envloop-cli 0.4.0__tar.gz → 1.4.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/.gitignore +3 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/PKG-INFO +1 -1
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/agent_runs.py +36 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/cli.py +43 -8
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/cli_errors.py +5 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/client.py +64 -9
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/http.py +5 -17
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/profile_updates.py +23 -3
- envloop_cli-1.4.0/envloop/rsi_cli.py +49 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/pyproject.toml +1 -1
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/README.pypi.md +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/__init__.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/auth_cli.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/auth_login.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/auth_workspace.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/cli_output.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/config.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/job_logs.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/lifecycle_cli.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/navigation_resources.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/profile_secrets.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/resource_cli.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/resources.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/sandbox.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/train.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/train_log_cli.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/training/__init__.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/training/credentials.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/training/sampling_client.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/training/service_client.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/utils/__init__.py +0 -0
- {envloop_cli-0.4.0 → envloop_cli-1.4.0}/envloop/utils/logging.py +0 -0
|
@@ -7,6 +7,7 @@ using the same rules as the Console's agent runs page.
|
|
|
7
7
|
"""
|
|
8
8
|
|
|
9
9
|
from datetime import datetime
|
|
10
|
+
from urllib.parse import quote, urlencode
|
|
10
11
|
|
|
11
12
|
from .train import TrainJobs
|
|
12
13
|
|
|
@@ -206,3 +207,38 @@ class AgentRuns:
|
|
|
206
207
|
if run not in runs:
|
|
207
208
|
raise AgentRunError("AGENT_RUN_NOT_FOUND")
|
|
208
209
|
return {**_detail(run, runs[run]), "observed_at": observed_at}
|
|
210
|
+
|
|
211
|
+
# The RSI loop service behind the platform (``/v1/agent-runs/rsi``). Starting spends money, so
|
|
212
|
+
# ``start`` needs a key with ``training.run``; ``control(stop)`` also needs ``training.stop``.
|
|
213
|
+
def _rsi(self, method: str, path: str, body=None) -> dict:
|
|
214
|
+
return self._train_jobs.transport.request(method, "/v1/agent-runs/rsi" + path, body)
|
|
215
|
+
|
|
216
|
+
def start(self, run: str, brief: str, *, model_usd_cap: float, usd_cap: float | None = None,
|
|
217
|
+
model_request_cap: int | None = None, domain: str | None = None, rounds: int | None = None,
|
|
218
|
+
dry_run: bool = False, student_model: str | None = None) -> dict:
|
|
219
|
+
"""Start a new RSI run; the platform validates every field. ``dry_run`` uses no GPU."""
|
|
220
|
+
limits = {"model_usd_cap": model_usd_cap}
|
|
221
|
+
for key, value in (("usd_cap", usd_cap), ("model_request_cap", model_request_cap)):
|
|
222
|
+
if value is not None:
|
|
223
|
+
limits[key] = value
|
|
224
|
+
body = {"run": run, "brief": brief, "limits": limits}
|
|
225
|
+
for key, value in (("domain", domain), ("rounds", rounds), ("student_model", student_model)):
|
|
226
|
+
if value is not None:
|
|
227
|
+
body[key] = value
|
|
228
|
+
if dry_run:
|
|
229
|
+
body["dry_run"] = True
|
|
230
|
+
return self._rsi("POST", "/runs", body)
|
|
231
|
+
|
|
232
|
+
def live(self, run: str, since: int | None = None) -> dict:
|
|
233
|
+
"""What a run does now; pass the previous ``next_cursor`` as ``since`` for newer events."""
|
|
234
|
+
query = "?" + urlencode({"since": since}) if since is not None else ""
|
|
235
|
+
return self._rsi("GET", f"/runs/{quote(run, safe='')}/live{query}")
|
|
236
|
+
|
|
237
|
+
def steer(self, run: str, text: str) -> dict:
|
|
238
|
+
"""Queue a message for the running operator (1 to 4000 characters)."""
|
|
239
|
+
return self._rsi("POST", f"/runs/{quote(run, safe='')}/steer", {"text": text})
|
|
240
|
+
|
|
241
|
+
def control(self, run: str, action: str, rounds: int | None = None) -> dict:
|
|
242
|
+
"""``extend`` (needs ``rounds``), ``stop`` (after the current round) or ``resume``."""
