeout": 120, |
| 126 | + "structured_output": None, |
| 127 | + "type_info": { |
| 128 | + "module": "distilabel.llms.oneai", |
| 129 | + "name": "OneAI", |
| 130 | + }, |
| 131 | + }, |
| 132 | + ), |
| 133 | + ( |
| 134 | + { |
| 135 | + "schema": DummyUserDetail.model_json_schema(), |
| 136 | + "mode": "tool_call", |
| 137 | + "max_retries": 1, |
| 138 | + }, |
| 139 | + { |
| 140 | + "model": "gpt-4", |
| 141 | + "generation_kwargs": {}, |
| 142 | + "max_retries": 6, |
| 143 | + "base_url": "https://api.01.ai/v1/chat/completions", |
| 144 | + "timeout": 120, |
| 145 | + "structured_output": { |
| 146 | + "schema": DummyUserDetail.model_json_schema(), |
| 147 | + "mode": "tool_call", |
| 148 | + "max_retries": 1, |
| 149 | + }, |
| 150 | + "type_info": { |
| 151 | + "module": "distilabel.llms.oneai", |
| 152 | + "name": "OneAI", |
| 153 | + }, |
| 154 | + }, |
| 155 | + ), |
| 156 | + ], |
| 157 | + ) |
| 158 | + def test_serialization( |
| 159 | + self, _: MagicMock, structured_output: Dict[str, Any], dump: Dict[str, Any] |
| 160 | + ) -> None: |
| 161 | + llm = OneAI(model=self.model_id, structured_output=structured_output) |
| 162 | + assert llm.dump() == dump |
| 163 | + assert isinstance(OneAI.from_dict(dump), OneAI) |
0 commit comments