forked from lancedb/lancedb
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtest_embeddings_slow.py
More file actions
518 lines (422 loc) · 16.5 KB
/
test_embeddings_slow.py
File metadata and controls
518 lines (422 loc) · 16.5 KB
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
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright The LanceDB Authors
import importlib
import io
import os
import lancedb
import numpy as np
import pandas as pd
import pyarrow as pa
import pytest
from lancedb.embeddings import get_registry
from lancedb.pydantic import LanceModel, Vector
# These are integration tests for embedding functions.
# They are slow because they require downloading models
# or connection to external api
try:
if importlib.util.find_spec("mlx.core") is not None:
_mlx = True
else:
_mlx = None
except Exception:
_mlx = None
try:
if importlib.util.find_spec("imagebind") is not None:
_imagebind = True
else:
_imagebind = None
except Exception:
_imagebind = None
@pytest.mark.slow
@pytest.mark.parametrize(
"alias", ["sentence-transformers", "openai", "huggingface", "ollama"]
)
def test_basic_text_embeddings(alias, tmp_path):
db = lancedb.connect(tmp_path)
registry = get_registry()
func = registry.get(alias).create(max_retries=0)
func2 = registry.get(alias).create(max_retries=0)
class Words(LanceModel):
text: str = func.SourceField()
text2: str = func2.SourceField()
vector: Vector(func.ndims()) = func.VectorField()
vector2: Vector(func2.ndims()) = func2.VectorField()
table = db.create_table("words", schema=Words)
table.add(
pd.DataFrame(
{
"text": [
"hello world",
"goodbye world",
"fizz",
"buzz",
"foo",
"bar",
"baz",
],
"text2": [
"to be or not to be",
"that is the question",
"for whether tis nobler",
"in the mind to suffer",
"the slings and arrows",
"of outrageous fortune",
"or to take arms",
],
}
)
)
query = "greeting"
actual = (
table.search(query, vector_column_name="vector").limit(1).to_pydantic(Words)[0]
)
vec = func.compute_query_embeddings(query)[0]
expected = (
table.search(vec, vector_column_name="vector").limit(1).to_pydantic(Words)[0]
)
assert actual.text == expected.text
assert actual.text == "hello world"
assert not np.allclose(actual.vector, actual.vector2)
actual = (
table.search(query, vector_column_name="vector2").limit(1).to_pydantic(Words)[0]
)
assert actual.text != "hello world"
assert not np.allclose(actual.vector, actual.vector2)
@pytest.mark.slow
def test_openclip(tmp_path):
import requests
from PIL import Image
db = lancedb.connect(tmp_path)
registry = get_registry()
func = registry.get("open-clip").create(max_retries=0)
class Images(LanceModel):
label: str
image_uri: str = func.SourceField()
image_bytes: bytes = func.SourceField()
vector: Vector(func.ndims()) = func.VectorField()
vec_from_bytes: Vector(func.ndims()) = func.VectorField()
table = db.create_table("images", schema=Images)
labels = ["cat", "cat", "dog", "dog", "horse", "horse"]
uris = [
"http://farm1.staticflickr.com/53/167798175_7c7845bbbd_z.jpg",
"http://farm1.staticflickr.com/134/332220238_da527d8140_z.jpg",
"http://farm9.staticflickr.com/8387/8602747737_2e5c2a45d4_z.jpg",
"http://farm5.staticflickr.com/4092/5017326486_1f46057f5f_z.jpg",
"http://farm9.staticflickr.com/8216/8434969557_d37882c42d_z.jpg",
"http://farm6.staticflickr.com/5142/5835678453_4f3a4edb45_z.jpg",
]
# get each uri as bytes
image_bytes = [requests.get(uri).content for uri in uris]
table.add(
pd.DataFrame({"label": labels, "image_uri": uris, "image_bytes": image_bytes})
)
# text search
actual = (
table.search("man's best friend", vector_column_name="vector")
.limit(1)
.to_pydantic(Images)[0]
)
assert actual.label == "dog"
frombytes = (
table.search("man's best friend", vector_column_name="vec_from_bytes")
.limit(1)
.to_pydantic(Images)[0]
)
assert actual.label == frombytes.label
assert np.allclose(actual.vector, frombytes.vector)
# image search
query_image_uri = "http://farm1.staticflickr.com/200/467715466_ed4a31801f_z.jpg"
image_bytes = requests.get(query_image_uri).content
query_image = Image.open(io.BytesIO(image_bytes))