|
|
243
|
+
return self._rsi("POST", f"/runs/{quote(run, safe='')}/control",
|
|
244
|
+
{"action": action, **({} if rounds is None else {"rounds": rounds})})
|
|
@@ -208,6 +208,27 @@ def parser():
|
|
|
208
208
|
runs_get = runs.add_parser("get", help="Show a run's rounds, Train Jobs and checkpoints")
|
|
209
209
|
runs_get.add_argument("run", metavar="RUN", help="run label: user_metadata.rsi_run or agent_run")
|
|
210
210
|
runs_get.add_argument("--json", action="store_true", help="Output the complete run as JSON (default: tables)")
|
|
211
|
+
runs_start = runs.add_parser("start", help="Start a new RSI loop run (paid; needs training.run)")
|
|
212
|
+
runs_start.add_argument("run", metavar="RUN", help="new run id: lower-case letters, digits and dashes")
|
|
213
|
+
brief = runs_start.add_mutually_exclusive_group(required=True)
|
|
214
|
+
brief.add_argument("--brief", help="the goal of the run (up to 4000 characters)")
|
|
215
|
+
brief.add_argument("--brief-file", metavar="PATH", help="read the goal from a file, or - for standard input")
|
|
216
|
+
runs_start.add_argument("--model-usd-cap", type=float, required=True, help="strong-model spend cap in USD (the hard stop)")
|
|
217
|
+
runs_start.add_argument("--usd-cap", type=float, help="cap in USD for model plus GPU training together")
|
|
218
|
+
runs_start.add_argument("--model-request-cap", type=int, help="cap on model requests")
|
|
219
|
+
runs_start.add_argument("--domain", help="task family (default: the service's default; an unknown value is rejected)")
|
|
220
|
+
runs_start.add_argument("--rounds", type=int, help="target rounds (default: the service's default)")
|
|
221
|
+
runs_start.add_argument("--dry-run", action="store_true", help="GPU-free run: trains nothing; exercises the loop, spends model budget only")
|
|
222
|
+
runs_start.add_argument("--student-model", help="gateway student model of a dry run")
|
|
223
|
+
runs_live = runs.add_parser("live", help="Show what a run does now; --follow streams new events")
|
|
224
|
+
runs_live.add_argument("run", metavar="RUN")
|
|
225
|
+
runs_live.add_argument("--since", type=int, help="only events after this cursor (the previous next_cursor)")
|
|
226
|
+
runs_live.add_argument("--follow", action="store_true", help="poll and print new events until the run ends; Ctrl-C exits without stopping it")
|
|
227
|
+
runs_steer = runs.add_parser("steer", help="Message the running operator now (needs training.run)")
|
|
228
|
+
runs_steer.add_argument("run", metavar="RUN")
|
|
229
|
+
runs_steer.add_argument("text", metavar="TEXT", help="1 to 4000 characters")
|
|
230
|
+
runs_stop = runs.add_parser("stop", help="Stop a run after its current round (needs training.run and training.stop)")
|
|
231
|
+
runs_stop.add_argument("run", metavar="RUN")
|
|
211
232
|
tasks = groups.add_parser("trail", help="Inspect and control existing trials").add_subparsers(dest="action", required=True)
|
|
212
233
|
listing = tasks.add_parser("list", help="List Agent, Sandbox, Train and Inference worker Trials")
|
|
213
234
|
listing.add_argument("--type", choices=("agent", "sandbox", "train", "inference"), help="Filter by Trial type")
|
|
@@ -230,11 +251,11 @@ def parser():
|
|
|
230
251
|
download.add_argument("trial_id")
|
|
231
252
|
download.add_argument("directory", type=Path)
|
|
232
253
|
profiles = groups.add_parser("profile", help="Manage platform runtime profiles").add_subparsers(dest="action", required=True)
|
|
233
|
-
put = profiles.add_parser("create", help="Create an immutable Profile version from JSON")
|
|
254
|
+
