chore: 添加虚拟环境到仓库

- 添加 backend_service/venv 虚拟环境
- 包含所有Python依赖包
- 注意：虚拟环境约393MB，包含12655个文件

backend_service/venv/lib/python3.13/site-packages/dashscope/tokenizers/__init__.py

@@ -0,0 +1,7 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+from .tokenization import Tokenization
+from .tokenizer import get_tokenizer, list_tokenizers
+from .tokenizer_base import Tokenizer
+
+__all__ = [Tokenization, Tokenizer, get_tokenizer, list_tokenizers]

backend_service/venv/lib/python3.13/site-packages/dashscope/tokenizers/qwen_tokenizer.py

@@ -0,0 +1,111 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+import base64
+import unicodedata
+from typing import Collection, Dict, List, Set, Union
+
+from .tokenizer_base import Tokenizer
+
+PAT_STR = r"""(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"""  # noqa E501
+ENDOFTEXT = '<|endoftext|>'
+IMSTART = '<|im_start|>'
+IMEND = '<|im_end|>'
+# as the default behavior is changed to allow special tokens in
+# regular texts, the surface forms of special tokens need to be
+# as different as possible to minimize the impact
+EXTRAS = tuple((f'<|extra_{i}|>' for i in range(205)))
+# changed to use actual index to avoid misconfiguration with vocabulary expansion
+SPECIAL_START_ID = 151643
+SPECIAL_TOKENS = tuple(
+    enumerate(
+        ((
+            ENDOFTEXT,
+            IMSTART,
+            IMEND,
+        ) + EXTRAS),
+        start=SPECIAL_START_ID,
+    ))
+SPECIAL_TOKENS_SET = set(t for i, t in SPECIAL_TOKENS)
+
+
+class QwenTokenizer(Tokenizer):
+    @staticmethod
+    def _load_tiktoken_bpe(tiktoken_bpe_file: str) -> Dict[bytes, int]:
+        with open(tiktoken_bpe_file, 'rb') as f:
+            contents = f.read()
+        return {
+            base64.b64decode(token): int(rank)
+            for token, rank in (line.split() for line in contents.splitlines()
+                                if line)
+        }
+
+    def __init__(self, vocab_file, errors='replace', extra_vocab_file=None):
+        self._errors = errors
+        self._vocab_file = vocab_file
+        self._extra_vocab_file = extra_vocab_file
+
+        self._mergeable_ranks = QwenTokenizer._load_tiktoken_bpe(
+            vocab_file)  # type: Dict[bytes, int]
+        self._special_tokens = {
+            token: index
+            for index, token in SPECIAL_TOKENS
+        }
+
+        # try load extra vocab from file
+        if extra_vocab_file is not None:
+            used_ids = set(self._mergeable_ranks.values()) | set(
+                self._special_tokens.values())
+            extra_mergeable_ranks = self._load_tiktoken_bpe(extra_vocab_file)
+            for token, index in extra_mergeable_ranks.items():
+                if token in self._mergeable_ranks:
+                    continue
+                if index in used_ids:
+                    continue
+                self._mergeable_ranks[token] = index
+            # the index may be sparse after this, but don't worry tiktoken.Encoding will handle this
+        import tiktoken
+        enc = tiktoken.Encoding(
+            'Qwen',
+            pat_str=PAT_STR,
+            mergeable_ranks=self._mergeable_ranks,
+            special_tokens=self._special_tokens,
+        )
+        assert (
+            len(self._mergeable_ranks) +
+            len(self._special_tokens) == enc.n_vocab
+        ), f'{len(self._mergeable_ranks) + len(self._special_tokens)} != {enc.n_vocab} in encoding'
+
+        self.decoder = {v: k
+                        for k, v in self._mergeable_ranks.items()
+                        }  # type: dict[int, bytes|str]
+        self.decoder.update({v: k for k, v in self._special_tokens.items()})
+
+        self._tokenizer = enc  # type: tiktoken.Encoding
+
+        self.eod_id = self._tokenizer.eot_token
+        self.im_start_id = self._special_tokens[IMSTART]
+        self.im_end_id = self._special_tokens[IMEND]
+
+    def encode(
+        self,
+        text: str,
+        allowed_special: Union[Set, str] = 'all',
+        disallowed_special: Union[Collection, str] = (),
+    ) -> Union[List[List], List]:
+        text = unicodedata.normalize('NFC', text)
+        return self._tokenizer.encode(text,
+                                      allowed_special=allowed_special,
+                                      disallowed_special=disallowed_special)
+
+    def decode(
+        self,
+        token_ids: Union[int, List[int]],
+        skip_special_tokens: bool = False,
+        errors: str = None,
+        **kwargs,
+    ) -> str:
+        if isinstance(token_ids, int):
+            token_ids = [token_ids]
+        if skip_special_tokens:
+            token_ids = [i for i in token_ids if i < self.eod_id]
+        return self._tokenizer.decode(token_ids, errors=errors or self._errors)

