conftest.py 10.6 KB
Newer Older
1
import sys
2
3
4
5
6
7
import subprocess
import contextlib
import pytest
import asyncio
import os
import docker
8
9
10
import json
import math
import time
11
import random
12
13

from docker.errors import NotFound
14
15
16
from typing import Optional, List, Dict
from syrupy.extensions.json import JSONSnapshotExtension
from aiohttp import ClientConnectorError, ClientOSError, ServerDisconnectedError
17
18

from text_generation import AsyncClient
19
from text_generation.types import Response, Details, InputToken, Token, BestOfSequence
20
21
22
23
24
25

DOCKER_IMAGE = os.getenv("DOCKER_IMAGE", None)
HUGGING_FACE_HUB_TOKEN = os.getenv("HUGGING_FACE_HUB_TOKEN", None)
DOCKER_VOLUME = os.getenv("DOCKER_VOLUME", "/data")


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
class ResponseComparator(JSONSnapshotExtension):
    def serialize(
        self,
        data,
        *,
        exclude=None,
        matcher=None,
    ):
        if isinstance(data, List):
            data = [d.dict() for d in data]

        data = self._filter(
            data=data, depth=0, path=(), exclude=exclude, matcher=matcher
        )
        return json.dumps(data, indent=2, ensure_ascii=False, sort_keys=False) + "\n"

    def matches(
        self,
        *,
        serialized_data,
        snapshot_data,
    ) -> bool:
        def convert_data(data):
            data = json.loads(data)

            if isinstance(data, Dict):
                return Response(**data)
            if isinstance(data, List):
                return [Response(**d) for d in data]
            raise NotImplementedError

        def eq_token(token: Token, other: Token) -> bool:
            return (
                token.id == other.id
                and token.text == other.text
                and math.isclose(token.logprob, other.logprob, rel_tol=0.2)
                and token.special == other.special
            )

65
        def eq_prefill_token(prefill_token: InputToken, other: InputToken) -> bool:
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
            try:
                return (
                    prefill_token.id == other.id
                    and prefill_token.text == other.text
                    and (
                        math.isclose(prefill_token.logprob, other.logprob, rel_tol=0.2)
                        if prefill_token.logprob is not None
                        else prefill_token.logprob == other.logprob
                    )
                )
            except TypeError:
                return False

        def eq_best_of(details: BestOfSequence, other: BestOfSequence) -> bool:
            return (
                details.finish_reason == other.finish_reason
                and details.generated_tokens == other.generated_tokens
                and details.seed == other.seed
                and len(details.prefill) == len(other.prefill)
                and all(
                    [
                        eq_prefill_token(d, o)
                        for d, o in zip(details.prefill, other.prefill)
                    ]
                )
                and len(details.tokens) == len(other.tokens)
                and all([eq_token(d, o) for d, o in zip(details.tokens, other.tokens)])
            )

        def eq_details(details: Details, other: Details) -> bool:
            return (
                details.finish_reason == other.finish_reason
                and details.generated_tokens == other.generated_tokens
                and details.seed == other.seed
                and len(details.prefill) == len(other.prefill)
                and all(
                    [
                        eq_prefill_token(d, o)
                        for d, o in zip(details.prefill, other.prefill)
                    ]
                )
                and len(details.tokens) == len(other.tokens)
                and all([eq_token(d, o) for d, o in zip(details.tokens, other.tokens)])
                and (
                    len(details.best_of_sequences)
                    if details.best_of_sequences is not None
                    else 0
                )
                == (
                    len(other.best_of_sequences)
                    if other.best_of_sequences is not None
                    else 0
                )
                and (
                    all(
                        [
                            eq_best_of(d, o)
                            for d, o in zip(
                                details.best_of_sequences, other.best_of_sequences
                            )
                        ]
                    )
                    if details.best_of_sequences is not None
                    else details.best_of_sequences == other.best_of_sequences
                )
            )

        def eq_response(response: Response, other: Response) -> bool:
            return response.generated_text == other.generated_text and eq_details(
                response.details, other.details
            )

        serialized_data = convert_data(serialized_data)
        snapshot_data = convert_data(snapshot_data)

        if not isinstance(serialized_data, List):
            serialized_data = [serialized_data]
        if not isinstance(snapshot_data, List):
            snapshot_data = [snapshot_data]

        return len(snapshot_data) == len(serialized_data) and all(
            [eq_response(r, o) for r, o in zip(serialized_data, snapshot_data)]
        )


class LauncherHandle:
    def __init__(self, port: int):
        self.client = AsyncClient(f"http://localhost:{port}")

    def _inner_health(self):
        raise NotImplementedError

    async def health(self, timeout: int = 60):
        assert timeout > 0
        for _ in range(timeout):
            if not self._inner_health():
                raise RuntimeError("Launcher crashed")

            try:
                await self.client.generate("test")
                return
            except (ClientConnectorError, ClientOSError, ServerDisconnectedError) as e:
                time.sleep(1)
        raise RuntimeError("Health check failed")


class ContainerLauncherHandle(LauncherHandle):
    def __init__(self, docker_client, container_name, port: int):
        super(ContainerLauncherHandle, self).__init__(port)
        self.docker_client = docker_client
        self.container_name = container_name

    def _inner_health(self) -> bool:
        container = self.docker_client.containers.get(self.container_name)
        return container.status in ["running", "created"]


class ProcessLauncherHandle(LauncherHandle):
    def __init__(self, process, port: int):
        super(ProcessLauncherHandle, self).__init__(port)
        self.process = process

    def _inner_health(self) -> bool:
        return self.process.poll() is None


192
@pytest.fixture
193
194
def response_snapshot(snapshot):
    return snapshot.use_extension(ResponseComparator)
195
196
197
198
199
200
201
202
203
204
205
206
207