put = profiles.add_parser("create", help="Create an immutable Profile version from TOML or JSON")
|
|
234
255
|
put.add_argument("name")
|
|
235
256
|
put.add_argument("version")
|
|
236
|
-
put.add_argument("--file", type=Path, required=True, help="
|
|
237
|
-
put.add_argument("--runner", metavar="RUNNER", help="Bind
|
|
257
|
+
put.add_argument("--file", type=Path, required=True, help="Profile TOML or JSON file")
|
|
258
|
+
put.add_argument("--runner", metavar="RUNNER", help="Bind a Runner revision to this version")
|
|
238
259
|
secret_input = put.add_mutually_exclusive_group()
|
|
239
260
|
secret_input.add_argument("--prompt-model-api-key", action="store_true", help="Read the model API key from a non-echoing prompt")
|
|
240
261
|
secret_input.add_argument("--model-api-key-stdin", action="store_true", help="Read the model API key from one line of standard input")
|
|
@@ -242,11 +263,16 @@ def parser():
|
|
|
242
263
|
update.add_argument("name")
|
|
243
264
|
update.add_argument("version", help="New version; existing versions cannot be overwritten")
|
|
244
265
|
update.add_argument("--from-version", help="Source version; omit to use the explicit default")
|
|
245
|
-
update.add_argument("--file", type=Path, required=True, help="JSON patch: merge
|
|
246
|
-
update.add_argument("--runner", metavar="RUNNER", help="Bind a Runner
|
|
266
|
+
update.add_argument("--file", type=Path, required=True, help="TOML or JSON patch: merge params keys, replace supplied metadata; preserve omitted fields")
|
|
267
|
+
update.add_argument("--runner", metavar="RUNNER", help="Bind a Runner revision; v1 requires a full commit; omit to preserve the source binding")
|
|
247
268
|
update_secret = update.add_mutually_exclusive_group()
|
|
248
269
|
update_secret.add_argument("--prompt-model-api-key", action="store_true", help="Read an encrypted model connection key from a non-echoing prompt")
|
|
249
270
|
update_secret.add_argument("--model-api-key-stdin", action="store_true", help="Read an encrypted model connection key from one line of standard input")
|
|
271
|
+
fork = profiles.add_parser('fork', help='Fork an existing Profile into a new immutable identity')
|
|
272
|
+
fork.add_argument('source_name')
|
|
273
|
+
fork.add_argument('source_version')
|
|
274
|
+
fork.add_argument('name')
|
|
275
|
+
fork.add_argument('version')
|
|
250
276
|
get = profiles.add_parser("get")
|
|
251
277
|
get.add_argument("name")
|
|
252
278
|
get.add_argument("--version")
|
|
@@ -406,7 +432,8 @@ def main(argv=None):
|
|
|
406
432
|
args.runner = runner_selector(args.runner)
|
|
407
433
|
except ValueError as exc:
|
|
408
434
|
cli.error(str(exc))
|
|
409
|
-
|
|
435
|
+
selections = sum(value is not None for value in (args.dataset, args.dataset_option, args.source))
|
|
436
|
+
if selections != 1 and not (selections == 0 and args.profile):
|
|
410
437
|
cli.error("provide exactly one positional dataset, --dataset or --source")
|
|
411
438
|
args.dataset = args.dataset if args.dataset is not None else args.dataset_option
|
|
412
439
|
if args.source is not None and any(getattr(args, key) is not None for key in ("runner", "task_id", "config", "split", "revision")):
|
|
@@ -430,7 +457,7 @@ def main(argv=None):
|
|
|
430
457
|
|
|
431
458
|
raw = args.file.read_bytes()
|
|
432
459
|
patch = read_profile_patch(raw)
|
|
433
|
-
if args.runner is not None:
|
|
460
|
+
if args.action == 'update' and args.runner is not None:
|
|
434
461
|
patch = with_runner(patch, args.runner)
|
|
435
462
|
raw = json.dumps(patch).encode('utf-8')
|
|
436
463
|
secrets = None
|
|
@@ -450,8 +477,10 @@ def main(argv=None):
|
|
|
450
477
|
