Implement Top-k Sampling in PyTorch

Mediumllm-decodingauto-graded

Select the k highest-probability tokens, zero out the rest, renormalize, and sample for controlled text generation.

Solve it

Check your answer

The grader verifies properties of your implementation, so a correct solution written differently from ours still passes.

pip install torchleet

from torchleet import check
check("top-k-sampling", top_k_sample)

Company tags

How these tags are sourced