Spaces:
Sleeping
Sleeping
File size: 932 Bytes
be3e455 25f063d be3e455 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 | """Pre-download models during Docker build to avoid cold-start penalty.
Called by the Dockerfile builder stage. This is separate from the main app
so it can be run once at build time and the downloaded caches get copied
into the final image.
"""
import logging
logging.basicConfig(level=logging.INFO)
# Trigger Docling model download
print("Pre-downloading Docling models...")
from src.core.converter import MarkdownConverterService # noqa: E402
MarkdownConverterService()
print("Docling models cached.")
# Trigger FlashRank model download
print("Pre-downloading FlashRank model...")
from src.retrieval.reranker import Reranker # noqa: E402
Reranker()
print("FlashRank model cached.")
# Trigger FastEmbed BM25 tokenizer download
print("Pre-downloading FastEmbed BM25 tokenizer...")
from fastembed import SparseTextEmbedding # noqa: E402
SparseTextEmbedding(model_name="Qdrant/bm25")
print("FastEmbed tokenizer cached.")
|