patch=patch, from_version=args.from_version, secrets=secrets)
|
|
451
478
|
else:
|
|
452
479
|
result = client.profiles.import_spec(
|
|
453
|
-
raw, name=args.name, version=args.version, secrets=secrets
|
|
480
|
+
raw, name=args.name, version=args.version, secrets=secrets, runner=args.runner
|
|
454
481
|
)
|
|
482
|
+
elif args.action == 'fork':
|
|
483
|
+
result = client.profiles.fork(args.source_name, args.source_version, name=args.name, version=args.version)
|
|
455
484
|
elif args.action == "get":
|
|
456
485
|
result = client.profiles.get(args.name, args.version)
|
|
457
486
|
elif args.action == "list":
|
|
@@ -495,6 +524,12 @@ def main(argv=None):
|
|
|
495
524
|
from .train_log_cli import read_logs
|
|
496
525
|
|
|
497
526
|
return read_logs(client.jobs, args)
|
|
527
|
+
elif args.action == "run" and args.run_action in {"start", "live", "steer", "stop"}:
|
|
528
|
+
from .rsi_cli import run_rsi
|
|
529
|
+
|
|
530
|
+
result = run_rsi(client.agent_runs, args)
|
|
531
|
+
if result is None:
|
|
532
|
+
return 0
|
|
498
533
|
elif args.action == "run":
|
|
499
534
|
result = client.agent_runs.list() if args.run_action == "list" else client.agent_runs.get(args.run)
|
|
500
535
|
elif args.action == "list":
|
|
@@ -82,6 +82,11 @@ _register("RUNNER_NOT_FOUND", "The Runner reference was not found or is not read
|
|
|
82
82
|
_register("INVALID_SANDBOX_RUNNER", "The Runner revision cannot execute as a standalone Sandbox Job.", "Check '{program} runner list' and select a frozen Runner revision whose environment supports execution without a Dataset.")
|
|
83
83
|
_register("INVALID_SANDBOX_IMAGE", "The Sandbox image reference is invalid.", "Use a Docker image reference name[:tag][@sha256:digest] and check '{command} --help'.")
|
|
84
84
|
|
|
85
|
+
_register("RSI_RUN_NOT_FOUND", "No RSI run with this name is visible in the current project.", "Check the name with '{program} job run list' and the scope with '{program} auth config show'; a run you just started appears once the service has accepted it.")
|
|
86
|
+
_register("INVALID_RSI_REQUEST INVALID_STEER INVALID_DIRECTIVE", "The RSI service rejected the request fields.", "Check '{command} --help' and https://docs.envloop.ai/agent-docs/en/guides/agent-runs.md for the run id format, the allowed domain and rounds, and the required spend cap.")
|
|
87
|
+
_register("RSI_CONFLICT RSI_RUN_NOT_ACTIVE", "The run id exists, the run is not in a state that allows this action, or no loop machine is free.", "Pick a new run id for a start; read the run with '{program} job run live RUN' before steering or stopping it.")
|
|
88
|
+
_register("RSI_CONTROL_UNCONFIGURED RSI_CONTROL_UNAVAILABLE RSI_CONTROL_TIMEOUT", "The RSI loop service is not available through this platform right now.", _RETRY_HINT)
|
|
89
|
+
|
|
85
90
|
_register("JOB_ACTION_UNSUPPORTED", "This Job type does not support the requested action.", "Check the Job type with '{program} job get JOB_ID' and use its supported actions.")
|
|
86
91
|
_register("INFERENCE_ROUTE_RETIRED", "The global inference route has been retired.", "Use the owning Job's base URL: API_BASE/v1/job/JOB_ID/Inference, and a model deployed by that Job.")
|
|
87
92
|
|
|
@@ -343,16 +343,38 @@ class Profiles:
|
|
|
343
343
|
|
|
344
344
|
Omitted fields and Secret markers are preserved. Supplied metadata,
|
|
345
345
|
including secret_params, replaces that field. Parameter values replace
|
|
346
|
-
whole values, including objects
|
|
346
|
+
whole values, including objects. V1 keys are literal; v2 keys are case
|
|
347
|
+
insensitive and live in spec.params. The default is unchanged.
