Spaces:
Running
Running
File size: 1,632 Bytes
a1e3907 | 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 | spec_version: 1
name: berth_planning
type: space
runtime: fastapi
app: berth_openenv.server:app
port: 8000
description: Dock planning at the Port of Barcelona on real 2024 container calls - re-plan one to three weeks at a quay with real crane fleets, movement limits, gales under the port's wind rules, crane outages, closures, late and bunched ships, diverted traffic, emergencies and priority cargo; graded once on submit against the CP-SAT optimum. 1,050 train and 50 eval tasks (dock-v1), plus the first 100-task pack (berth-v1).
version: "0.1.0"
variables:
BERTH_TASKS_DIR: "" # colon-separated pack directories; defaults to tasks/dock-v1-eval:tasks/dock-v1-train
BERTH_RUNS_DIR: "" # rollout runs shown in the viewer
BERTH_MAX_CHECKS: "10"
BERTH_MAX_TOOL_CALLS: "24"
MAX_CONCURRENT_ENVS: "64"
validation:
reward:
range: [0.0, 1.0]
# oracle/solve.py submits the CP-SAT optimum and scores exactly 1.0; no plan scores 0, any rule broken <= 0.2
oracle_tolerance: 0.001
floor_margin: 0.5
resources:
# pure Python: the grader re-checks a plan in milliseconds, CP-SAT never runs on the server
cpu: 2.0
memory_mb: 4096
disk_mb: 2048
episode_timeout_s: 3600.0
network:
mode: no-network
capabilities:
verifier:
kind: reward_channel
oracle:
form: script
location: oracle/solve.py
rubric_tree: true
task_api: true
declared_tools: [get_situation, check_plan, submit_plan]
declared_task_count:
eval: 50
train: 1050
types:
tags: [simulation, logistics, scheduling, operations-research, mcp, real-world-data]
|