Expand description
Fused operations for maximum memory efficiency
Structs§
- Adam
Config - Adam optimizer configuration
Functions§
- fused_
adam_ update - Fused Adam update operation: combines momentum, variance, and parameter update in one pass
- fused_
apply_ constraints - Fused parameter constraint application
- fused_
gradient_ clip_ normalize - Fused gradient clipping and normalization
- fused_
sgd_ update - Fused SGD with momentum and weight decay