Skip to content
mlmentorship

Book II

Model training and research

Optimization, reliable experiments, implementation, debugging, and research judgment.

9 chapters · 52 entries

Chapters

9
II.1

Objectives and regularization

Start with training objectives, then control fit and generalization.

Scope
Core
Difficulty
Foundation
Useful for
All ML roles
II.2

Optimization and schedules

Move from SGD and Adam to schedules, clipping, and initialization.

Scope
Core
Difficulty
Intermediate
Useful for
All ML roles
II.3

Numerics and training stability

Understand normalization, precision, floating-point behavior, and loss failures.

Scope
Core
Difficulty
Intermediate
Useful for
MLE, RE, RS
II.4

Efficiency, data, and scaling

Reduce memory and work, build training data, and reason about scale.

Scope
Role-specific
Difficulty
Advanced
Useful for
RE, RS, MLE
II.5

Training recipe and case study

Connect the parts into one training plan, then inspect a real open training log.

Scope
Core
Difficulty
Intermediate
Useful for
AS, RS, RE, MLE
II.6

Training and debugging questions

Practice diagnosis, numerical decisions, imbalance, sequence training, and optimizer trade-offs.

Scope
Core
Difficulty
Mixed
Useful for
MLE, RE, AS
II.7

ML implementation

Implement core algorithms and transformer mechanisms with tests and explicit contracts.

Scope
Core
Difficulty
Mixed
Useful for
MLE, RE, RS
II.8

Training code and codebase debugging

Debug a training loop and extend an unfamiliar ML repository safely.

Scope
Role-specific
Difficulty
Advanced
Useful for
MLE, RE
II.9

Research methods and derivations

Derive, critique, design ablations, and investigate behavior from evidence.

Scope
Core
Difficulty
Advanced
Useful for
RS, RE, AS