wandb 0.13.11__py3-none-any.whl → 0.14.0__py3-none-any.whl
Sign up to get free protection for your applications and to get access to all the features.
- wandb/__init__.py +1 -1
- wandb/apis/importers/__init__.py +4 -0
- wandb/apis/importers/base.py +312 -0
- wandb/apis/importers/mlflow.py +113 -0
- wandb/apis/internal.py +9 -0
- wandb/apis/public.py +0 -2
- wandb/cli/cli.py +100 -72
- wandb/docker/__init__.py +33 -5
- wandb/proto/v3/wandb_telemetry_pb2.py +10 -10
- wandb/proto/v4/wandb_telemetry_pb2.py +10 -10
- wandb/sdk/internal/internal_api.py +85 -9
- wandb/sdk/launch/_project_spec.py +45 -55
- wandb/sdk/launch/agent/agent.py +80 -18
- wandb/sdk/launch/builder/build.py +16 -74
- wandb/sdk/launch/builder/docker_builder.py +36 -8
- wandb/sdk/launch/builder/kaniko_builder.py +78 -37
- wandb/sdk/launch/builder/templates/_wandb_bootstrap.py +68 -18
- wandb/sdk/launch/environment/aws_environment.py +4 -0
- wandb/sdk/launch/launch.py +1 -6
- wandb/sdk/launch/launch_add.py +0 -5
- wandb/sdk/launch/registry/abstract.py +12 -0
- wandb/sdk/launch/registry/elastic_container_registry.py +31 -1
- wandb/sdk/launch/registry/google_artifact_registry.py +32 -0
- wandb/sdk/launch/registry/local_registry.py +15 -1
- wandb/sdk/launch/runner/abstract.py +0 -14
- wandb/sdk/launch/runner/kubernetes_runner.py +25 -19
- wandb/sdk/launch/runner/local_container.py +7 -8
- wandb/sdk/launch/runner/local_process.py +0 -3
- wandb/sdk/launch/runner/sagemaker_runner.py +0 -3
- wandb/sdk/launch/runner/vertex_runner.py +0 -2
- wandb/sdk/launch/sweeps/scheduler.py +39 -10
- wandb/sdk/launch/utils.py +52 -4
- wandb/sdk/wandb_run.py +3 -10
- wandb/sync/sync.py +1 -0
- wandb/util.py +1 -0
- {wandb-0.13.11.dist-info → wandb-0.14.0.dist-info}/METADATA +1 -1
- {wandb-0.13.11.dist-info → wandb-0.14.0.dist-info}/RECORD +41 -38
- {wandb-0.13.11.dist-info → wandb-0.14.0.dist-info}/WHEEL +1 -1
- {wandb-0.13.11.dist-info → wandb-0.14.0.dist-info}/LICENSE +0 -0
- {wandb-0.13.11.dist-info → wandb-0.14.0.dist-info}/entry_points.txt +0 -0
- {wandb-0.13.11.dist-info → wandb-0.14.0.dist-info}/top_level.txt +0 -0
wandb/__init__.py
CHANGED
@@ -11,7 +11,7 @@ For scripts and interactive notebooks, see https://github.com/wandb/examples.
|
|
11
11
|
|
12
12
|
For reference documentation, see https://docs.wandb.com/ref/python.
|
13
13
|
"""
|
14
|
-
__version__ = "0.
|
14
|
+
__version__ = "0.14.0"
|
15
15
|
|
16
16
|
# Used with pypi checks and other messages related to pip
|
17
17
|
_wandb_module = "wandb"
|
@@ -0,0 +1,312 @@
|
|
1
|
+
import json
|
2
|
+
import platform
|
3
|
+
from abc import ABC, abstractmethod
|
4
|
+
from concurrent.futures import ProcessPoolExecutor, as_completed
|
5
|
+
from contextlib import contextmanager
|
6
|
+
from typing import Any, Dict, Iterable, List, Optional, Tuple
|
7
|
+
|
8
|
+
from tqdm.auto import tqdm
|
9
|
+
|
10
|
+
import wandb
|
11
|
+
from wandb.proto import wandb_internal_pb2 as pb
|
12
|
+
from wandb.proto import wandb_telemetry_pb2 as telem_pb
|
13
|
+
from wandb.sdk.interface.interface import file_policy_to_enum
|
14
|
+
from wandb.sdk.interface.interface_queue import InterfaceQueue
|
15
|
+
from wandb.sdk.internal.sender import SendManager
|
16
|
+
|
17
|
+
Name = str
|
18
|
+
Path = str
|
19
|
+
|
20
|
+
|
21
|
+
def coalesce(*arg: Any) -> Any:
|
22
|
+
"""Return the first non-none value in the list of arguments. Similar to ?? in C#."""
