AI Architecture & Systems
1
Microarchitecture
2
Kernel
3
Compiler
4
Architecture
5
System
6
Algorithms
Compiler
/
§3.4
Automatic Differentiation and Training-Graph Compilation
3.4.1
Reverse-Mode and Forward-Mode AD
#
3.4.2
Backward Graph, VJP, and JVP
#
3.4.3
AOT Autograd and Joint Graph Optimization
#
3.4.4
Activation Liveness and Rematerialization
#
3.4.5
Gradient Accumulation, Gradient Communication, and the Optimizer Graph
#
← Previous
3.3 Intermediate Representation
→ Next
3.5 Graph-Level Optimization