Example 1
Arguments: [
[
5,
1,
4,
2
],
4,
2
]
Expected result: [
[
1,
2
],
[
3
],
[
0
]
]Lengths 1,4,2 belong to bucket 0 in original index order [1,2,3], then index 0 forms bucket 1.
Batching and serving algorithms · Easy
Stable Length-Bucketed Batches
Problem 43 of 50 · Python and NumPy
Problem
Group requests by prompt-length bucket floor((length−1)/bucket_width). Process buckets in ascending order, preserving original request order within each bucket. Split each bucket into batches of at most batch_size without mixing buckets. Return batches of original zero-based request indices.
def length_bucket_batches(lengths, bucket_width, batch_size):Constraints
- lengths contains 0..64 positive integers ≤128.
- 1 ≤ bucket_width,batch_size ≤32. Bucket 0 covers lengths 1..bucket_width inclusive.
- Inputs satisfy these constraints; invalid-input handling is not graded. Use only Python standard library and NumPy.
Examples
Example 2
Arguments: [
[
4,
2,
3,
1
],
4,
3
]
Expected result: [
[
0,
1,
2
],
[
3
]
]All requests share a bucket, so retain their input order rather than sorting by length.