AlignmentMemoryScenario
Training scenarios · Implementation: mlsysim/engine/scenario_definitions/training.py.
Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.
PPO and DPO alignment of policy with critic-sized reward and value models.
Notes
Equation: Compose mlsysim.solvers.AlignmentMemoryModel.
Assumptions: The stated fields are inputs to mlsysim.solvers.AlignmentMemoryModel; its documented validity regime, units, and modeling choices apply unchanged.
Excludes: This scenario adds no measurement, empirical calibration, or mechanisms beyond mlsysim.solvers.AlignmentMemoryModel.
References
Derivation: this scenario binds its input fields to the referenced implementation contracts: mlsysim.solvers.AlignmentMemoryModel. The methods below identify any additional arithmetic.
Examples
>>> from mlsysim import Scenarios
>>> case = Scenarios.Llama2_70B_PpoAlignment
>>> type(case.result).__name__
'AlignmentMemoryResult'Attributes
| Name | Description |
|---|---|
| critic | |
| device_memory | |
| dp_size | |
| policy | |
| pp_size | |
| result | Evaluate this scenario through mlsysim.solvers.AlignmentMemoryModel.solve. Equation: |
| rollout_sequences | |
| rollout_tokens | |
| tp_size |
policy
Type: Workload.
Declared default: required.
critic
Type: Workload.
Declared default: required.
tp_size
Type: int.
Declared default: required.
pp_size
Type: int.
Declared default: required.
dp_size
Type: int.
Declared default: 1.
rollout_sequences
Type: int.
Declared default: required.
rollout_tokens
Type: int.
Declared default: required.
device_memory
Type: Optional[Quantity].
Declared default: None.
result
engine.scenario_types.AlignmentMemoryScenario.result
Evaluate this scenario through mlsysim.solvers.AlignmentMemoryModel.solve. Equation: pass the stated fields to that implementation; its contract defines the calculation and feasibility tests. Units: each returned field follows the referenced result schema; no independent equation is evaluated by this wrapper.