@pytest.fixture(scope="module")
def event_loop():
    loop = asyncio.get_event_loop()
    yield loop
    loop.close()


@pytest.fixture(scope="module")
def launcher(event_loop):
    @contextlib.contextmanager
    def local_launcher(
208
209
210
211
        model_id: str,
        num_shard: Optional[int] = None,
        quantize: Optional[str] = None,
        trust_remote_code: bool = False,
212
    ):
213
214
        port = random.randint(8000, 10_000)
        master_port = random.randint(10_000, 20_000)
215

216
217
218
        shard_uds_path = (
            f"/tmp/tgi-tests-{model_id.split('/')[-1]}-{num_shard}-{quantize}-server"
        )
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235

        args = [
            "text-generation-launcher",
            "--model-id",
            model_id,
            "--port",
            str(port),
            "--master-port",
            str(master_port),
            "--shard-uds-path",
            shard_uds_path,
        ]

        if num_shard is not None:
            args.extend(["--num-shard", str(num_shard)])
        if quantize:
            args.append("--quantize")
236
237
238
            args.append("bitsandbytes")
        if trust_remote_code:
            args.append("--trust-remote-code")
239

240
241
242
        env = os.environ
        env["LOG_LEVEL"] = "info,text_generation_router=debug"

243
        with subprocess.Popen(
244
            args, stdout=subprocess.PIPE, stderr=subprocess.PIPE, env=env
245
        ) as process:
246
            yield ProcessLauncherHandle(process, port)
247
248
249
250
251

            process.terminate()
            process.wait(60)

            launcher_output = process.stdout.read().decode("utf-8")
252
            print(launcher_output, file=sys.stderr)
253
254
255
256
257
258

            process.stdout.close()
            process.stderr.close()

    @contextlib.contextmanager
    def docker_launcher(
259
260
261
262
        model_id: str,
        num_shard: Optional[int] = None,
        quantize: Optional[str] = None,
        trust_remote_code: bool = False,
263
    ):
264
        port = random.randint(8000, 10_000)
265
266
267
268
269
270
271

        args = ["--model-id", model_id, "--env"]

        if num_shard is not None:
            args.extend(["--num-shard", str(num_shard)])
        if quantize:
            args.append("--quantize")
272
273
274
            args.append("bitsandbytes")
        if trust_remote_code:
            args.append("--trust-remote-code")
275
276
277
278
279
280
281
282
283
284
285
286
287
288

        client = docker.from_env()

        container_name = f"tgi-tests-{model_id.split('/')[-1]}-{num_shard}-{quantize}"

        try:
            container = client.containers.get(container_name)
            container.stop()
            container.wait()
        except NotFound:
            pass

        gpu_count = num_shard if num_shard is not None else 1

289
        env = {"LOG_LEVEL": "info,text_generation_router=debug"}
290
291
292
293
294
295
296
297
298
299
300
301
        if HUGGING_FACE_HUB_TOKEN is not None:
            env["HUGGING_FACE_HUB_TOKEN"] = HUGGING_FACE_HUB_TOKEN

        volumes = []
        if DOCKER_VOLUME:
            volumes = [f"{DOCKER_VOLUME}:/data"]

        container = client.containers.run(
            DOCKER_IMAGE,
            command=args,
            name=container_name,
            environment=env,
302
            auto_remove=False,
303
304
305
306
307
308
309
310
            detach=True,
            device_requests=[
                docker.types.DeviceRequest(count=gpu_count, capabilities=[["gpu"]])
            ],
            volumes=volumes,
            ports={"80/tcp": port},
        )

311
        yield ContainerLauncherHandle(client, container.name, port)
312

313
314
315
316
317
        try:
            container.stop()
            container.wait()
        except NotFound:
            pass
318
319

        container_output = container.logs().decode("utf-8")
320
        print(container_output, file=sys.stderr)
321

322
323
        container.remove()

324
325
326
327
328
329
330
331
332
333
334
    if DOCKER_IMAGE is not None:
        return docker_launcher
    return local_launcher


@pytest.fixture(scope="module")
def generate_load():
    async def generate_load_inner(
        client: AsyncClient, prompt: str, max_new_tokens: int, n: int
    ) -> List[Response]:
        futures = [
335
336
337
338
            client.generate(
                prompt, max_new_tokens=max_new_tokens, decoder_input_details=True
            )
            for _ in range(n)
339
340
        ]

341
        return await asyncio.gather(*futures)
342
343

    return generate_load_inner