-
Notifications
You must be signed in to change notification settings - Fork 1.3k
Expand file tree
/
Copy pathtest_compute.py
More file actions
331 lines (279 loc) · 10.9 KB
/
test_compute.py
File metadata and controls
331 lines (279 loc) · 10.9 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
from datetime import timedelta
from typing import cast
from unittest.mock import MagicMock
import pandas as pd
import pytest
from tqdm import tqdm
from feast import BatchFeatureView, Field
from feast.aggregation import Aggregation
from feast.infra.common.materialization_job import (
MaterializationJobStatus,
MaterializationTask,
)
from feast.infra.common.retrieval_task import HistoricalRetrievalTask
from feast.infra.compute_engines.ray.compute import RayComputeEngine
from feast.infra.compute_engines.ray.config import RayComputeEngineConfig
from feast.infra.compute_engines.ray.job import RayDAGRetrievalJob
from feast.infra.offline_stores.contrib.ray_offline_store.ray import (
RayOfflineStore,
)
from feast.transformation.ray_transformation import RayTransformation
from feast.types import Float32, Int32, Int64
from tests.component.ray.ray_shared_utils import (
driver,
now,
)
@pytest.mark.integration
@pytest.mark.xdist_group(name="ray")
def test_ray_compute_engine_get_historical_features(
ray_environment, feature_dataset, entity_df
):
"""Test Ray compute engine historical feature retrieval."""
fs = ray_environment.feature_store
registry = fs.registry
def transform_feature(df: pd.DataFrame) -> pd.DataFrame:
df["sum_conv_rate"] = df["sum_conv_rate"] * 2
df["avg_acc_rate"] = df["avg_acc_rate"] * 2
return df
driver_stats_fv = BatchFeatureView(
name="driver_hourly_stats",
entities=[driver],
mode="pandas",
aggregations=[
Aggregation(column="conv_rate", function="sum"),
Aggregation(column="acc_rate", function="avg"),
],
udf=transform_feature,
udf_string="transform_feature",
ttl=timedelta(days=3),
schema=[
Field(name="conv_rate", dtype=Float32),
Field(name="acc_rate", dtype=Float32),
Field(name="avg_daily_trips", dtype=Int64),
Field(name="driver_id", dtype=Int32),
],
online=False,
offline=False,
source=feature_dataset,
)
fs.apply([driver, driver_stats_fv])
# Build retrieval task
task = HistoricalRetrievalTask(
project=ray_environment.project,
entity_df=entity_df,
feature_view=driver_stats_fv,
full_feature_name=False,
registry=registry,
)
engine = RayComputeEngine(
repo_config=ray_environment.config,
offline_store=RayOfflineStore(),
online_store=MagicMock(),
)
ray_dag_retrieval_job = engine.get_historical_features(registry, task)
ray_dataset = cast(RayDAGRetrievalJob, ray_dag_retrieval_job).to_ray_dataset()
df_out = ray_dataset.to_pandas().sort_values("driver_id")
assert df_out.driver_id.to_list() == [1001, 1002]
assert abs(df_out["sum_conv_rate"].to_list()[0] - 1.6) < 1e-6
assert abs(df_out["sum_conv_rate"].to_list()[1] - 2.0) < 1e-6
assert abs(df_out["avg_acc_rate"].to_list()[0] - 1.0) < 1e-6
assert abs(df_out["avg_acc_rate"].to_list()[1] - 1.0) < 1e-6
@pytest.mark.integration
@pytest.mark.xdist_group(name="ray")
def test_ray_compute_engine_materialize(ray_environment, feature_dataset):
"""Test Ray compute engine materialization."""
fs = ray_environment.feature_store
registry = fs.registry
def transform_feature(df: pd.DataFrame) -> pd.DataFrame:
df["sum_conv_rate"] = df["sum_conv_rate"] * 2
df["avg_acc_rate"] = df["avg_acc_rate"] * 2
return df
driver_stats_fv = BatchFeatureView(
name="driver_hourly_stats",
entities=[driver],
mode="pandas",
aggregations=[
Aggregation(column="conv_rate", function="sum"),
Aggregation(column="acc_rate", function="avg"),
],
udf=transform_feature,
udf_string="transform_feature",
ttl=timedelta(days=3),
schema=[
Field(name="conv_rate", dtype=Float32),
Field(name="acc_rate", dtype=Float32),
Field(name="avg_daily_trips", dtype=Int64),
Field(name="driver_id", dtype=Int32),
],
online=True,
offline=False,
source=feature_dataset,
)
def tqdm_builder(length):
return tqdm(length, ncols=100)
fs.apply([driver, driver_stats_fv])
task = MaterializationTask(
project=ray_environment.project,
feature_view=driver_stats_fv,
start_time=now - timedelta(days=2),
end_time=now,
tqdm_builder=tqdm_builder,
)
engine = RayComputeEngine(
repo_config=ray_environment.config,
offline_store=RayOfflineStore(),
online_store=MagicMock(),
)
ray_materialize_jobs = engine.materialize(registry, task)
assert len(ray_materialize_jobs) == 1
assert ray_materialize_jobs[0].status() == MaterializationJobStatus.SUCCEEDED
@pytest.mark.integration
@pytest.mark.xdist_group(name="ray")
def test_ray_compute_engine_config():
"""Test Ray compute engine configuration."""