|
|
347
348
|
"""
|
|
348
|
-
from .profile_updates import checked_patch
|
|
349
|
+
from .profile_updates import checked_patch, merge_v2_params
|
|
349
350
|
|
|
350
351
|
patch = checked_patch(patch, name, version)
|
|
351
352
|
original = self.get(name, from_version)["spec"]
|
|
352
|
-
spec = {**original, **patch, "name": name, "version": version
|
|
353
|
-
|
|
353
|
+
spec = {**original, **patch, "name": name, "version": version}
|
|
354
|
+
if original.get('schema_version') == 'envplatform.profile/v2':
|
|
355
|
+
if 'params' in patch:
|
|
356
|
+
from .http import RemoteError
|
|
357
|
+
raise RemoteError('INVALID_PROFILE', 422)
|
|
358
|
+
before, changes = original.get('spec', {}), patch.get('spec', {})
|
|
359
|
+
spec['spec'] = {**before, **changes, 'params': merge_v2_params(before.get('params', {}), changes.get('params', {}), spec.get('secret_params', []))}
|
|
360
|
+
else:
|
|
361
|
+
spec['params'] = {**original.get('params', {}), **patch.get('params', {})}
|
|
354
362
|
return self.put(spec, secrets=secrets)
|
|
355
363
|
|
|
364
|
+
def fork(self, source_name, source_version, *, name, version, patch=None, secrets=None):
|
|
365
|
+
"""Publish a new identity from an authorized source; never change its default."""
|
|
366
|
+
from .profile_updates import checked_patch, merge_v2_params
|
|
367
|
+
|
|
368
|
+
original = self.get(source_name, source_version)['spec']
|
|
369
|
+
changes = checked_patch({} if patch is None else patch, name, version)
|
|
370
|
+
result = {**original, **changes, 'name': name, 'version': version}
|
|
371
|
+
if original.get('schema_version') == 'envplatform.profile/v2':
|
|
372
|
+
before, updates = original.get('spec', {}), changes.get('spec', {})
|
|
373
|
+
result['spec'] = {**before, **updates, 'params': merge_v2_params(before.get('params', {}), updates.get('params', {}), result.get('secret_params', []))}
|
|
374
|
+
else:
|
|
375
|
+
result['params'] = {**original.get('params', {}), **changes.get('params', {})}
|
|
376
|
+
return self.put(result, secrets=secrets)
|
|
377
|
+
|
|
356
378
|
def get(self, name: str, version: str | None = None) -> dict:
|
|
357
379
|
return self._server.get_profile(name, version)
|
|
358
380
|
|
|
@@ -377,7 +399,31 @@ class Profiles:
|
|
|
377
399
|
"""Revoke model access for a scoped immutable secret revision."""
|
|
378
400
|
return self._server.revoke_profile_secret(name, version)
|
|
379
401
|
|
|
380
|
-
def import_spec(self, raw: bytes, *, name=None, version=None, secrets=None):
|
|
402
|
+
def import_spec(self, raw: bytes, *, name=None, version=None, secrets=None, runner=None):
|
|
403
|
+
import json
|
|
404
|
+
|
|
405
|
+
from .profile_updates import read_profile_patch, with_runner
|
|
406
|
+
|
|
407
|
+
value = read_profile_patch(raw)
|
|
408
|
+
if isinstance(value, dict) and value.get('schema_version') == 'envhub.runner/v1':
|
|
409
|
+
from .http import RemoteError
|
|
410
|
+
|
|
411
|
+
definition = value.get('runner')
|
|
412
|
+
declarations = value.get('params', {})
|
|
413
|
+
if (not isinstance(definition, dict) or not isinstance(definition.get('name'), str)
|
|
414
|
+
or not definition['name'] or not isinstance(declarations, dict)
|
|
415
|
