handler.py

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

dtype = torch.bfloat16 if torch.cuda.get_device_capability()[0] == 8 else torch.float16

class EndpointHandler:
    def __init__(self, path=""):
        tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
        model = AutoModelForCausalLM.from_pretrained(
            path,
            return_dict=True,
            device_map="auto",
            load_in_8bit=True,
            torch_dtype=dtype,
            trust_remote_code=True,
        )

        self.generation_config = model.generation_config
        self.generation_config.max_new_tokens = 1000
        self.generation_config.temperature = 0.7  # Changed from 0 to 0.7
        self.generation_config.num_return_sequences = 1
        self.generation_config.pad_token_id = tokenizer.eos_token_id
        self.generation_config.eos_token_id = tokenizer.eos_token_id

        self.pipeline = transformers.pipeline(
            "text-generation", model=model, tokenizer=tokenizer
        )

    def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
        prompt = data.pop("inputs", data)
        result = self.pipeline(
            prompt,
            max_length=1000,  # Added this line to set max_length
            temperature=0.7,  # Added this line to set temperature
            top_p=0.9,  # Added this line to set top_p
            num_return_sequences=1,  # Added this line to set num_return_sequences
            pad_token_id=self.generation_config.pad_token_id,
            eos_token_id=self.generation_config.eos_token_id,
            return_full_text=True  # Added this line to return full text
        )
        return {"generated_text": result}