File size: 1,416 Bytes
c8b5fa1 dc32044 c8b5fa1 dc32044 c8b5fa1 dc32044 c8b5fa1 dc32044 c8b5fa1 dc32044 c8b5fa1 dc32044 |
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 |
from typing import Dict, List, Any
import logging
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftConfig, PeftModel
LOGGER = logging.getLogger(__name__)
class EndpointHandler():
def __init__(self, path=""):
config = PeftConfig.from_pretrained(path)
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=True, device_map='auto')
self.tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
# Load the Lora model
self.model = PeftModel.from_pretrained(model, path)
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
"""
Args:
data (Dict): The payload with the text prompt and generation parameters.
"""
LOGGER.info(f"Received data: {data}")
# Get inputs
prompt = data.pop("prompt", data)
parameters = data.pop("parameters", None)
# Preprocess
inputs = self.tokenizer(prompt, return_tensors="pt")
# Forward
if parameters is not None:
outputs = self.model.generate(**inputs, **parameters)
else:
outputs = self.model.generate(**inputs)
# Postprocess
prediction = self.tokenizer.decode(outputs[0])
LOGGER.info(f"Generated text: {prediction}")
return [{"generated_text": prediction}]
|