Sign up to save your podcasts
Or
Share Memory Efficiency at Scale
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
Memory Efficiency at Scale
- The 224x Memory Reduction: A breakdown of how the Mohawk Protocol's streaming aggregation achieves this massive efficiency gain.
- Streaming vs. Monolithic Aggregation: The technical shift from loading massive, monolithic tensor payloads into memory to processing updates as a continuous stream of chunks.
- Scaling to 100M+ Nodes: How bounded working memory makes high fan-out participation feasible on commodity infrastructure by scaling with the window size rather than the global update size.
- Inline Verification: How trust and policy checks are executed incrementally alongside the aggregation stream to prevent bottlenecks
View all episodes
By Ryan Williams- The 224x Memory Reduction: A breakdown of how the Mohawk Protocol's streaming aggregation achieves this massive efficiency gain.
- Streaming vs. Monolithic Aggregation: The technical shift from loading massive, monolithic tensor payloads into memory to processing updates as a continuous stream of chunks.
- Scaling to 100M+ Nodes: How bounded working memory makes high fan-out participation feasible on commodity infrastructure by scaling with the window size rather than the global update size.
- Inline Verification: How trust and policy checks are executed incrementally alongside the aggregation stream to prevent bottlenecks