+
or any(not isinstance(item, dict) for item in declarations.values())):
|
|
416
|
+
raise RemoteError('INVALID_PROFILE', 422)
|
|
417
|
+
if runner is not None and runner.partition('@')[0] != definition['name']:
|
|
418
|
+
raise RemoteError('INVALID_PROFILE_RUNNER', 422)
|
|
419
|
+
defaults = {key: declaration['default'] for key, declaration in declarations.items()
|
|
420
|
+
if 'default' in declaration}
|
|
421
|
+
value = {'schema_version': 'envplatform.profile/v2', 'name': name, 'version': version,
|
|
422
|
+
'runner': runner if runner is not None else definition['name'] + '@main',
|
|
423
|
+
'spec': {'params': defaults, **({'env': value['env']} if 'env' in value else {})}}
|
|
424
|
+
if runner is not None:
|
|
425
|
+
value = with_runner(value, runner)
|
|
426
|
+
raw = json.dumps(value, allow_nan=False, ensure_ascii=False).encode('utf-8')
|
|
381
427
|
options = {} if secrets is None else {"secrets": secrets}
|
|
382
428
|
return self._server.import_profile(raw, name=name, version=version, **options)
|
|
383
429
|
|
|
@@ -454,9 +500,10 @@ class Jobs:
|
|
|
454
500
|
body['image'] = validate_image(image)
|
|
455
501
|
validate_public_execution(body)
|
|
456
502
|
return self._server.submit_job(body)
|
|
457
|
-
|
|
503
|
+
profile_only = all(value is None for value in (dataset, source, task_ref)) and (profile_id is not None or options.get("profile_name") is not None)
|
|
504
|
+
if sum(value is not None for value in (dataset, source, task_ref)) != 1 and not profile_only:
|
|
458
505
|
raise ValueError("provide exactly one dataset, source or task_ref")
|
|
459
|
-
if dataset is None and (runner is not None or any(value is not None for value in (task_id, namespace, config, split, revision))):
|
|
506
|
+
if dataset is None and not profile_only and (runner is not None or any(value is not None for value in (task_id, namespace, config, split, revision))):
|
|
460
507
|
raise ValueError("dataset selection options require a dataset")
|
|
461
508
|
if runner is not None:
|
|
462
509
|
runner = runner_selector(runner)
|
|
@@ -464,7 +511,9 @@ class Jobs:
|
|
|
464
511
|
options["command"] = validate_command(command)
|
|
465
512
|
selection = {key: value for key, value in dict(task_id=task_id, namespace=namespace, config=config,
|
|
466
513
|
split=split, revision=revision).items() if value is not None}
|
|
467
|
-
if
|
|
514
|
+
if profile_only:
|
|
515
|
+
source = {"dataset": selection} if selection else {}
|
|
516
|
+
elif source is not None:
|
|
468
517
|
source = {"task": load_task(source)}
|
|
469
518
|
elif task_ref is not None:
|
|
470
519
|
source = {"task_ref": dict(task_ref)}
|
|
@@ -535,7 +584,12 @@ class Jobs:
|
|
|
535
584
|
def stop(self, job_id):
|
|
536
585
|
return self._server.stop_job(job_id)
|
|
537
586
|
|
|
538
|
-
def wait(self, job_id, *, timeout=600, poll_interval=.1):
|
|
587
|
+
def wait(self, job_id, *, timeout=600, poll_interval=.1, max_poll_interval=5):
|
|
588
|
+
"""Poll until terminal; the interval grows 1.5x per poll up to ``max_poll_interval``.
|
|
589
|
+
|
|
590
|
+
Jobs that run for tens of minutes (large-model training) would otherwise
|
|
591
|
+
issue ten status requests a second for the whole wait.