config = RayComputeEngineConfig(
type="ray.engine",
ray_address="ray://localhost:10001",
broadcast_join_threshold_mb=200,
enable_distributed_joins=True,
max_parallelism_multiplier=4,
target_partition_size_mb=128,
window_size_for_joins="2H",
max_workers=4,
enable_optimization=True,
)
assert config.type == "ray.engine"
assert config.ray_address == "ray://localhost:10001"
assert config.broadcast_join_threshold_mb == 200
assert config.window_size_timedelta == timedelta(hours=2)
@pytest.mark.integration
@pytest.mark.xdist_group(name="ray")
def test_ray_transformation_compute_engine(ray_environment, feature_dataset, entity_df):
"""Test Ray compute engine with Ray transformation mode."""
import ray.data
fs = ray_environment.feature_store
registry = fs.registry
def ray_transformation_udf(ds: ray.data.Dataset) -> ray.data.Dataset:
"""Ray native transformation that processes data in parallel."""
def process_batch(batch: pd.DataFrame) -> pd.DataFrame:
# Simulate some computation (e.g., feature engineering)
if "conv_rate" in batch.columns:
batch["processed_conv_rate"] = batch["conv_rate"] * 2.0
if "acc_rate" in batch.columns:
batch["processed_acc_rate"] = batch["acc_rate"] * 1.5
return batch
return ds.map_batches(
process_batch,
batch_format="pandas",
concurrency=2, # Use 2 parallel workers
)
# Create Ray transformation
ray_transform = RayTransformation(
udf=ray_transformation_udf,
udf_string="def ray_transformation_udf(ds): return ds.map_batches(...)",
)
driver_stats_fv = BatchFeatureView(
name="driver_hourly_stats_ray",
entities=[driver],
mode="ray", # Use Ray transformation mode
feature_transformation=ray_transform,
ttl=timedelta(days=3),
schema=[
Field(name="conv_rate", dtype=Float32),
Field(name="acc_rate", dtype=Float32),
Field(name="processed_conv_rate", dtype=Float32),
Field(name="processed_acc_rate", dtype=Float32),
Field(name="avg_daily_trips", dtype=Int64),
Field(name="driver_id", dtype=Int32),
],
online=False,
offline=False,
source=feature_dataset,
)
fs.apply([driver, driver_stats_fv])
# Build retrieval task
task = HistoricalRetrievalTask(
project=ray_environment.project,
entity_df=entity_df,
feature_view=driver_stats_fv,
full_feature_name=False,
registry=registry,
)
engine = RayComputeEngine(
repo_config=ray_environment.config,
offline_store=RayOfflineStore(),
online_store=MagicMock(),
)
ray_dag_retrieval_job = engine.get_historical_features(registry, task)
ray_dataset = cast(RayDAGRetrievalJob, ray_dag_retrieval_job).to_ray_dataset()
df_out = ray_dataset.to_pandas().sort_values("driver_id")
# Verify the transformation was applied
assert df_out.driver_id.to_list() == [1001, 1002]
# Check that original columns are present
assert "conv_rate" in df_out.columns
assert "acc_rate" in df_out.columns
# Check that transformed columns are present
assert "processed_conv_rate" in df_out.columns
assert "processed_acc_rate" in df_out.columns
# Verify the transformation logic was applied
for idx, row in df_out.iterrows():
assert abs(row["processed_conv_rate"] - row["conv_rate"] * 2.0) < 1e-6
assert abs(row["processed_acc_rate"] - row["acc_rate"] * 1.5) < 1e-6
@pytest.mark.integration
@pytest.mark.xdist_group(name="ray")
def test_ray_transformation_materialization(ray_environment, feature_dataset):
"""Test Ray transformation during materialization."""
import ray.data
fs = ray_environment.feature_store
registry = fs.registry
def ray_embedding_udf(ds: ray.data.Dataset) -> ray.data.Dataset:
"""Simulate embedding generation with Ray native processing."""
def generate_embeddings(batch: pd.DataFrame) -> pd.DataFrame:
# Simulate embedding generation
if "conv_rate" in batch.columns:
# Create a simple embedding based on conv_rate
batch["embedding"] = batch["conv_rate"].apply(
lambda x: [x * 0.1, x * 0.2, x * 0.3]
)
return batch
return ds.map_batches(generate_embeddings, batch_format="pandas", concurrency=2)
# Create Ray transformation for embeddings
ray_embedding_transform = RayTransformation(
udf=ray_embedding_udf,
udf_string="def ray_embedding_udf(ds): return ds.map_batches(...)",
)
driver_embeddings_fv = BatchFeatureView(
name="driver_embeddings",
entities=[driver],
mode="ray",
feature_transformation=ray_embedding_transform,
ttl=timedelta(days=3),
schema=[
Field(name="conv_rate", dtype=Float32),
Field(
name="embedding", dtype=Float32
), # This would be Array(Float32) in real usage
Field(name="driver_id", dtype=Int32),
],
online=True,
offline=False,
source=feature_dataset,
)
def tqdm_builder(length):
return tqdm(length, ncols=100)
fs.apply([driver, driver_embeddings_fv])
task = MaterializationTask(
project=ray_environment.project,
feature_view=driver_embeddings_fv,
start_time=now - timedelta(days=2),
end_time=now,
tqdm_builder=tqdm_builder,
)
engine = RayComputeEngine(
repo_config=ray_environment.config,
offline_store=RayOfflineStore(),
online_store=MagicMock(),
)
ray_materialize_jobs = engine.materialize(registry, task)
assert len(ray_materialize_jobs) == 1
assert ray_materialize_jobs[0].status() == MaterializationJobStatus.SUCCEEDED