import os
from typing import Any, Dict, Union

import torch
from modelscope.models.base import Model, TorchModel
from modelscope.models.builder import MODELS
from modelscope.pipelines.base import Pipeline
from modelscope.pipelines.builder import PIPELINES
from modelscope.utils.constant import Tasks
from modelscope.utils.logger import get_logger
from transformers import AutoModelForCausalLM, AutoTokenizer, LlamaForCausalLM

if 'CUDA_VISIBLE_DEVICES' not in os.environ:
    os.environ['CUDA_VISIBLE_DEVICES'] = '0'


@PIPELINES.register_module(Tasks.text_generation, module_name='ChatYuan-7B-text-generation-pipe')
class ChatYuan7BTextGenerationPipeline(Pipeline):
    def __init__(
            self,
            model: Union[Model, str],
            *args,
            **kwargs):
        model = ChatYuan7BTextGeneration(model) if isinstance(model, str) else model
        super().__init__(model=model, **kwargs)
    
    def preprocess(self, inputs, **preprocess_params) -> Dict[str, Any]:
        return inputs
    
    # define the forward pass
    def forward(self, inputs: Dict, **forward_params) -> Dict[str, Any]:
        return self.model(inputs)
    
    # format the outputs from pipeline
    def postprocess(self, input, **kwargs) -> Dict[str, Any]:
        return input



@MODELS.register_module(Tasks.text_generation, module_name='ChatYuan-7B')
class ChatYuan7BTextGeneration(TorchModel):
    def __init__(self, model_dir=None, *args, **kwargs):
        super().__init__(model_dir, *args, **kwargs)
        self.logger = get_logger()
        # loading tokenizer
        self.tokenizer = AutoTokenizer.from_pretrained(model_dir)
        self.model = LlamaForCausalLM.from_pretrained(model_dir, torch_dtype=torch.float16, device_map="auto")
        self.model = self.model.eval()
    
    def forward(self,input: Dict) -> Dict[str, Any]:
        output = {}
        res = self.infer(input)
        output['text'] = res
        return output
    
    def quantize(self, bits: int):
        self.model = self.model.quantize(bits)
        return self
    
    def infer(self, input):
        device = self.model.device

        input_ids = self.tokenizer(input, return_tensors="pt").input_ids.to(device)
        generate_ids = self.model.generate(input_ids, max_new_tokens=1024, do_sample = True, temperature = 0.7)
        output = self.tokenizer.batch_decode(generate_ids)[0]
        response = output[len(input):]

        return response