|
23
|
+
return next((a for a in arg if a is not None), None)
|
24
|
+
|
25
|
+
|
26
|
+
@contextmanager
|
27
|
+
def send_manager(root_dir):
|
28
|
+
sm = SendManager.setup(root_dir, resume=False)
|
29
|
+
try:
|
30
|
+
yield sm
|
31
|
+
finally:
|
32
|
+
# flush any remaining records
|
33
|
+
while sm:
|
34
|
+
data = next(sm)
|
35
|
+
sm.send(data)
|
36
|
+
sm.finish()
|
37
|
+
|
38
|
+
|
39
|
+
class ImporterRun:
|
40
|
+
def __init__(self) -> None:
|
41
|
+
self.interface = InterfaceQueue()
|
42
|
+
self.run_dir = f"./wandb-importer/{self.run_id()}"
|
43
|
+
|
44
|
+
def run_id(self) -> str:
|
45
|
+
_id = wandb.util.generate_id()
|
46
|
+
wandb.termwarn(f"`run_id` not specified. Autogenerating id: {_id}")
|
47
|
+
return _id
|
48
|
+
|
49
|
+
def entity(self) -> str:
|
50
|
+
_entity = "unspecified-entity"
|
51
|
+
wandb.termwarn(f"`entity` not specified. Defaulting to: {_entity}")
|
52
|
+
return _entity
|
53
|
+
|
54
|
+
def project(self) -> str:
|
55
|
+
_project = "unspecified-project"
|
56
|
+
wandb.termwarn(f"`project` not specified. Defaulting to: {_project}")
|
57
|
+
return _project
|
58
|
+
|
59
|
+
def config(self) -> Dict[str, Any]:
|
60
|
+
return {}
|
61
|
+
|
62
|
+
def summary(self) -> Dict[str, float]:
|
63
|
+
return {}
|
64
|
+
|
65
|
+
def metrics(self) -> List[Dict[str, float]]:
|
66
|
+
"""Metrics for the run.
|
67
|
+
|
68
|
+
We expect metrics in this shape:
|
69
|
+
|
70
|
+
[
|
71
|
+
{'metric1': 1, 'metric2': 1, '_step': 0},
|
72
|
+
{'metric1': 2, 'metric2': 4, '_step': 1},
|
73
|
+
{'metric1': 3, 'metric2': 9, '_step': 2},
|
74
|
+
...
|
75
|
+
]
|
76
|
+
|
77
|
+
You can also submit metrics in this shape:
|
78
|
+
[
|
79
|
+
{'metric1': 1, '_step': 0},
|
80
|
+
{'metric2': 1, '_step': 0},
|
81
|
+
{'metric1': 2, '_step': 1},
|
82
|
+
{'metric2': 4, '_step': 1},
|
83
|
+
...
|
84
|
+
]
|
85
|
+
"""
|
86
|
+
return []
|
87
|
+
|
88
|
+
def run_group(self) -> Optional[str]:
|
89
|
+
...
|
90
|
+
|
91
|
+
def job_type(self) -> Optional[str]:
|
92
|
+
...
|
93
|
+
|
94
|
+
def display_name(self) -> str:
|
95
|
+
return self.run_id()
|
96
|
+
|
97
|
+
def notes(self) -> Optional[str]:
|
98
|
+
...
|
99
|
+
|
100
|
+
def tags(self) -> Optional[List[str]]:
|
101
|
+
...
|
102
|
+
|
103
|
+
def artifacts(self) -> Optional[Iterable[Tuple[Name, Path]]]:
|
104
|
+
...
|
105
|
+
|
106
|
+
def os_version(self) -> Optional[str]:
|
107
|
+
...
|
108
|
+
|
109
|
+
def python_version(self) -> Optional[str]:
|
110
|
+
...
|
111
|
+
|
112
|
+
def cuda_version(self) -> Optional[str]:
|
113
|
+
...
|
114
|
+
|
115
|
+
def program(self) -> Optional[str]:
|
116
|
+
...
|
117
|
+
|
118
|
+
def host(self) -> Optional[str]:
|
119
|
+
...
