File size: 6,829 Bytes
779dd0e 8f37329 779dd0e 095f297 779dd0e 8f37329 9de8f1b 7105268 9de8f1b 8f37329 779dd0e 8f37329 779dd0e 7e8e4e0 779dd0e 8f37329 779dd0e 8f37329 779dd0e 095f297 5711ab3 779dd0e 8f37329 779dd0e 8f37329 779dd0e 095f297 779dd0e 8f37329 779dd0e 8f37329 779dd0e 5711ab3 779dd0e 095f297 779dd0e | 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 | """Submit the reviewed translation Job without hard-coding a moving runner SHA.
This launcher is run by an operator with access to the ``hf-doc-build``
namespace. It resolves the current immutable commit of the public runner
dataset, then uses that same SHA for an in-Job snapshot download and the
``RUNNER_REVISION`` manifest field.
"""
from __future__ import annotations
import argparse
import json
import os
import re
from typing import Any
RUNNER_REPOSITORY = "hf-doc-build/translation-runner"
MODEL_REPOSITORY = "google/translategemma-27b-it"
MODEL_REVISION = "7d10f0b72f89a2d0f268cea30727d8b77c0d25c2"
DOC_BUILDER_REVISION = "e60a538eea9817ab312196d0d233604b01697265"
IMAGE = "huggingface/transformers-pytorch-gpu:latest"
NAMESPACE = "hf-doc-build"
RUNNER_LOCAL_DIR = "/tmp/translation-runner"
CONFIG_PATH = f"{RUNNER_LOCAL_DIR}/configs/transformers-ja-job.yml"
BOOTSTRAP = """\
import json
import os
from pathlib import Path
import subprocess
import sys
subprocess.run(
[
sys.executable,
"-m",
"pip",
"install",
"--disable-pip-version-check",
"--no-cache-dir",
"--upgrade",
"huggingface-hub==1.8.0",
"wheel>=0.38",
],
check=True,
)
from huggingface_hub import snapshot_download
revision = os.environ["RUNNER_REVISION"]
runner = Path(
snapshot_download(
repo_id="hf-doc-build/translation-runner",
repo_type="dataset",
revision=revision,
local_dir="/tmp/translation-runner",
cache_dir="/tmp/runner-download-cache",
force_download=True,
)
)
config_path = runner / "configs" / "transformers-ja-job.yml"
if not config_path.is_file():
raise RuntimeError(f"runner snapshot is incomplete: {config_path}")
runner_arguments = json.loads(os.environ["RUNNER_ARGUMENTS"])
if not isinstance(runner_arguments, list) or not all(isinstance(item, str) for item in runner_arguments):
raise RuntimeError("RUNNER_ARGUMENTS must be a JSON array of strings")
environment = os.environ.copy()
runner_src = str(runner / "src")
existing_pythonpath = environment.get("PYTHONPATH")
environment["PYTHONPATH"] = os.pathsep.join(
value for value in (runner_src, existing_pythonpath) if value
)
command = ["python3", "-m", "hf_doc_translation.sync", *runner_arguments]
os.execvpe(command[0], command, environment)
"""
def _runner_revision(api: Any) -> str:
revision = str(api.dataset_info(RUNNER_REPOSITORY).sha or "")
if not re.fullmatch(r"[0-9a-f]{40}", revision):
raise RuntimeError(f"Hub did not return an immutable runner revision: {revision!r}")
return revision
def _volumes(Volume: Any) -> list[Any]:
return [
Volume(type="model", source=MODEL_REPOSITORY, mount_path="/model", revision=MODEL_REVISION, read_only=True),
Volume(type="bucket", source="hf-doc-build/doc-translation-cache", mount_path="/translation-cache"),
Volume(
type="bucket",
source="hf-doc-build/doc-build-cache",
mount_path="/doc-build-cache",
read_only=True,
),
]
def _common_args() -> list[str]:
return [
"--config",
CONFIG_PATH,
"--repository",
"stevhliu/transformers",
"--base-ref",
"ja-translation",
"--environment",
"staging",
"--model-path",
"/model",
"--cache-dir",
"/translation-cache",
"--runner-revision",
"{runner_revision}",
]
def _command_arguments(mode: str, runner_revision: str) -> list[str]:
arguments = ["smoke-batching"] if mode == "smoke" else []
arguments.extend(value.format(runner_revision=runner_revision) for value in _common_args())
if mode == "backfill":
arguments.append("--force-backfill")
return arguments
def _github_secrets(mode: str) -> dict[str, str] | None:
"""Read the GitHub credential only from the submitter's environment."""
if mode not in {"backfill", "schedule"}:
return None
token = os.environ.get("GITHUB_TOKEN") or os.environ.get("GH_TOKEN")
if not token:
raise RuntimeError(
"publishing requires GITHUB_TOKEN or GH_TOKEN in the local environment; "
"the value is passed encrypted to the Job and is never read from the repository"
)
return {"GITHUB_TOKEN": token}
def _status_json(job: Any) -> dict[str, str | None] | None:
status = getattr(job, "status", None)
if status is None:
return None
stage = getattr(status, "stage", None)
return {
"stage": getattr(stage, "value", str(stage)) if stage is not None else None,
"message": getattr(status, "message", None),
}
def submit(mode: str, soak_complete: bool = False) -> dict[str, Any]:
from huggingface_hub import HfApi, Volume
if mode == "schedule" and not soak_complete:
raise RuntimeError("refusing to enable the daily schedule before the 30-day staging soak is complete")
api = HfApi()
runner_revision = _runner_revision(api)
command_mode = "backfill" if mode in {"backfill", "schedule"} else "smoke"
runner_arguments = _command_arguments(command_mode, runner_revision)
environment = {
"RUNNER_REVISION": runner_revision,
"RUNNER_ARGUMENTS": json.dumps(runner_arguments),
"TRANSLATEGEMMA_REVISION": MODEL_REVISION,
"DOC_BUILDER_REVISION": DOC_BUILDER_REVISION,
}
secrets = _github_secrets(mode)
kwargs = {
"image": IMAGE,
"command": ["python3", "-c", BOOTSTRAP],
"env": environment,
"secrets": secrets,
"flavor": "a100-large",
"timeout": "12h" if command_mode == "backfill" else "2h",
"labels": {"purpose": "transformers-ja-doc-sync", "environment": "staging"},
"volumes": _volumes(Volume),
"namespace": NAMESPACE,
}
job = api.create_scheduled_job(schedule="@daily", suspend=False, concurrency=False, **kwargs) if mode == "schedule" else api.run_job(**kwargs)
return {
"mode": mode,
"runner_revision": runner_revision,
"job_id": getattr(job, "id", None),
"job_url": getattr(job, "url", None),
"status": _status_json(job),
}
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("mode", choices=("smoke", "backfill", "schedule"))
parser.add_argument(
"--soak-complete",
action="store_true",
help="required safety acknowledgement before enabling the daily schedule",
)
args = parser.parse_args()
try:
result = submit(args.mode, soak_complete=args.soak_complete)
except RuntimeError as exc:
parser.error(str(exc))
print(json.dumps(result, indent=2, sort_keys=True))
return 0
if __name__ == "__main__":
raise SystemExit(main())
|