Implement Beam Search for LLM Decoding in PyTorch
Mediumllm-decodingauto-graded
Maintain and expand top-scoring partial sequences at each decoding step with length normalization and early stopping.
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("beam-search", beam_search)