mirror of
https://github.com/osmarks/meme-search-engine.git
synced 2025-02-22 14:00:09 +00:00
e57931d47f
While profiling suggests that most operations are cheap and IO-bound rather than CPU-bound, the GEMM for deduplication is pretty slow. As such, use multiple threads for higher throughput.
30 lines
890 B
Python
30 lines
890 B
Python
import numpy as np
|
|
import aiohttp
|
|
import asyncio
|
|
import sys
|
|
|
|
queries = np.random.randn(1000, 1152)
|
|
|
|
async def main():
|
|
async with aiohttp.ClientSession() as sess:
|
|
async with asyncio.TaskGroup() as tg:
|
|
sem = asyncio.Semaphore(100)
|
|
async def lookup(embedding):
|
|
async with sess.post("http://localhost:5601", json={
|
|
"terms": [{ "embedding": list(float(x) for x in embedding) }], # sorry
|
|
"k": 10
|
|
}) as res:
|
|
sys.stdout.write(".")
|
|
sys.stdout.flush()
|
|
return (await res.json())["matches"]
|
|
|
|
async def dispatch(i):
|
|
await lookup(queries[i])
|
|
sem.release()
|
|
|
|
for i in range(1000):
|
|
await sem.acquire()
|
|
tg.create_task(dispatch(i))
|
|
|
|
asyncio.run(main())
|