mirror of
https://github.com/harvard-edge/cs249r_book.git
synced 2026-05-07 18:18:42 -05:00
The solver.py refactoring renamed most solver classes from *Solver to *Model (e.g. DistributedSolver → DistributedModel). The docs still referenced the old names, causing the Quarto site build to fail with: ImportError: cannot import name 'DistributedSolver' from 'mlsysim' - Fix executable code cells in tutorials/distributed.qmd - Update non-executable code examples across 10 doc files - Rename 19 API reference files from *Solver.qmd to *Model.qmd - SensitivitySolver and SynthesisSolver retain their names (correct) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
51 lines
2.2 KiB
Plaintext
51 lines
2.2 KiB
Plaintext
# core.solver.ResponsibleEngineeringModel { #mlsysim.core.solver.ResponsibleEngineeringModel }
|
|
|
|
```python
|
|
core.solver.ResponsibleEngineeringModel()
|
|
```
|
|
|
|
Analyzes the 'Responsibility Tax' — the systems cost of ethical mandates.
|
|
|
|
This solver models the physical overheads (compute, data, carbon) required
|
|
to satisfy mandates for privacy (DP-SGD), fairness (sub-group representation),
|
|
and sustainability (embodied manufacturing carbon).
|
|
|
|
Literature Source:
|
|
1. Abadi et al. (2016), "Deep Learning with Differential Privacy."
|
|
2. Gupta et al. (2021), "ACT: Architectural Carbon Modeling."
|
|
3. Janapa Reddi et al. (2025), Chapter 15 (Responsible Engineering).
|
|
|
|
## Methods
|
|
|
|
| Name | Description |
|
|
| --- | --- |
|
|
| [solve](#mlsysim.core.solver.ResponsibleEngineeringModel.solve) | Solves for the overheads of responsible engineering mandates. |
|
|
|
|
### solve { #mlsysim.core.solver.ResponsibleEngineeringModel.solve }
|
|
|
|
```python
|
|
core.solver.ResponsibleEngineeringModel.solve(
|
|
workload,
|
|
hardware,
|
|
privacy_epsilon=None,
|
|
fairness_disparity=1.0,
|
|
)
|
|
```
|
|
|
|
Solves for the overheads of responsible engineering mandates.
|
|
|
|
#### Parameters {.doc-section .doc-section-parameters}
|
|
|
|
| Name | Type | Description | Default |
|
|
|--------------------|--------------|--------------------------------------------------------------------------------|------------|
|
|
| workload | Workload | The model workload. | _required_ |
|
|
| hardware | HardwareNode | The execution hardware. | _required_ |
|
|
| privacy_epsilon | float | Differential Privacy budget (ε). Lower means more privacy but higher overhead. | `None` |
|
|
| fairness_disparity | float | Current representation disparity (e.g., 10.0 for 10x disparity). | `1.0` |
|
|
|
|
#### Returns {.doc-section .doc-section-returns}
|
|
|
|
| Name | Type | Description |
|
|
|--------|------------------|----------------------------------------------------------|
|
|
| | Dict\[str, Any\] | Overhead metrics for compute, data, and embodied carbon. |
|