Spaces:
Running
Running
File size: 8,360 Bytes
a36db1b | 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 | from __future__ import annotations
import random
from dataclasses import dataclass
from enum import Enum
from typing import Any
class JobStatus(str, Enum):
QUEUED = "queued"
RUNNING = "running"
COMPLETE = "complete"
FAILED = "failed"
TIMED_OUT = "timed_out"
@dataclass
class GPUJob:
job_id: str
priority: int
memory_required: int
steps_to_complete: int
deadline: int
owner: str
status: JobStatus = JobStatus.QUEUED
assigned_gpu: str | None = None
actual_progress: float = 0.0
reported_progress: float = 0.0
completed_at: int | None = None
class JobQueue:
"""Job queue with hidden priorities, deadlines, and progress tracking."""
def __init__(self, jobs: list[GPUJob] | None = None) -> None:
self._jobs: dict[str, GPUJob] = {}
for job in jobs or []:
self.submit(job)
@classmethod
def generate(
cls,
count: int,
seed: int | None = None,
min_memory: int = 10,
max_memory: int = 75,
min_steps: int = 2,
max_steps: int = 12,
deadline_min: int = 12,
deadline_max: int = 120,
) -> "JobQueue":
if count <= 0:
raise ValueError("count must be positive.")
rng = random.Random(seed)
jobs = [
GPUJob(
job_id=f"JOB-{idx:03d}",
priority=rng.randint(1, 5),
memory_required=rng.randint(min_memory, max_memory),
steps_to_complete=rng.randint(min_steps, max_steps),
deadline=rng.randint(deadline_min, deadline_max),
owner=f"team-{rng.randint(1, 4)}",
)
for idx in range(count)
]
return cls(jobs)
def submit(self, job: GPUJob) -> str:
if job.job_id in self._jobs:
raise ValueError(f"Duplicate job_id: {job.job_id}")
if not 1 <= job.priority <= 5:
raise ValueError("priority must be in range 1..5.")
if job.memory_required <= 0:
raise ValueError("memory_required must be positive.")
if job.steps_to_complete <= 0:
raise ValueError("steps_to_complete must be positive.")
self._jobs[job.job_id] = job
return job.job_id
def get(self, job_id: str) -> GPUJob:
if job_id not in self._jobs:
raise KeyError(f"Unknown job_id: {job_id}")
return self._jobs[job_id]
def assign(self, job_id: str, gpu_id: str) -> bool:
job = self.get(job_id)
if job.status not in (JobStatus.QUEUED, JobStatus.RUNNING):
return False
job.status = JobStatus.RUNNING
job.assigned_gpu = gpu_id
return True
def unassign(self, job_id: str) -> bool:
job = self.get(job_id)
if job.status != JobStatus.RUNNING:
return False
job.status = JobStatus.QUEUED
job.assigned_gpu = None
return True
def tick(self, current_step: int, active_job_ids: set[str] | None = None) -> list[str]:
"""
Advance job progress and mark deadlines.
active_job_ids lets the environment pass jobs currently allocated on
GPUs. If omitted, all RUNNING jobs advance.
"""
timed_out: list[str] = []
for job in self._jobs.values():
if job.status in (JobStatus.COMPLETE, JobStatus.FAILED, JobStatus.TIMED_OUT):
continue
if current_step > job.deadline:
job.status = JobStatus.TIMED_OUT
job.assigned_gpu = None
timed_out.append(job.job_id)
continue
if job.status == JobStatus.RUNNING and (
active_job_ids is None or job.job_id in active_job_ids
):
increment = 1.0 / job.steps_to_complete
job.actual_progress = min(1.0, job.actual_progress + increment)
job.reported_progress = max(job.reported_progress, job.actual_progress)
if job.actual_progress >= 1.0:
job.status = JobStatus.COMPLETE
job.completed_at = current_step
job.assigned_gpu = None
return timed_out
def advance(
self,
job_id: str,
current_step: int,
progress_multiplier: float = 1.0,
) -> bool:
"""
Advance one running job by a worker-specific speed multiplier.
Returns True when the job is complete after this advancement.
"""
job = self.get(job_id)
if job.status != JobStatus.RUNNING:
return job.status == JobStatus.COMPLETE
if current_step > job.deadline:
job.status = JobStatus.TIMED_OUT
job.assigned_gpu = None
return False
increment = max(0.0, progress_multiplier) / job.steps_to_complete
job.actual_progress = min(1.0, job.actual_progress + increment)
job.reported_progress = max(job.reported_progress, job.actual_progress)
if job.actual_progress >= 1.0:
job.status = JobStatus.COMPLETE
job.completed_at = current_step
job.assigned_gpu = None
return True
return False
def complete(self, job_id: str, actual: bool = True, current_step: int | None = None) -> float:
job = self.get(job_id)
if actual:
job.actual_progress = 1.0
job.reported_progress = 1.0
job.status = JobStatus.COMPLETE
job.completed_at = current_step
job.assigned_gpu = None
return 1.0
job.reported_progress = 1.0
return 0.0
def fail(self, job_id: str) -> bool:
job = self.get(job_id)
if job.status in (JobStatus.COMPLETE, JobStatus.TIMED_OUT):
return False
job.status = JobStatus.FAILED
job.assigned_gpu = None
return True
def pending_jobs(self) -> list[GPUJob]:
return [job for job in self._jobs.values() if job.status == JobStatus.QUEUED]
def running_jobs(self) -> list[GPUJob]:
return [job for job in self._jobs.values() if job.status == JobStatus.RUNNING]
def active_job_ids(self) -> set[str]:
return {job.job_id for job in self.running_jobs()}
def deadline_pressure(self, current_step: int, window: int = 10) -> list[GPUJob]:
return [
job for job in self._jobs.values()
if job.status in (JobStatus.QUEUED, JobStatus.RUNNING)
and current_step <= job.deadline <= current_step + window
]
def completion_rate(self) -> float:
if not self._jobs:
return 0.0
completed = sum(1 for job in self._jobs.values() if job.status == JobStatus.COMPLETE)
return completed / len(self._jobs)
def deadline_hit_rate(self) -> float:
completed = [job for job in self._jobs.values() if job.status == JobStatus.COMPLETE]
if not completed:
return 0.0
hits = sum(1 for job in completed if job.completed_at is not None and job.completed_at <= job.deadline)
return hits / len(completed)
def snapshot(self, include_hidden: bool = False) -> list[dict[str, Any]]:
rows: list[dict[str, Any]] = []
for job in self._jobs.values():
row = {
"job_id": job.job_id,
"memory_required": job.memory_required,
"steps_to_complete": job.steps_to_complete,
"deadline": job.deadline,
"owner": job.owner,
"status": job.status.value,
"assigned_gpu": job.assigned_gpu,
"reported_progress": round(job.reported_progress, 3),
}
if include_hidden:
row["priority"] = job.priority
row["actual_progress"] = round(job.actual_progress, 3)
rows.append(row)
return rows
def summary(self) -> dict[str, Any]:
statuses = {status.value: 0 for status in JobStatus}
for job in self._jobs.values():
statuses[job.status.value] += 1
return {
"jobs_total": len(self._jobs),
"statuses": statuses,
"completion_rate": round(self.completion_rate(), 4),
"deadline_hit_rate": round(self.deadline_hit_rate(), 4),
}
|