|
120
|
+
|
121
|
+
def username(self) -> Optional[str]:
|
122
|
+
...
|
123
|
+
|
124
|
+
def executable(self) -> Optional[str]:
|
125
|
+
...
|
126
|
+
|
127
|
+
def gpus_used(self) -> Optional[str]:
|
128
|
+
...
|
129
|
+
|
130
|
+
def cpus_used(self) -> Optional[int]: # can we get the model?
|
131
|
+
...
|
132
|
+
|
133
|
+
def memory_used(self) -> Optional[int]:
|
134
|
+
...
|
135
|
+
|
136
|
+
def runtime(self) -> Optional[int]:
|
137
|
+
...
|
138
|
+
|
139
|
+
def start_time(self) -> Optional[int]:
|
140
|
+
...
|
141
|
+
|
142
|
+
def _make_run_record(self) -> pb.Record:
|
143
|
+
run = pb.RunRecord()
|
144
|
+
run.run_id = self.run_id()
|
145
|
+
run.entity = self.entity()
|
146
|
+
run.project = self.project()
|
147
|
+
run.display_name = coalesce(self.display_name())
|
148
|
+
run.notes = coalesce(self.notes(), "")
|
149
|
+
run.tags.extend(coalesce(self.tags(), list()))
|
150
|
+
# run.start_time.FromMilliseconds(self.start_time())
|
151
|
+
# run.runtime = self.runtime()
|
152
|
+
run_group = self.run_group()
|
153
|
+
if run_group is not None:
|
154
|
+
run.run_group = run_group
|
155
|
+
self.interface._make_config(
|
156
|
+
data=self.config(),
|
157
|
+
obj=run.config,
|
158
|
+
) # is there a better way?
|
159
|
+
return self.interface._make_record(run=run)
|
160
|
+
|
161
|
+
def _make_summary_record(self) -> pb.Record:
|
162
|
+
d: dict = {
|
163
|
+
**self.summary(),
|
164
|
+
"_runtime": self.runtime(), # quirk of runtime -- it has to be here!
|
165
|
+
# '_timestamp': self.start_time()/1000,
|
166
|
+
}
|
167
|
+
summary = self.interface._make_summary_from_dict(d)
|
168
|
+
return self.interface._make_record(summary=summary)
|
169
|
+
|
170
|
+
def _make_history_records(self) -> Iterable[pb.Record]:
|
171
|
+
for _, metrics in enumerate(self.metrics()):
|
172
|
+
history = pb.HistoryRecord()
|
173
|
+
for k, v in metrics.items():
|
174
|
+
item = history.item.add()
|
175
|
+
item.key = k
|
176
|
+
item.value_json = json.dumps(v)
|
177
|
+
yield self.interface._make_record(history=history)
|
178
|
+
|
179
|
+
def _make_files_record(self, files_dict) -> pb.Record:
|
180
|
+
# when making the metadata file, it captures most things correctly
|
181
|
+
# but notably it doesn't capture the start time!
|
182
|
+
files_record = pb.FilesRecord()
|
183
|
+
for path, policy in files_dict["files"]:
|
184
|
+
f = files_record.files.add()
|
185
|
+
f.path = path
|
186
|
+
f.policy = file_policy_to_enum(policy) # is this always "end"?
|
187
|
+
return self.interface._make_record(files=files_record)
|
188
|
+
|
189
|
+
def _make_metadata_files_record(self) -> pb.Record:
|
190
|
+
self._make_metadata_file(self.run_dir)
|
191
|
+
return self._make_files_record(
|
192
|
+
{"files": [[f"{self.run_dir}/files/wandb-metadata.json", "end"]]}
|
193
|
+
)
|
194
|
+
|
195
|
+
def _make_artifact_record(self) -> pb.Record:
|
196
|
+
art = wandb.Artifact(self.display_name(), "imported-artifacts")
|
197
|
+
artifacts = self.artifacts()
|
198
|
+
if artifacts is not None:
|
199
|
+
for name, path in artifacts:
|
200
|
+
art.add_file(path, name)
|
201
|
+
proto = self.interface._make_artifact(art)
|
202
|
+
proto.run_id = self.run_id()
|
203
|
+
proto.project = self.project()
|
204
|
+
proto.entity = self.entity()
|
205
|
+
proto.user_created = False
|
206
|
+
proto.use_after_commit = False
|
207
|
+
proto.finalize = True
|
208
|
+
for tag in ["latest", "imported"]:
|
209
|
+
proto.aliases.append(tag)
|
210
|
+
return self.interface._make_record(artifact=proto)
|
211
|
+
|
212
|
+
def _make_telem_record(self) -> pb.Record:
|
213
|
+
feature = telem_pb.Feature()
|
214
|
+
feature.importer_mlflow = True
|
215
|
+
|
216
|
+
telem = telem_pb.TelemetryRecord()
|
217
|
+
telem.feature.CopyFrom(feature)
|
218
|
+
telem.python_version = platform.python_version() # importer's python version
|
219
|
+
telem.cli_version = wandb.__version__
|
220
|
+
return self.interface._make_record(telemetry=telem)
|
221
|
+
|
222
|
+
def _make_metadata_file(self, run_dir: str) -> None:
|
223
|
+
missing_text = "MLFlow did not capture this info."