actual = (
table.search(query_image, vector_column_name="vector")
.limit(1)
.to_pydantic(Images)[0]
)
assert actual.label == "dog"
other = (
table.search(query_image, vector_column_name="vec_from_bytes")
.limit(1)
.to_pydantic(Images)[0]
)
assert actual.label == other.label
arrow_table = table.search().select(["vector", "vec_from_bytes"]).to_arrow()
assert np.allclose(
arrow_table["vector"].combine_chunks().values.to_numpy(),
arrow_table["vec_from_bytes"].combine_chunks().values.to_numpy(),
)
@pytest.mark.skipif(
_imagebind is None,
reason="skip if imagebind not installed.",
)
@pytest.mark.slow
def test_imagebind(tmp_path):
import os
import shutil
import tempfile
import pandas as pd
import requests
from lancedb.embeddings import get_registry
from lancedb.pydantic import LanceModel, Vector
with tempfile.TemporaryDirectory() as temp_dir:
print(f"Created temporary directory {temp_dir}")
def download_images(image_uris):
downloaded_image_paths = []
for uri in image_uris:
try:
response = requests.get(uri, stream=True)
if response.status_code == 200:
# Extract image name from URI
image_name = os.path.basename(uri)
image_path = os.path.join(temp_dir, image_name)
with open(image_path, "wb") as out_file:
shutil.copyfileobj(response.raw, out_file)
downloaded_image_paths.append(image_path)
except Exception as e: # noqa: PERF203
print(f"Failed to download {uri}. Error: {e}")
return temp_dir, downloaded_image_paths
db = lancedb.connect(tmp_path)
registry = get_registry()
func = registry.get("imagebind").create(max_retries=0)
class Images(LanceModel):
label: str
image_uri: str = func.SourceField()
vector: Vector(func.ndims()) = func.VectorField()
table = db.create_table("images", schema=Images)
labels = ["cat", "cat", "dog", "dog", "horse", "horse"]
uris = [
"http://farm1.staticflickr.com/53/167798175_7c7845bbbd_z.jpg",
"http://farm1.staticflickr.com/134/332220238_da527d8140_z.jpg",
"http://farm9.staticflickr.com/8387/8602747737_2e5c2a45d4_z.jpg",
"http://farm5.staticflickr.com/4092/5017326486_1f46057f5f_z.jpg",
"http://farm9.staticflickr.com/8216/8434969557_d37882c42d_z.jpg",
"http://farm6.staticflickr.com/5142/5835678453_4f3a4edb45_z.jpg",
]
temp_dir, downloaded_images = download_images(uris)
table.add(pd.DataFrame({"label": labels, "image_uri": downloaded_images}))
# text search
actual = (
table.search("man's best friend", vector_column_name="vector")
.limit(1)
.to_pydantic(Images)[0]
)
assert actual.label == "dog"
# image search
query_image_uri = [
"https://live.staticflickr.com/65535/33336453970_491665f66e_h.jpg"
]
temp_dir, downloaded_images = download_images(query_image_uri)
query_image_uri = downloaded_images[0]
actual = (
table.search(query_image_uri, vector_column_name="vector")
.limit(1)
.to_pydantic(Images)[0]
)
assert actual.label == "dog"
if os.path.isdir(temp_dir):
shutil.rmtree(temp_dir)
print(f"Deleted temporary directory {temp_dir}")
@pytest.mark.slow
@pytest.mark.skipif(
os.environ.get("COHERE_API_KEY") is None, reason="COHERE_API_KEY not set"
) # also skip if cohere not installed
def test_cohere_embedding_function():
cohere = (
get_registry()
.get("cohere")
.create(name="embed-multilingual-v2.0", max_retries=0)
)
class TextModel(LanceModel):
text: str = cohere.SourceField()
vector: Vector(cohere.ndims()) = cohere.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect("~/lancedb")
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == cohere.ndims()
@pytest.mark.slow
def test_instructor_embedding(tmp_path):
model = get_registry().get("instructor").create(max_retries=0)
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
@pytest.mark.slow
@pytest.mark.skipif(
os.environ.get("GOOGLE_API_KEY") is None, reason="GOOGLE_API_KEY not set"
)
def test_gemini_embedding(tmp_path):
model = get_registry().get("gemini-text").create(max_retries=0)
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
@pytest.mark.skipif(
_mlx is None,
reason="mlx tests only required for apple users.",
)
@pytest.mark.slow
def test_gte_embedding(tmp_path):
model = get_registry().get("gte-text").create()
