MoCo keeps a fixed-size first-in, first-out queue of keys from past batches. Each training step pushes the newest keys in and drops the oldest, so the negatives are always from recent batches.
Implement enqueue(queue, new_keys, K), which returns the queue with the new keys appended and capped at K entries.
K rows.Concatenate the new keys after the queue, then keep the last K rows.
No results yet
Run your code to check it against the examples, or submit to run every test.