Write a cached-vs-full-forward equivalence test for incremental decoding

Medium

Research Engineering & Debugging4 / 8

Chapter 254 / 8

Research Engineering & Debugging

Write a cached-vs-full-forward equivalence test for incremental decoding

Medium

Target interface

test_cache_equals_full(model, input_ids, rtol=0.0001, atol=1e-05)

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

def test_cache_equals_full(model, input_ids, rtol=0.0001, atol=1e-05):
    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 254 / 8

Research Engineering & Debugging

Write a cached-vs-full-forward equivalence test for incremental decoding

Medium

Target interface

test_cache_equals_full(model, input_ids, rtol=0.0001, atol=1e-05)

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

Hints0 / 2

Reference solution

Your own code stays in the editor.