Fine-tune LLMs from one YAML. Layer streaming trains an 8B model on a 4 GB laptop GPU.
Python cliconsumer-gpudpofine-tuningggufhuggingfacellmllmopslocal-ailocal-llmloralow-vrammachine-learningollamapeftpythonpytorchqlorasfttransformers
"""Build Web UI test clients using the address served by create_app."""
from __future__ import annotations
from typing import TYPE_CHECKING
if TYPE_CHECKING:
from fastapi.testclient import TestClient
def ui_client(app=None, *, host: str = "127.0.0.1", port: int = 7860, **kwargs) -> TestClient:
"""Use the served address; IPv6 binds need explicit Host headers on requests."""
from fastapi.testclient import TestClient
from soup_cli.ui.app import create_app
if app is None:
app = create_app(host=host, port=port)
# Starlette 1.6 TestClient cannot parse an IPv6 netloc.
url_host = "127.0.0.1" if ":" in host else host
return TestClient(app, base_url=f"http://{url_host}:{port}", **kwargs)