Registry / ai-ml / pylate

pylate

JSON →

library1.5.0pypypi✓ verified 82d ago

A library for training and retrieval with ColBERT. Current version is 1.5.0, release cadence is irregular.

ai-ml

pip install pylate

github ↗homepage ↗

INSTALL

IMPORT

SIG · PYLATE

pylate

ai-mlpythonv1.5.0

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

annotations

✓ from pylate import annotations

✗ from pylate import ColBERT

Basic usage: encode documents, build index, and retrieve.

from pylate import ColBERT, IndexConfig, Reranker import os colbert = ColBERT() index_config = IndexConfig( nbits=2, kmeans_niters=20, ncentroids=256, index_key="IVF256,PQ2", index_type="ivf", gpu=False, ) # Example documents documents = [ "ColBERT is a fast and accurate retrieval model.", "pylate provides a simple interface for training and indexing.", ] # Encode and index embeddings = colbert.encode(documents) colbert.index.add(embeddings, index_config=index_config) # Query query = "What is ColBERT?" query_embedding = colbert.encode([query])[0] scores = colbert.index.search(query_embedding, k=2) print(scores)

Debug

Known issues

breakingVersion 1.0.0 changed the API significantly: IndexConfig and index management moved to top-level. Old patterns like `pylate.models.ColBERT` no longer work.

fix

Import from top-level: `from pylate import ColBERT, IndexConfig`.

affects: >=1.0.0

gotchaThe `encode` method returns a list of tensors, not a single tensor. Iterating over documents yields one embedding per document.

fix

Use list comprehension or loop to handle individual embeddings.

affects: all

deprecated`ColBERT.load()` is deprecated in favor of `ColBERT.from_pretrained()`. The old method may be removed in a future release.

fix

Use `ColBERT.from_pretrained('colbert-base')` instead.

affects: >=1.2.0

gotchaThe index is not automatically persisted. After calling `colbert.index.add()`, the index is in memory only; you must call `colbert.index.save()` to persist to disk.

fix

Call `colbert.index.save('path/to/index')` after adding embeddings.

affects: all

Upgrade

Version history

1.5.0latest on PyPI · released May 5, 2026

Audit

Dependencies

torchrequiredRequired for tensor operations and neural network execution.

numpyrequiredUsed for numerical operations and array handling.

Agent activity

10 hits · last 30 days

node

OpenAI (training)

Resources

githubhttps://github.com/lightonai/pylate.git ↗

homepagehttps://github.com/lightonai/pylate.git ↗