Vijay Janapa Reddi f213260153 feat(mlsysim): align analytical solvers with industry-standard literature ...

Updated solvers to use literature-grade models for:
- Roofline Performance (Williams et al. 2009)
- Transformer Scaling (6PD rule, Kaplan et al. 2020)
- Training Memory (Shoeybi et al. 2019)
- Pipeline Parallelism (Huang et al. 2019)
- LLM Serving (Pope et al. 2023)
- Reliability (Young-Daly 1974/2006)

Introduced Hierarchical Communication Modeling and MFU/HFU metrics.
Fixed test suite imports and return key mismatches.
Updated Smart Doorbell scorecard reference in ml_systems.qmd.
Restored core __init__.py exports for backward compatibility.

2026-03-07 15:02:26 -05:00

..

__init__.py

feat(mlsysim): align analytical solvers with industry-standard literature

2026-03-07 15:02:26 -05:00