Palash DebnathandClaude Opus 4.8
5a723c0408
Windows users reported 'create demo voice runs indefinitely, no audio, no error'. Root cause: the first /generate triggers OmniVoice.from_pretrained() which downloads multi-GB weights via the legacy LFS path (HF_HUB_DISABLE_XET=1), with NO timeout anywhere. A stalled socket (proxy/firewall/AV) blocks the GPU- pool worker forever inside get_model() -- before the try/except that would surface an error -- and the frontend /generate fetch had no abort, so the spinner spun forever with no toast. - backend/main.py: set HF_HUB_ETAG_TIMEOUT=15 + HF_HUB_DOWNLOAD_TIMEOUT=30 (per-read timeout: resets on each chunk, so slow-but-progressing downloads are never punished; only a dead socket trips it). Set before hf import. - model_manager: get_model()/preload_model() now load via _load_model_with_timeout(), an asyncio.wait_for backstop (OMNIVOICE_MODEL_LOAD_TIMEOUT, default 1200s) that drops the poisoned GPU pool and raises a clear, actionable RuntimeError so a retry gets a fresh worker instead of queueing behind the wedged one. - useTTS.js: AbortController backstop on /generate so the UI never spins forever even if the backend is unreachable; friendly timeout toast. - tests: watchdog raises + resets pool + releases lock; env/floor parsing. Cross-platform (no OS-specific behavior); backward-compatible with installed models; local-first preserved. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
64 lines
2.3 KiB
Python
64 lines
2.3 KiB
Python
"""Regression: model load/download must never hang forever (Windows demo-voice
|
|
"runs indefinitely, no audio, no error" report).
|
|
|
|
`get_model()` wraps the blocking load in an `asyncio.wait_for` deadline; on
|
|
timeout it drops the poisoned GPU pool and raises a clear RuntimeError instead
|
|
of leaving `/generate` (and the UI spinner) wedged forever.
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import sys
|
|
import threading
|
|
|
|
import pytest
|
|
|
|
|
|
@pytest.fixture
|
|
def model_manager(monkeypatch):
|
|
for mod_name in ("core.config", "services.model_manager"):
|
|
if getattr(sys.modules.get(mod_name), "__file__", None) is None:
|
|
sys.modules.pop(mod_name, None)
|
|
|
|
import services.model_manager as mm
|
|
|
|
monkeypatch.setattr(mm, "model", None)
|
|
return mm
|
|
|
|
|
|
def test_model_load_timeout_respects_env(model_manager, monkeypatch):
|
|
mm = model_manager
|
|
monkeypatch.delenv("OMNIVOICE_MODEL_LOAD_TIMEOUT", raising=False)
|
|
assert mm._model_load_timeout() == 1200.0
|
|
monkeypatch.setenv("OMNIVOICE_MODEL_LOAD_TIMEOUT", "5000")
|
|
assert mm._model_load_timeout() == 5000.0
|
|
monkeypatch.setenv("OMNIVOICE_MODEL_LOAD_TIMEOUT", "not-a-number")
|
|
assert mm._model_load_timeout() == 1200.0
|
|
monkeypatch.setenv("OMNIVOICE_MODEL_LOAD_TIMEOUT", "1") # below the safety floor
|
|
assert mm._model_load_timeout() == 30.0
|
|
|
|
|
|
def test_get_model_times_out_and_resets_pool(model_manager, monkeypatch):
|
|
mm = model_manager
|
|
# Isolated lock so we never reuse one bound to a previous test's loop.
|
|
monkeypatch.setattr(mm, "_model_lock", asyncio.Lock())
|
|
monkeypatch.setattr(mm, "_model_load_timeout", lambda: 0.3)
|
|
|
|
release = threading.Event()
|
|
|
|
def _hang(): # simulates a wedged download that never returns
|
|
release.wait(2.0)
|
|
return object()
|
|
|
|
monkeypatch.setattr(mm, "_load_model_sync", _hang)
|
|
assert mm._get_gpu_pool() is not None # a pool exists before the timeout
|
|
|
|
try:
|
|
with pytest.raises(RuntimeError, match="timed out"):
|
|
asyncio.run(mm.get_model())
|
|
assert mm.model is None # no half-loaded model
|
|
assert mm._gpu_pool_singleton is None # poisoned pool was dropped
|
|
assert not mm._model_lock.locked() # lock released for a retry
|
|
finally:
|
|
release.set() # let the orphaned worker exit immediately
|