|
224
|
+
|
225
|
+
d = {}
|
226
|
+
if self.os_version() is not None:
|
227
|
+
d["os"] = self.os_version()
|
228
|
+
else:
|
229
|
+
d["os"] = missing_text
|
230
|
+
|
231
|
+
if self.python_version() is not None:
|
232
|
+
d["python"] = self.python_version()
|
233
|
+
else:
|
234
|
+
d["python"] = missing_text
|
235
|
+
|
236
|
+
if self.program() is not None:
|
237
|
+
d["program"] = self.program()
|
238
|
+
else:
|
239
|
+
d["program"] = missing_text
|
240
|
+
|
241
|
+
if self.cuda_version() is not None:
|
242
|
+
d["cuda"] = self.cuda_version()
|
243
|
+
if self.host() is not None:
|
244
|
+
d["host"] = self.host()
|
245
|
+
if self.username() is not None:
|
246
|
+
d["username"] = self.username()
|
247
|
+
if self.executable() is not None:
|
248
|
+
d["executable"] = self.executable()
|
249
|
+
gpus_used = self.gpus_used()
|
250
|
+
if gpus_used is not None:
|
251
|
+
d["gpu_devices"] = json.dumps(gpus_used)
|
252
|
+
d["gpu_count"] = json.dumps(len(gpus_used))
|
253
|
+
cpus_used = self.cpus_used()
|
254
|
+
if cpus_used is not None:
|
255
|
+
d["cpu_count"] = json.dumps(self.cpus_used())
|
256
|
+
mem_used = self.memory_used()
|
257
|
+
if mem_used is not None:
|
258
|
+
d["memory"] = json.dumps({"total": self.memory_used()})
|
259
|
+
|
260
|
+
with open(f"{run_dir}/files/wandb-metadata.json", "w") as f:
|
261
|
+
f.write(json.dumps(d))
|
262
|
+
|
263
|
+
|
264
|
+
class Importer(ABC):
|
265
|
+
@abstractmethod
|
266
|
+
def download_all_runs(self) -> Iterable[ImporterRun]:
|
267
|
+
...
|
268
|
+
|
269
|
+
def import_all(self, overrides: Optional[Dict[str, Any]] = None) -> None:
|
270
|
+
for run in tqdm(self.download_all_runs(), desc="Sending runs"):
|
271
|
+
self.import_one(run, overrides)
|
272
|
+
|
273
|
+
def import_all_parallel(
|
274
|
+
self, overrides: Optional[Dict[str, Any]] = None, **pool_kwargs: Any
|
275
|
+
) -> None:
|
276
|
+
runs = list(self.download_all_runs())
|
277
|
+
with tqdm(total=len(runs)) as pbar:
|
278
|
+
with ProcessPoolExecutor(**pool_kwargs) as exc:
|
279
|
+
futures = {
|
280
|
+
exc.submit(self.import_one, run, overrides=overrides): run
|
281
|
+
for run in runs
|
282
|
+
}
|
283
|
+
for future in as_completed(futures):
|
284
|
+
run = futures[future]
|
285
|
+
pbar.update(1)
|
286
|
+
pbar.set_description(
|
287
|
+
f"Imported Run: {run.run_group()} {run.display_name()}"
|
288
|
+
)
|
289
|
+
|
290
|
+
def import_one(
|
291
|
+
self,
|
292
|
+
run: ImporterRun,
|
293
|
+
overrides: Optional[Dict[str, Any]] = None,
|
294
|
+
) -> None:
