Text_to_signmotion / client_test.py
drrobot9's picture
Upload text to sign motion backend
81abf7f verified
Raw History Blame Contribute Delete
6.54 kB
import argparse
import asyncio
import json
import struct
import time
import urllib.request
import numpy as np
FRAME_DIMENSION = 133
BLOCK_HEADER = struct.Struct("<III")
class MotionCollector:
def __init__(self) -> None:
self.started_at = time.perf_counter()
self.first_block_seconds = None
self.total_frames_received = 0
self.blocks_received = 0
self.last_block = None
def elapsed(self) -> float:
return time.perf_counter() - self.started_at
def consume(self, message) -> bool:
if isinstance(message, (bytes, bytearray)):
request_index, frame_offset, frame_count = BLOCK_HEADER.unpack_from(message, 0)
expected_length = BLOCK_HEADER.size + frame_count * FRAME_DIMENSION * 4
if len(message) != expected_length:
print("Bad block length", len(message), "expected", expected_length)
return True
frames = np.frombuffer(message, dtype="<f4", offset=BLOCK_HEADER.size)
frames = frames.reshape(frame_count, FRAME_DIMENSION)
if self.first_block_seconds is None:
self.first_block_seconds = self.elapsed()
self.total_frames_received += frame_count
self.blocks_received += 1
self.last_block = frames
return False
data = json.loads(message)
message_type = data.get("type")
if message_type == "hello":
print("hello:", "ready =", data.get("ready"), "fps =", data.get("fps"), "transport =", data.get("transport"))
return False
if message_type == "sentence":
print(
"sentence", data["sentence_index"] + 1, "of", data["sentence_count"],
"frames =", data["frame_count"],
"generation_ms =", data["generation_ms"],
"cache_hit =", data["cache_hit"],
"text =", data["text"],
)
return False
if message_type == "done":
print("done: total_frames =", data["total_frames"], "server_elapsed_ms =", data["elapsed_ms"])
return True
if message_type == "error":
print("error:", data)
return True
print("message:", data)
return False
def report(self) -> None:
print("blocks received:", self.blocks_received)
print("frames received:", self.total_frames_received)
if self.first_block_seconds is not None:
print("seconds until first motion block:", round(self.first_block_seconds, 3))
print("seconds until finished:", round(self.elapsed(), 3))
if self.last_block is not None:
print("last block shape:", self.last_block.shape)
print("last block value range:", float(self.last_block.min()), float(self.last_block.max()))
def build_generate_message(text: str) -> str:
return json.dumps({"type": "generate", "id": "test-1", "text": text})
def http_request(url: str, token: str, payload=None) -> dict:
headers = {"Content-Type": "application/json"}
if token:
headers["Authorization"] = "Bearer " + token
request_body = None
if payload is not None:
request_body = json.dumps(payload).encode("utf-8")
request = urllib.request.Request(url, data=request_body, headers=headers)
with urllib.request.urlopen(request, timeout=60) as response:
return json.loads(response.read().decode("utf-8"))
async def run_websocket_test(base_url: str, text: str, token: str) -> None:
import websockets
websocket_url = base_url.replace("http", "ws", 1) + "/ws"
if token:
websocket_url += "?token=" + token
print("WEBSOCKET TEST", websocket_url)
collector = MotionCollector()
async with websockets.connect(websocket_url, max_size=None) as socket:
await socket.send(build_generate_message(text))
collector.started_at = time.perf_counter()
while True:
message = await asyncio.wait_for(socket.recv(), timeout=180)
if collector.consume(message):
break
collector.report()
async def run_webrtc_test(base_url: str, text: str, token: str) -> None:
from aiortc import RTCConfiguration, RTCIceServer, RTCPeerConnection, RTCSessionDescription
print("WEBRTC TEST", base_url)
ice_response = await asyncio.to_thread(http_request, base_url + "/ice", token)
ice_servers = []
for entry in ice_response["iceServers"]:
ice_servers.append(
RTCIceServer(
urls=entry["urls"],
username=entry.get("username"),
credential=entry.get("credential"),
)
)
peer = RTCPeerConnection(configuration=RTCConfiguration(iceServers=ice_servers))
channel = peer.createDataChannel("motion", ordered=True)
collector = MotionCollector()
finished = asyncio.get_event_loop().create_future()
@channel.on("open")
def on_open() -> None:
print("data channel open")
collector.started_at = time.perf_counter()
channel.send(build_generate_message(text))
@channel.on("message")
def on_message(message) -> None:
if collector.consume(message) and not finished.done():
finished.set_result(True)
offer = await peer.createOffer()
await peer.setLocalDescription(offer)
answer = await asyncio.to_thread(
http_request,
base_url + "/webrtc/offer",
token,
{"sdp": peer.localDescription.sdp, "type": peer.localDescription.type},
)
await peer.setRemoteDescription(RTCSessionDescription(sdp=answer["sdp"], type=answer["type"]))
try:
await asyncio.wait_for(finished, timeout=180)
finally:
await peer.close()
collector.report()
async def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("--base", default="http://127.0.0.1:7860")
parser.add_argument("--text", default="Hello. How are you today? The doctor will help you now.")
parser.add_argument("--mode", choices=["websocket", "webrtc", "both"], default="both")
parser.add_argument("--token", default="")
arguments = parser.parse_args()
if arguments.mode in ("websocket", "both"):
await run_websocket_test(arguments.base, arguments.text, arguments.token)
if arguments.mode in ("webrtc", "both"):
await run_webrtc_test(arguments.base, arguments.text, arguments.token)
if __name__ == "__main__":
asyncio.run(main())