handler.py

import warnings
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from typing import Any, Dict

class EndpointHandler:
    INSTRUCTION_KEY = "### Instruction:"
    RESPONSE_KEY = "### Response:"
    END_KEY = "### End"
    INTRO_BLURB = "Below is an instruction that describes a task. Write a response that appropriately completes the request."
    PROMPT_FOR_GENERATION_FORMAT = """{intro}
    {instruction_key}
    {instruction}
    {response_key}
    """.format(
        intro=INTRO_BLURB,
        instruction_key=INSTRUCTION_KEY,
        instruction="{instruction}",
        response_key=RESPONSE_KEY,
    )

    def __init__(
        self,
        path,
        torch_dtype=torch.bfloat16,
        trust_remote_code=True,
    ) -> None:
        self.model = AutoModelForCausalLM.from_pretrained(
            path,
            torch_dtype=torch_dtype,
            trust_remote_code=trust_remote_code
        )
        tokenizer = AutoTokenizer.from_pretrained(
            "mosaicml/mpt-7b-instruct",
            trust_remote_code=trust_remote_code
        )
        if tokenizer.pad_token_id is None:
            warnings.warn(
                "pad_token_id is not set for the tokenizer. Using eos_token_id as pad_token_id."
            )
            tokenizer.pad_token = tokenizer.eos_token

        tokenizer.padding_side = "right"
        self.tokenizer = tokenizer

        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        self.model.eval()
        self.model.to(device=self.device, dtype=torch_dtype)

        self.generate_kwargs = {
            "temperature": 0.01,
            "top_p": 0.92,
            "top_k": 0,
            "max_new_tokens": 512,
            "use_cache": True,
            "do_sample": True,
            "eos_token_id": self.tokenizer.eos_token_id,
            "pad_token_id": self.tokenizer.pad_token_id,
            "repetition_penalty": 1.0
        }

    def format_instruction(self, instruction):
        return self.PROMPT_FOR_GENERATION_FORMAT.format(instruction=instruction)

    def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
        # process input
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", None)

        # preprocess
        s = self.format_instruction(instruction=inputs)
        input_ids = self.tokenizer(s, return_tensors="pt").input_ids.to(self.device)
        gkw = {**self.generate_kwargs, **parameters}
        # pass inputs with all kwargs in data  
        with torch.no_grad():
            output_ids = self.model.generate(input_ids, **gkw)
        # Slice the output_ids tensor to get only new tokens
        new_tokens = output_ids[0, len(input_ids[0]) :]
        output_text = self.tokenizer.decode(new_tokens, skip_special_tokens=True)
        return [{"generated_text": output_text}]