Skip to content
mlmentorship

Book V · Chapter 3

Training plans and reliability

Size a run, choose a layout, and preserve consistent state through failures.

Priority
Role-specific
Difficulty
Advanced
Useful for
RE, Systems MLE, RS
Interview rounds
Systems design, Training incident

Read first: Distributed training and parallelism

Chapter contents

5 entries · read in order
  1. 01
    Walk me through how you'd train a 100B parameter model
    Question
  2. 02
    Plan and cost a 70B transformer training run
    Question
  3. 03
    Fault-tolerant collectives
    Concept
  4. 04
    Design fault-tolerant distributed training
    Question
  5. 05
    Debug a frontier LLM training run that suddenly diverges
    Question