backend_service/venv/lib/python3.13/site-packages/dashscope/tokenizers/tokenization.py

@@ -0,0 +1,125 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+import copy
+from typing import Any, List
+
+from dashscope.api_entities.dashscope_response import (DashScopeAPIResponse,
+                                                       Message, Role)
+from dashscope.client.base_api import BaseApi
+from dashscope.common.constants import (CUSTOMIZED_MODEL_ID,
+                                        DEPRECATED_MESSAGE, HISTORY, MESSAGES,
+                                        PROMPT)
+from dashscope.common.error import InputRequired, ModelRequired
+from dashscope.common.logging import logger
+
+
+class Tokenization(BaseApi):
+    FUNCTION = 'tokenizer'
+    """API for get tokenizer result..
+
+    """
+    class Models:
+        """List of models currently supported
+        """
+        qwen_turbo = 'qwen-turbo'
+        qwen_plus = 'qwen-plus'
+        qwen_7b_chat = 'qwen-7b-chat'
+        qwen_14b_chat = 'qwen-14b-chat'
+        llama2_7b_chat_v2 = 'llama2-7b-chat-v2'
+        llama2_13b_chat_v2 = 'llama2-13b-chat-v2'
+        text_embedding_v2 = 'text-embedding-v2'
+        qwen_72b_chat = 'qwen-72b-chat'
+
+    @classmethod
+    def call(cls,
+             model: str,
+             input: Any = None,
+             prompt: Any = None,
+             history: list = None,
+             api_key: str = None,
+             messages: List[Message] = None,
+             workspace: str = None,
+             **kwargs) -> DashScopeAPIResponse:
+        """Call tokenization.
+
+        Args:
+            model (str): The requested model, such as qwen-v1
+            input: (Any): The model input body.
+            prompt (Any): The input prompt, for qwen serial model.
+            history (list):The user provided history,
+                deprecated, use messages instead.
+                examples:
+                    [{'user':'The weather is fine today.',
+                    'bot': 'Suitable for outings'}].
+                Defaults to None.
+            api_key (str, optional): The api api_key, can be None,
+                if None, will get by default rule(TODO: api key doc).
+            messages (list): The generation messages.
+                examples:
+                    [{'role': 'user',
+                      'content': 'The weather is fine today.'},
+                      {'role': 'assistant', 'content': 'Suitable for outings'}]
+            workspace (str): The dashscope workspace id.
+            **kwargs:
+                see model input.
+
+        Raises:
+            InputRequired: input is required.
+            ModelRequired: model is required.
+
+        Returns:
+            DashScopeAPIResponse: The tokenizer output.
+        """
+        if (input is None or not input) and \
+            (prompt is None or not prompt) and \
+                (messages is None or not messages):
+            raise InputRequired('prompt or messages is required!')
+        if model is None or not model:
+            raise ModelRequired('Model is required!')
+        if input is None:
+            input, parameters = cls._build_llm_parameters(
+                model, prompt, history, messages, **kwargs)
+        else:
+            parameters = kwargs
+
+        if kwargs.pop('stream', False):  # not support stream
+            logger.warning('streaming option not supported for tokenization.')
+
+        return super().call(model=model,
+                            task_group=None,
+                            function=cls.FUNCTION,
+                            api_key=api_key,
+                            input=input,
+                            is_service=False,
+                            workspace=workspace,
+                            **parameters)
+
+    @classmethod
+    def _build_llm_parameters(cls, model, prompt, history, messages, **kwargs):
+        parameters = {}
+        input = {}
+        if history is not None:
+            logger.warning(DEPRECATED_MESSAGE)
+            input[HISTORY] = history
+            if prompt is not None and prompt:
+                input[PROMPT] = prompt
+        elif messages is not None:
+            msgs = copy.deepcopy(messages)
+            if prompt is not None and prompt:
+                msgs.append({'role': Role.USER, 'content': prompt})
+            input = {MESSAGES: msgs}
+        else:
+            input[PROMPT] = prompt
+
+        if model.startswith('qwen'):
+            enable_search = kwargs.pop('enable_search', False)
+            if enable_search:
+                parameters['enable_search'] = enable_search
+        elif model.startswith('bailian'):
+            customized_model_id = kwargs.pop('customized_model_id', None)
+            if customized_model_id is None:
+                raise InputRequired('customized_model_id is required for %s' %
+                                    model)
+            input[CUSTOMIZED_MODEL_ID] = customized_model_id
+
+        return input, {**parameters, **kwargs}