|
|
592
|
+
"""
|
|
539
593
|
if not math.isfinite(timeout) or timeout < 0:
|
|
540
594
|
raise ValueError("timeout must be finite and non-negative")
|
|
541
595
|
if not math.isfinite(poll_interval) or poll_interval <= 0:
|
|
@@ -549,6 +603,7 @@ class Jobs:
|
|
|
549
603
|
if remaining <= 0:
|
|
550
604
|
raise TimeoutError("job did not reach a terminal state")
|
|
551
605
|
time.sleep(min(poll_interval, remaining))
|
|
606
|
+
poll_interval = max(poll_interval, min(poll_interval * 1.5, max_poll_interval))
|
|
552
607
|
|
|
553
608
|
|
|
554
609
|
class Client:
|
|
@@ -307,22 +307,10 @@ class HttpService:
|
|
|
307
307
|
sort_keys=True, separators=(",", ":"), allow_nan=False).encode("utf-8")
|
|
308
308
|
|
|
309
309
|
def import_profile(self, raw, *, name=None, version=None, secrets=None):
|
|
310
|
-
|
|
311
|
-
result = {}
|
|
312
|
-
for key, value in pairs:
|
|
313
|
-
if key in result:
|
|
314
|
-
raise ValueError("duplicate key")
|
|
315
|
-
result[key] = value
|
|
316
|
-
return result
|
|
310
|
+
from .profile_updates import read_profile_patch
|
|
317
311
|
|
|
318
|
-
|
|
319
|
-
|
|
320
|
-
|
|
321
|
-
|
|
322
|
-
spec = json.loads(raw, object_pairs_hook=unique_object, parse_constant=invalid_constant)
|
|
323
|
-
if (not isinstance(spec, dict) or (name is not None and spec.get("name") != name)
|
|
324
|
-
or (version is not None and spec.get("version") != version)):
|
|
325
|
-
raise ValueError("Profile name/version mismatch")
|
|
326
|
-
except (ValueError, UnicodeError, TypeError):
|
|
327
|
-
raise RemoteError("INVALID_PROFILE", 422) from None
|
|
312
|
+
spec = read_profile_patch(raw)
|
|
313
|
+
if (not isinstance(spec, dict) or (name is not None and spec.get("name") != name)
|
|
314
|
+
or (version is not None and spec.get("version") != version)):
|
|
315
|
+
raise RemoteError("INVALID_PROFILE", 422)
|
|
328
316
|
return self.put_profile(profile_payload(spec, secrets))
|
|
@@ -1,5 +1,6 @@
|
|
|
1
1
|
"""Build new Profile versions without changing the source or its default."""
|
|
2
2
|
import json
|
|
3
|
+
import tomllib
|
|
3
4
|
from collections.abc import Mapping
|
|
4
5
|
|
|
5
6
|
from .http import RemoteError
|
|
@@ -18,13 +19,19 @@ def read_profile_patch(raw):
|
|
|
18
19
|
raise ValueError
|
|
19
20
|
|
|
20
21
|
try:
|
|
21
|
-
|
|
22
|
+
text = raw.decode('utf-8') if isinstance(raw, bytes) else raw
|
|
23
|
+
try:
|
|
24
|
+
value = json.loads(text, object_pairs_hook=unique, parse_constant=invalid_constant)
|
|
25
|
+
except ValueError:
|
|
26
|
+
value = tomllib.loads(text)
|
|
27
|
+
json.dumps(value, allow_nan=False)
|
|
28
|
+
return value
|
|
22
29
|
except (ValueError, UnicodeError, TypeError):
|
|
23
30
|
raise RemoteError("INVALID_PROFILE", 422) from None
|
|
24
31
|
|
|
25
32
|
|
|
26
33
|
def checked_patch(patch, name, version):
|
|
27
|
-
allowed = {"schema_version", "name", "version", "params", "description", "created_for", "secret_params", "runner"}
|
|
34
|
+
allowed = {"schema_version", "name", "version", "params", "description", "created_for", "secret_params", "agent", "runner", "dataset", "dataset_mode", "task", "tags", "spec"}
|
|
28
35
|
|
|
29
36
|
def string_keys(value):
|
|
30
37
|
if isinstance(value, Mapping):
|
|
@@ -39,7 +46,8 @@ def checked_patch(patch, name, version):
|
|
|
39
46
|
try:
|
|
40
47
|
if (not isinstance(patch, Mapping) or set(patch) - allowed
|
|
41
48
|
or patch.get("name", name) != name or patch.get("version", version) != version
|
|
42
|
-
or not isinstance(patch.get("params", {}), Mapping)
|
|
49
|
+
or not isinstance(patch.get("params", {}), Mapping)
|
|
50
|
+
or not isinstance(patch.get("spec", {}), Mapping)):
|
|
43
51
|
raise ValueError
|
|
44
52
|
# Reject non-finite numbers for SDK input as well as CLI JSON files.
|
|
45
53
|
json.dumps(patch, allow_nan=False)
|
|
@@ -49,6 +57,18 @@ def checked_patch(patch, name, version):
|
|
|
49
57
|
return dict(patch)
|
|
50
58
|
|
|
51
59
|
|
|
60
|
+
def merge_v2_params(base, override, secrets=()):
|
|
61
|
+
"""Overlay parameter values using the Runner's case insensitive names."""
