| import torch | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| from typing import Dict, List, Any | |
| class EndpointHandler(): | |
| def __init__(self, path=""): | |
| path = "tiiuae/falcon-40b" | |
| self.model = AutoModelForCausalLM.from_pretrained(path, | |
| torch_dtype=torch.bfloat16, | |
| device_map="auto", | |
| load_in_8bit=True, | |
| trust_remote_code=True) | |
| self.tokenizer = AutoTokenizer.from_pretrained(path) | |
| self.device = "cuda" if torch.cuda.is_available() else "cpu" | |
| def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]: | |
| input_text = data.pop("inputs", data) | |
| inputs = self.tokenizer(input_text, | |
| return_tensors="pt") | |
| input_ids = inputs.input_ids.to(self.device) | |
| attention_mask = inputs.attention_mask.to(self.device) | |
| score = self.model(input_ids=input_ids, | |
| attention_mask=attention_mask, | |
| labels=input_ids).loss.item() | |
| return score | |