backend_service/venv/lib/python3.13/site-packages/dashscope/tokenizers/tokenizer.py

@@ -0,0 +1,45 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+import os
+from typing import List
+
+from dashscope.common.error import UnsupportedModel
+from dashscope.tokenizers.qwen_tokenizer import QwenTokenizer
+
+from .tokenizer_base import Tokenizer
+
+QWEN_SERIALS = ['qwen-7b-chat', 'qwen-turbo', 'qwen-plus', 'qwen-max']
+current_path = os.path.dirname(os.path.abspath(__file__))
+root_path = os.path.dirname(current_path)
+
+
+def get_tokenizer(model: str) -> Tokenizer:
+    """Get a tokenizer based on model name.
+
+    Args:
+        model (str): The model name.
+
+    Raises:
+        UnsupportedModel: Not support model
+
+    Returns:
+        Tokenizer: The  `Tokenizer` of the model.
+    """
+    if model in QWEN_SERIALS:
+        return QwenTokenizer(
+            os.path.join(root_path, 'resources', 'qwen.tiktoken'))
+    elif model.startswith('qwen'):
+        return QwenTokenizer(
+            os.path.join(root_path, 'resources', 'qwen.tiktoken'))
+    else:
+        raise UnsupportedModel(
+            f'Not support model: {model}, currently only support qwen models.')
+
+
+def list_tokenizers() -> List[str]:
+    """List support models
+
+    Returns:
+        List[str]: The model list.
+    """
+    return QWEN_SERIALS

backend_service/venv/lib/python3.13/site-packages/dashscope/tokenizers/tokenizer_base.py

@@ -0,0 +1,32 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+from typing import List
+
+
+class Tokenizer:
+    """Base tokenizer interface for local tokenizers.
+    """
+    def __init__(self):
+        pass
+
+    def encode(self, text: str, **kwargs) -> List[int]:
+        """Encode input text string to token ids.
+
+        Args:
+            text (str): The string to be encoded.
+
+        Returns:
+            List[int]: The token ids.
+        """
+        pass
+
+    def decode(self, token_ids: List[int], **kwargs) -> str:
+        """Decode token ids to string.
+
+        Args:
+            token_ids (List[int]): The input token ids.
+
+        Returns:
+            str: The string of the token ids.
+        """
+        pass