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
def aws_setup():
try:
import boto3
sts = boto3.client("sts")
sts.get_caller_identity()
return True
except Exception:
return False
@pytest.mark.slow
@pytest.mark.skipif(
not aws_setup(), reason="AWS credentials not set or libraries not installed"
)
def test_bedrock_embedding(tmp_path):
for name in [
"amazon.titan-embed-text-v1",
"cohere.embed-english-v3",
"cohere.embed-multilingual-v3",
]:
model = get_registry().get("bedrock-text").create(max_retries=0, name=name)
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
@pytest.mark.slow
@pytest.mark.skipif(
os.environ.get("OPENAI_API_KEY") is None, reason="OPENAI_API_KEY not set"
)
def test_openai_embedding(tmp_path):
def _get_table(model):
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
return tbl
model = get_registry().get("openai").create(max_retries=0)
tbl = _get_table(model)
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
model = (
get_registry()
.get("openai")
.create(max_retries=0, name="text-embedding-3-large")
)
tbl = _get_table(model)
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
model = (
get_registry()
.get("openai")
.create(max_retries=0, name="text-embedding-3-large", dim=1024)
)
tbl = _get_table(model)
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
@pytest.mark.slow
@pytest.mark.skipif(
os.environ.get("WATSONX_API_KEY") is None
or os.environ.get("WATSONX_PROJECT_ID") is None,
reason="WATSONX_API_KEY and WATSONX_PROJECT_ID not set",
)
def test_watsonx_embedding(tmp_path):
from lancedb.embeddings import WatsonxEmbeddings
for name in WatsonxEmbeddings.model_names():
model = get_registry().get("watsonx").create(max_retries=0, name=name)
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
db = lancedb.connect("~/.lancedb")
tbl = db.create_table("watsonx_test", schema=TextModel, mode="overwrite")
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
@pytest.mark.slow
@pytest.mark.skipif(
os.environ.get("OPENAI_API_KEY") is None, reason="OPENAI_API_KEY not set"
)
def test_openai_with_empty_strs(tmp_path):
model = get_registry().get("openai").create(max_retries=0)
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
df = pd.DataFrame({"text": ["hello world", ""]})
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df, on_bad_vectors="skip")
tb = tbl.to_arrow()
assert tb.schema.field_by_name("vector").type == pa.list_(
pa.float32(), model.ndims()
)
assert len(tb) == 2
assert tb["vector"].is_null().to_pylist() == [False, True]
@pytest.mark.slow
@pytest.mark.skipif(
importlib.util.find_spec("ollama") is None, reason="Ollama not installed"
)
def test_ollama_embedding(tmp_path):
model = get_registry().get("ollama").create(max_retries=0)
class TextModel(LanceModel):
text: str = model.SourceField()
vector: Vector(model.ndims()) = model.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect(tmp_path)
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == model.ndims()
result = tbl.search("hello").limit(1).to_pandas()
assert result["text"][0] == "hello world"
# Test safe_model_dump
dumped_model = model.safe_model_dump()
assert isinstance(dumped_model, dict)
assert "name" in dumped_model
assert "max_retries" in dumped_model
assert dumped_model["max_retries"] == 0
assert all(not k.startswith("_") for k in dumped_model.keys())
# Test serialization of the dumped model
import json
try:
json.dumps(dumped_model)
except TypeError:
pytest.fail("Failed to JSON serialize the dumped model")
@pytest.mark.slow
@pytest.mark.skipif(
os.environ.get("VOYAGE_API_KEY") is None, reason="VOYAGE_API_KEY not set"
)
def test_voyageai_embedding_function():
voyageai = get_registry().get("voyageai").create(name="voyage-3", max_retries=0)
class TextModel(LanceModel):
text: str = voyageai.SourceField()
vector: Vector(voyageai.ndims()) = voyageai.VectorField()
df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
db = lancedb.connect("~/lancedb")
tbl = db.create_table("test", schema=TextModel, mode="overwrite")
tbl.add(df)
assert len(tbl.to_pandas()["vector"][0]) == voyageai.ndims()