Lists, tuples, and sets

Lists: the workhorse

A list is an ordered, mutable sequence: docs = ["a", "b"].

• Index: docs[0], docs[-1]
• Slice: docs[:2] (first two — think “mini batch”)
• Add: append, extend
• Length: len(docs)

Lists can hold mixed types, but in ML code keep them homogeneous (all strings, or all floats). Mixed lists become bugs at tensor conversion time.

list.append(x) returns None. A classic bug is docs = docs.append(x), which wipes the list. Append in place; do not assign the result.

Tuples and sets

A tuple is ordered and immutable: shape = (32, 768). Use tuples for records that should not grow — image size, embedding dim, (train, val, test) split sizes.

A set stores unique unordered values: seen = {"id-1", "id-2"}. Membership tests (x in seen) are fast. Use sets to drop duplicate retrieved ids in RAG.

Convert: set(list_of_ids) deduplicates. list(the_set) if you need order back — but set order is not meaningful, so sort if you need stability: sorted(set(ids)).

Copy-paste examples

Copy into your own editor, or load one into the compiler below and press Run.

chunks = [
    "Python is used for ML pipelines.",
    "NumPy stores arrays for tensors.",
    "pandas cleans tabular data.",
    "FastAPI serves models.",
]
batch = chunks[:3]
print("batch size:", len(batch))
print("last in batch:", batch[-1])
batch.append("LangChain wires LLM calls.")
print("after append:", len(batch))

retrieved = ["d1", "d4", "d1", "d9", "d4"]
unique_ids = list(dict.fromkeys(retrieved))  # unique, keep order
print("raw:", retrieved)
print("unique ordered:", unique_ids)
print("as set:", set(retrieved))

batch_size = 8
hidden = 768
shape = (batch_size, hidden)
print("shape:", shape)
print("rank (ndim):", len(shape))
print("total values:", shape[0] * shape[1])

Takeaways

• Lists are ordered and mutable — batches, token lists, labels.
• Never assign the result of append(); it returns None.
• Sets drop duplicates; tuples hold fixed records like shapes.