Llama-2/llama2.py

# Copyright 2021 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
from typing import List

from huggingface_hub import hf_hub_download
from llama_cpp import Llama

from towhee.operator.base import PyOperator, SharedType


class LlamaCpp(PyOperator):
    '''Wrapper of Dolly inference'''
    def __init__(self,
                 model_name_or_file: str = 'llama-2-7b-chat',
                 **kwargs
                 ):
        self.kwargs = kwargs
        supported_model_names = self.supported_model_names()
        
        if model_name_or_file in supported_model_names:
            model_info = supported_model_names[model_name_or_file]
            hf_id = model_info['hf_id']
            model_filename = model_info['filename']
            self.model_path = hf_hub_download(repo_id=hf_id, filename=model_filename)
        else:
            self.model_path = model_name_or_file
        assert os.path.isfile(self.model_path), f'Invalid model path: {self.model_path}'

        init_kwargs = {}
        for k in vars(Llama).keys() & self.kwargs.keys():
            init_kwargs[k] = self.kwargs.pop(k)
        self.model = Llama(model_path=self.model_path, **init_kwargs)

    def __call__(self, messages: List[dict]):
        messages = self.parse_inputs(messages)
        resp = self.model.create_chat_completion(messages, **self.kwargs)
        answer = self.parse_outputs(resp)
        return answer
        
    def parse_inputs(self, messages: List[dict]):
        assert isinstance(messages, list), \
            'Inputs must be a list of dictionaries with keys from ["system", "question", "answer"].'
        new_messages = []
        for m in messages:
            for k, v in m.items():
                if k == 'system':
                    new_messages.append({'role': 'system', 'content': v})
                elif k == 'question':
                    new_messages.append({'role': 'user', 'content': v})
                elif k == 'answer':
                    new_messages.append({'role': 'assistant', 'content': v})
                else:
                    raise KeyError(f'Invalid key of message: {k}')
        return new_messages
    
    def parse_outputs(self, response):
        return response['choices'][0]['message']['content']

    @staticmethod
    def supported_model_names():
        models = {
            'llama-2-7b-chat': {
                'hf_id': 'TheBloke/Llama-2-7B-Chat-GGML',
                'filename': 'llama-2-7b-chat.ggmlv3.q4_0.bin'
                },
            'llama-2-13b-chat': {
                'hf_id': 'TheBloke/Llama-2-13B-chat-GGML',
                'filename': 'llama-2-13b-chat.ggmlv3.q4_0.bin'
            }
        }
        return models

    @property
    def shared_type(self):
        return SharedType.Shareable