_pitch_preserving_stretch ran a blocking subprocess.run() inside the `_stream` async generator (on the event loop). Each ffmpeg atempo call is ~50-100 ms, so on a multi-segment time_stretch dub job it froze health checks, status SSE, and every other concurrent request for seconds. Convert to asyncio.create_subprocess_exec + await communicate() (same pattern as run_proc_streaming_stderr); await the call site in _stream. Drop the now-unused `import subprocess`. Tests: async coroutine + target-length + no-op cases (real ffmpeg). Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
32 lines
1.1 KiB
Python
32 lines
1.1 KiB
Python
"""`_pitch_preserving_stretch` must be async so it doesn't block the event
|
|
loop (Greptile P1 on PR #133's dub-timing path). It's awaited from the
|
|
`_stream` async generator; each ffmpeg invocation is ~50-100ms and on a
|
|
multi-segment job that froze health-checks / SSE / every concurrent request.
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
|
|
import numpy as np
|
|
import torch
|
|
|
|
from api.routers.dub_generate import _pitch_preserving_stretch
|
|
|
|
|
|
def test_stretch_is_a_coroutine_and_hits_target_length():
|
|
sr = 16000
|
|
wav = torch.zeros((1, sr), dtype=torch.float32) # 1s of silence
|
|
target = sr // 2 # compress to 0.5s
|
|
coro = _pitch_preserving_stretch(wav, target, sr)
|
|
assert asyncio.iscoroutine(coro), "must be async so it can't block the loop"
|
|
out = asyncio.run(coro)
|
|
assert out.shape == (1, target)
|
|
assert out.dtype == torch.float32
|
|
|
|
|
|
def test_stretch_noop_when_already_target_length():
|
|
sr = 16000
|
|
wav = torch.zeros((1, sr), dtype=torch.float32)
|
|
out = asyncio.run(_pitch_preserving_stretch(wav, sr, sr))
|
|
assert out.shape[-1] == sr
|