Skip to content

A straightforward method for training your LLM, from downloading data to generating text.

200 B
Raw Blame
"""Inference helpers: sampling, speculative decoding and weight quantization."""

from src.inference.sampling import filter_logits, sample_next_token

__all__ = ["filter_logits", "sample_next_token"]