Analysis of Memory Scaling in Sequence Processing
Consider two different methods for processing a long sequence of data items one by one.
- Method A: At each step
i, it calculates an output by attending to all previous data items from 1 toi. To do this, it must keep every single past data item in memory. - Method B: At each step
i, it updates a fixed-size summary of the past. This summary from stepi-1is combined with the current data item at stepito create the new summary for stepi. Only this summary is kept in memory.
Analyze how the memory requirement for each method changes as the number of processed data items (i) grows very large. Which method is more suitable for processing extremely long or continuous data streams, and why?
0
1
Tags
Ch.2 Generative Models - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A model is designed to process a continuous, unending stream of input data. At each time step
i, it calculates an output that depends on all inputs from step 1 toi. The model's internal state is updated using a method where the state at stepiis a function of only the state at stepi-1and the current input at stepi. What is the primary implication of this update method for the model's memory requirements as the stream continues indefinitely?Architectural Trade-offs for Long-Sequence Processing
Analysis of Memory Scaling in Sequence Processing