35 lines
814 B
Python
35 lines
814 B
Python
from pathlib import Path
|
|
|
|
import torch
|
|
|
|
from astrai.config.param_config import ModelParameter
|
|
from astrai.inference.generator import GenerationRequest, GeneratorFactory
|
|
|
|
PROJECT_ROOT = Path(__file__).resolve().parents[2]
|
|
PARAMETER_ROOT = Path(PROJECT_ROOT, "params")
|
|
|
|
|
|
def generate_text():
|
|
param = ModelParameter.load(PARAMETER_ROOT, disable_init=True)
|
|
param.to(device="cuda", dtype=torch.bfloat16)
|
|
|
|
query = input(">> ")
|
|
|
|
request = GenerationRequest(
|
|
query=query,
|
|
temperature=0.8,
|
|
top_p=0.95,
|
|
top_k=50,
|
|
max_len=param.config.max_len,
|
|
history=None,
|
|
system_prompt=None,
|
|
)
|
|
generator = GeneratorFactory.create(param, request)
|
|
response = generator.generate(request)
|
|
|
|
print(response)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
generate_text()
|