Implement YaRN RoPE scaling (frequency grouping + attention-logit scaling) and demonstrate

Super-hard

Attention Efficiency & Long Context5 / 5

Chapter 035 / 5

Attention Efficiency & Long Context

Implement YaRN RoPE scaling (frequency grouping + attention-logit scaling) and demonstrate

Super-hard

Implement YaRN RoPE scaling (frequency grouping + attention-logit scaling) and demonstrate length extrapolation on a toy retrieval task.

Target interface

yarn_freqs(d, base=10000.0, L=256, scale=8.0, alpha=1.0, beta=32.0)

Implement the function/class skeleton in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.

solution.pyPython

Local draft

import torch, math

def yarn_freqs(d, base=10000.0, L=256, scale=8.0, alpha=1.0, beta=32.0):
    raise NotImplementedError

def build_rope(inv_freq, T, mscale=1.0):
    raise NotImplementedError

def rotate_half(x):
    raise NotImplementedError

def apply_rope(x, cos, sin):
    raise NotImplementedError

⌘/Ctrl↵to submit

AI review

No review yet

Submit your solution for a verdict, a score out of 100 and concrete feedback. Any correct approach passes.

Chapter 035 / 5

Attention Efficiency & Long Context

Implement YaRN RoPE scaling (frequency grouping + attention-logit scaling) and demonstrate

Super-hard

Implement YaRN RoPE scaling (frequency grouping + attention-logit scaling) and demonstrate length extrapolation on a toy retrieval task.

Target interface

yarn_freqs(d, base=10000.0, L=256, scale=8.0, alpha=1.0, beta=32.0)

Implement the function/class skeleton in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.