|
|
62
|
+
if not isinstance(base, Mapping) or not isinstance(override, Mapping):
|
|
63
|
+
raise RemoteError("INVALID_PROFILE", 422)
|
|
64
|
+
targets = [key.upper() for key in override]
|
|
65
|
+
if len(set(targets)) != len(targets):
|
|
66
|
+
raise RemoteError("INVALID_PROFILE", 422)
|
|
67
|
+
marked = {key.upper(): key for key in secrets}
|
|
68
|
+
changes = {marked.get(key.upper(), key): item for key, item in override.items()}
|
|
69
|
+
return {key: item for key, item in base.items() if key.upper() not in targets} | changes
|
|
70
|
+
|
|
71
|
+
|
|
52
72
|
def with_runner(value, runner):
|
|
53
73
|
"""Keep a file binding and explicit CLI binding unambiguous."""
|
|
54
74
|
if not isinstance(value, Mapping) or value.get('runner', runner) != runner:
|
|
@@ -0,0 +1,49 @@
|
|
|
1
|
+
"""``el job run start|live|steer|stop``: thin wrappers over the platform's RSI routes."""
|
|
2
|
+
|
|
3
|
+
import sys
|
|
4
|
+
import time
|
|
5
|
+
|
|
6
|
+
FINISHED = frozenset({"finished", "failed", "stopped", "unreadable"})
|
|
7
|
+
POLL_SECONDS = 3.0
|
|
8
|
+
|
|
9
|
+
|
|
10
|
+
def _brief(args):
|
|
11
|
+
if args.brief is not None:
|
|
12
|
+
return args.brief
|
|
13
|
+
if args.brief_file == "-":
|
|
14
|
+
return sys.stdin.read()
|
|
15
|
+
with open(args.brief_file, encoding="utf-8") as handle:
|
|
16
|
+
return handle.read()
|
|
17
|
+
|
|
18
|
+
|
|
19
|
+
def _follow(runs, run, since):
|
|
20
|
+
"""Print new events as ``seq kind: text`` until the run ends; the run itself is left alone."""
|
|
21
|
+
while True:
|
|
22
|
+
live = runs.live(run, since)
|
|
23
|
+
for event in live.get("events") or []:
|
|
24
|
+
print(f"{event.get('seq')} {event.get('kind')}: {event.get('text') or ''}", flush=True)
|
|
25
|
+
if isinstance(live.get("next_cursor"), int):
|
|
26
|
+
since = live["next_cursor"]
|
|
27
|
+
if live.get("state") in FINISHED:
|
|
28
|
+
print(f"state: {live['state']}", flush=True)
|
|
29
|
+
return
|
|
30
|
+
time.sleep(POLL_SECONDS)
|
|
31
|
+
|
|
32
|
+
|
|
33
|
+
def run_rsi(runs, args):
|
|
34
|
+
"""Run one action; returns the result to print, or None when it printed its own output."""
|
|
35
|
+
if args.run_action == "start":
|
|
36
|
+
return runs.start(args.run, _brief(args), model_usd_cap=args.model_usd_cap, usd_cap=args.usd_cap,
|
|
37
|
+
model_request_cap=args.model_request_cap, domain=args.domain, rounds=args.rounds,
|
|
38
|
+
dry_run=args.dry_run, student_model=args.student_model)
|
|
39
|
+
if args.run_action == "live":
|
|
40
|
+
if not args.follow:
|
|
41
|
+
return runs.live(args.run, args.since)
|
|
42
|
+
try:
|
|
43
|
+
_follow(runs, args.run, args.since)
|
|
44
|
+
except KeyboardInterrupt:
|
|
45
|
+
pass
|
|
46
|
+
return None
|
|
47
|
+
if args.run_action == "steer":
|
|
48
|
+
return runs.steer(args.run, args.text)
|
|
49
|
+
return runs.control(args.run, "stop")
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|