QuotaLendingScenario
Optimization scenarios · Implementation: mlsysim/engine/scenario_definitions/optimization.py.
Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.
A team guaranteed guarantee GPUs that uses used_share of them; the rest can be lent.
Lent GPUs stay reclaimable: borrowers run them as lower-priority, preemptible work and return them when the owner’s demand comes back.
Notes
Equation: used devices = round(guarantee * used share); lendable = guarantee - used.
Assumptions: Usage share describes the guaranteed pool and rounds with Python nearest- integer semantics.
Excludes: Borrower demand, fairness policy, and reclaim latency.
References
Derivation: used devices = round(guarantee * used share); lendable = guarantee - used. The individual properties state units and intermediate accounting; this class introduces no fitted coefficient.
Examples
The named scenario supplies illustrative inputs; the output is calculated.
>>> from mlsysim import Scenarios
>>> case = Scenarios.TeamQuotaLending500
>>> round(float(case.used), 3)
300.0Attributes
| Name | Description |
|---|---|
| borrower_guarantee | |
| guarantee | |
| lendable | Guaranteed devices remaining after rounded usage; borrower demand is not modeled. |
| used | Used device count, rounding guarantee times the declared usage share to the |
| used_share |
guarantee
Type: int.
Declared default: required.
borrower_guarantee
Type: Optional[int].
Declared default: None.
used
engine.scenario_types.QuotaLendingScenario.used
Used device count, rounding guarantee times the declared usage share to the nearest integer.
Equation: return round(guarantee * used_share).
lendable
engine.scenario_types.QuotaLendingScenario.lendable
Guaranteed devices remaining after rounded usage; borrower demand is not modeled.
Equation: return guarantee - used.