|
295
|
+
# does this need to be here for pmap?
|
296
|
+
if overrides:
|
297
|
+
for k, v in overrides.items():
|
298
|
+
# `lambda: v` won't work!
|
299
|
+
# https://stackoverflow.com/questions/10802002/why-deepcopy-doesnt-create-new-references-to-lambda-function
|
300
|
+
setattr(run, k, lambda v=v: v)
|
301
|
+
self._import_one(run)
|
302
|
+
|
303
|
+
def _import_one(self, run: ImporterRun) -> None:
|
304
|
+
with send_manager(run.run_dir) as sm:
|
305
|
+
sm.send(run._make_run_record())
|
306
|
+
sm.send(run._make_summary_record())
|
307
|
+
sm.send(run._make_metadata_files_record())
|
308
|
+
for history_record in run._make_history_records():
|
309
|
+
sm.send(history_record)
|
310
|
+
if run.artifacts() is not None:
|
311
|
+
sm.send(run._make_artifact_record())
|
312
|
+
sm.send(run._make_telem_record())
|
@@ -0,0 +1,113 @@
|
|
1
|
+
from typing import Any, Dict, Iterable, Optional
|
2
|
+
|
3
|
+
from wandb.util import get_module
|
4
|
+
|
5
|
+
from .base import Importer, ImporterRun
|
6
|
+
|
7
|
+
mlflow = get_module(
|
8
|
+
"mlflow",
|
9
|
+
required="To use the MlflowImporter, please install mlflow: `pip install mlflow`",
|
10
|
+
)
|
11
|
+
|
12
|
+
|
13
|
+
class MlflowRun(ImporterRun):
|
14
|
+
def __init__(self, run, mlflow_client):
|
15
|
+
self.run = run
|
16
|
+
self.mlflow_client = mlflow_client
|
17
|
+
super().__init__()
|
18
|
+
|
19
|
+
def run_id(self):
|
20
|
+
return self.run.info.run_id
|
21
|
+
|
22
|
+
def entity(self):
|
23
|
+
return self.run.info.user_id
|
24
|
+
|
25
|
+
def project(self):
|
26
|
+
return "imported-from-mlflow"
|
27
|
+
|
28
|
+
def config(self):
|
29
|
+
return self.run.data.params
|
30
|
+
|
31
|
+
def summary(self):
|
32
|
+
return self.run.data.metrics
|
33
|
+
|
34
|
+
def metrics(self):
|
35
|
+
def wandbify(metrics):
|
36
|
+
for step, t in enumerate(metrics):
|
37
|
+
d = {m.key: m.value for m in t}
|
38
|
+
d["_step"] = step
|
39
|
+
yield d
|
40
|
+
|
41
|
+
metrics = [
|
42
|
+
self.mlflow_client.get_metric_history(self.run.info.run_id, k)
|
43
|
+
for k in self.run.data.metrics.keys()
|
44
|
+
]
|
45
|
+
metrics = zip(*metrics) # transpose
|
46
|
+
return wandbify(metrics)
|
47
|
+
|
48
|
+
# Alternate: Might be slower but use less mem
|
49
|
+
# Can't make this a generator. See mlflow get_metric_history internals
|
50
|
+
# https://github.com/mlflow/mlflow/blob/master/mlflow/tracking/_tracking_service/client.py#L74-L93
|
51
|
+
# for k in self.run.data.metrics.keys():
|
52
|
+
# history = self.mlflow_client.get_metric_history(self.run.info.run_id, k)
|
53
|
+
# yield wandbify(history)
|
54
|
+
|
55
|
+
def run_group(self):
|
56
|
+
# this is nesting? Parent at `run.info.tags.get("mlflow.parentRunId")`
|
57
|
+
return f"Experiment {self.run.info.experiment_id}"
|
58
|
+
|
59
|
+
def job_type(self):
|
60
|
+
# Is this the right approach?
|
61
|
+
return f"User {self.run.info.user_id}"
|
62
|
+
|
63
|
+
def display_name(self):
|
64
|
+
return self.run.info.run_name
|
65
|
+
|
66
|
+
def notes(self):
|
67
|
+
return self.run.data.tags.get("mlflow.note.content")
|
68
|
+
|
69
|
+
def tags(self):
|
70
|
+
return {
|
71
|
+
k: v for k, v in self.run.data.tags.items() if not k.startswith("mlflow.")
