TensorParallelScenario
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.
The activation AllReduces of one tensor-parallel step of workload.
Notes
Equation: Compose mlsysim.solvers.TensorParallelCommModel.
Assumptions: The stated fields are inputs to mlsysim.solvers.TensorParallelCommModel; its documented validity regime, units, and modeling choices apply unchanged.
Excludes: This scenario adds no measurement, empirical calibration, or mechanisms beyond mlsysim.solvers.TensorParallelCommModel.
References
Derivation: this scenario binds its input fields to the referenced implementation contracts: mlsysim.solvers.TensorParallelCommModel. The methods below identify any additional arithmetic.
Examples
>>> from mlsysim import Scenarios
>>> case = Scenarios.Gpt3TensorParallelStepDgxH100
>>> type(case.result).__name__
'TensorParallelCommResult'Attributes
| Name | Description |
|---|---|
| collectives_per_layer | |
| compute_time | |
| link_bandwidth | |
| microbatch | |
| result | Evaluate this scenario through mlsysim.solvers.TensorParallelCommModel.solve. |
| seq_len | |
| tp_size | |
| workload |
workload
Type: Workload.
Declared default: required.
tp_size
Type: int.
Declared default: required.
microbatch
Type: int.
Declared default: required.
link_bandwidth
Type: Quantity.
Declared default: required.
compute_time
Type: Quantity.
Declared default: required.
seq_len
Type: Optional[int].
Declared default: None.
collectives_per_layer
Type: int.
Declared default: 2.
result
engine.scenario_types.TensorParallelScenario.result
Evaluate this scenario through mlsysim.solvers.TensorParallelCommModel.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.