AllToAllDispatchScenario
Communication scenarios · Implementation: mlsysim/engine/scenario_definitions/communication.py.
Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.
One expert-parallel dispatch: total_tokens of hidden_dim values over n_devices holding n_experts.
Notes
Equation: Compose mlsysim.solvers.AllToAllDispatchModel.
Assumptions: The stated fields are inputs to mlsysim.solvers.AllToAllDispatchModel; its documented validity regime, units, and modeling choices apply unchanged.
Excludes: This scenario adds no measurement, empirical calibration, or mechanisms beyond mlsysim.solvers.AllToAllDispatchModel.
References
Derivation: this scenario binds its input fields to the referenced implementation contracts: mlsysim.solvers.AllToAllDispatchModel. The methods below identify any additional arithmetic.
Examples
>>> from mlsysim import Scenarios
>>> case = Scenarios.MoeDispatch8Gpu64Experts
>>> type(case.result).__name__
'AllToAllDispatchResult'Attributes
| Name | Description |
|---|---|
| alpha | |
| bandwidth | |
| bytes_per_element | |
| hidden_dim | |
| n_devices | |
| n_experts | |
| receive_skew | |
| result | Evaluate this scenario through mlsysim.solvers.AllToAllDispatchModel.solve. |
| total_tokens |
total_tokens
Type: int.
Declared default: required.
n_devices
Type: int.
Declared default: required.
n_experts
Type: int.
Declared default: required.
bytes_per_element
Type: Quantity.
Declared default: required.
alpha
Type: Quantity.
Declared default: required.
bandwidth
Type: Quantity.
Declared default: required.
receive_skew
Type: float.
Declared default: 1.0.
result
engine.scenario_types.AllToAllDispatchScenario.result
Evaluate this scenario through mlsysim.solvers.AllToAllDispatchModel.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.