|
72
|
+
}
|
73
|
+
|
74
|
+
def start_time(self):
|
75
|
+
return self.run.info.start_time // 1000
|
76
|
+
|
77
|
+
def runtime(self):
|
78
|
+
return self.run.info.end_time // 1_000 - self.start_time()
|
79
|
+
|
80
|
+
def git(self):
|
81
|
+
...
|
82
|
+
|
83
|
+
def artifacts(self):
|
84
|
+
for f in self.mlflow_client.list_artifacts(self.run.info.run_id):
|
85
|
+
dir_path = mlflow.artifacts.download_artifacts(run_id=self.run.info.run_id)
|
86
|
+
full_path = dir_path + f.path
|
87
|
+
yield (f.path, full_path)
|
88
|
+
|
89
|
+
|
90
|
+
class MlflowImporter(Importer):
|
91
|
+
def __init__(
|
92
|
+
self, mlflow_tracking_uri, mlflow_registry_uri=None, wandb_base_url=None
|
93
|
+
) -> None:
|
94
|
+
super().__init__()
|
95
|
+
self.mlflow_tracking_uri = mlflow_tracking_uri
|
96
|
+
|
97
|
+
mlflow.set_tracking_uri(self.mlflow_tracking_uri)
|
98
|
+
if mlflow_registry_uri:
|
99
|
+
mlflow.set_registry_uri(mlflow_registry_uri)
|
100
|
+
self.mlflow_client = mlflow.tracking.MlflowClient(mlflow_tracking_uri)
|
101
|
+
|
102
|
+
def import_one(
|
103
|
+
self,
|
104
|
+
run: ImporterRun,
|
105
|
+
overrides: Optional[Dict[str, Any]] = None,
|
106
|
+
) -> None:
|
107
|
+
mlflow.set_tracking_uri(self.mlflow_tracking_uri)
|
108
|
+
super().import_one(run, overrides)
|
109
|
+
|
110
|
+
def download_all_runs(self) -> Iterable[MlflowRun]:
|
111
|
+
for exp in self.mlflow_client.search_experiments():
|
112
|
+
for run in self.mlflow_client.search_runs(exp.experiment_id):
|
113
|
+
yield MlflowRun(run, self.mlflow_client)
|
wandb/apis/internal.py
CHANGED
@@ -161,6 +161,9 @@ class Api:
|
|
161
161
|
def get_run_state(self, *args, **kwargs):
|
162
162
|
return self.api.get_run_state(*args, **kwargs)
|
163
163
|
|
164
|
+
def entity_is_team(self, *args, **kwargs):
|
165
|
+
return self.api.entity_is_team(*args, **kwargs)
|
166
|
+
|
164
167
|
def get_project_run_queues(self, *args, **kwargs):
|
165
168
|
return self.api.get_project_run_queues(*args, **kwargs)
|
166
169
|
|
@@ -182,6 +185,12 @@ class Api:
|
|
182
185
|
def launch_agent_introspection(self, *args, **kwargs):
|
183
186
|
return self.api.launch_agent_introspection(*args, **kwargs)
|
184
187
|
|
188
|
+
def fail_run_queue_item_introspection(self, *args, **kwargs):
|
189
|
+
return self.api.fail_run_queue_item_introspection(*args, **kwargs)
|
190
|
+
|
191
|
+
def fail_run_queue_item(self, *args, **kwargs):
|
192
|
+
return self.api.fail_run_queue_item(*args, **kwargs)
|
193
|
+
|
185
194
|
def get_launch_agent(self, *args, **kwargs):
|
186
195
|
return self.api.get_launch_agent(*args, **kwargs)
|
187
196
|
|
wandb/apis/public.py
CHANGED
@@ -5562,7 +5562,6 @@ class Job:
|
|
5562
5562
|
queue=None,
|
5563
5563
|
resource="local-container",
|
5564
5564
|
resource_args=None,
|
5565
|
-
cuda=False,
|
5566
5565
|
project_queue=None,
|
5567
5566
|
):
|
5568
5567
|
from wandb.sdk.launch import launch_add
|
@@ -5589,6 +5588,5 @@ class Job:
|
|
5589
5588
|
resource=resource,
|
5590
5589
|
project_queue=project_queue,
|
5591
5590
|
resource_args=resource_args,
|
5592
|
-
cuda=cuda,
|
5593
5591
|
)
|
5594
5592
|
return queued_run
|