Spaces:
Running
on
Zero
Running
on
Zero
File size: 2,029 Bytes
6a83074 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 |
import argparse
import json
import requests
from llava_llama3.conversation import default_conversation
def main():
if args.worker_address:
worker_addr = args.worker_address
else:
controller_addr = args.controller_address
ret = requests.post(controller_addr + "/refresh_all_workers")
ret = requests.post(controller_addr + "/list_models")
models = ret.json()["models"]
models.sort()
print(f"Models: {models}")
ret = requests.post(controller_addr + "/get_worker_address",
json={"model": args.model_name})
worker_addr = ret.json()["address"]
print(f"worker_addr: {worker_addr}")
if worker_addr == "":
return
conv = default_conversation.copy()
conv.append_message(conv.roles[0], args.message)
prompt = conv.get_prompt()
headers = {"User-Agent": "LLaVA Client"}
pload = {
"model": args.model_name,
"prompt": prompt,
"max_new_tokens": args.max_new_tokens,
"temperature": 0.7,
"stop": conv.sep,
}
response = requests.post(worker_addr + "/worker_generate_stream", headers=headers,
json=pload, stream=True)
print(prompt.replace(conv.sep, "\n"), end="")
for chunk in response.iter_lines(chunk_size=8192, decode_unicode=False, delimiter=b"\0"):
if chunk:
data = json.loads(chunk.decode("utf-8"))
output = data["text"].split(conv.sep)[-1]
print(output, end="\r")
print("")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--controller-address", type=str, default="http://localhost:21001")
parser.add_argument("--worker-address", type=str)
parser.add_argument("--model-name", type=str, default="facebook/opt-350m")
parser.add_argument("--max-new-tokens", type=int, default=32)
parser.add_argument("--message", type=str, default=
"Tell me a story with more than 1000 words.")
args = parser.parse_args()
main()
|