Upload folder using huggingface_hub

.gitignore

@@ -0,0 +1,160 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+.idea/

ChatWorld/ChatWorld.py

@@ -0,0 +1,84 @@
+from jinja2 import Template
+import torch
+
+from .models import qwen_model
+
+from .NaiveDB import NaiveDB
+from .utils import *
+
+
+class ChatWorld:
+    def __init__(self, pretrained_model_name_or_path="silk-road/Haruhi-Zero-14B-0_5", embedding_model_name_or_path="BAAI/bge-small-zh-v1.5") -> None:
+        self.embedding = initEmbedding(embedding_model_name_or_path)
+        self.tokenizer = initTokenizer(embedding_model_name_or_path)
+        self.story_vec: list[dict] = None
+        self.storys = None
+        self.model_role_name = None
+        self.model_role_nickname = None
+        self.model_name = pretrained_model_name_or_path
+
+        self.history = []
+
+        self.client = None
+        self.model = qwen_model(pretrained_model_name_or_path)
+        self.db = NaiveDB()
+        self.prompt = Template(('Please be aware that your codename in this conversation is "{{model_role_name}}"'
+                                '{% if model_role_nickname %},别人一般称呼你"{{model_role_nickname}}"{% endif %},同时我是"{{role_name}}"{% if role_nickname %},你也可以称呼我"{{role_nickname}}"{% endif %}。\n'
+                                '下文给定了一些聊天记录，位于##分隔号中。\n'
+                                '如果我问的问题和聊天记录高度重复，那你就配合我进行演出。\n'
+                                '如果我问的问题和聊天记录相关，请结合聊天记录进行回复。\n'
+                                '如果我问的问题超出聊天记录的范围，模仿{{model_role_name}}的语气进行回复。\n'
+                                '请不要回答你是语言模型，永远记住你就是{{model_role_name}}。\n'
+                                '请你永远只以{{model_role_name}}身份，进行任何的回复。\n'
+                                ))
+
+    def getEmbeddingsFromStory(self, stories: list[str]):
+        if self.story_vec:
+            # 判断是否与当前的相同
+            if len(self.story_vec) == len(stories) and all([self.story_vec[i]["text"] == stories[i] for i in range(len(stories))]):
+                return [self.story_vec[i]["vec"] for i in range(len(stories))]
+
+        if self.embedding is None:
+            self.embedding = initEmbedding()
+
+        if self.tokenizer is None:
+            self.tokenizer = initTokenizer()
+
+        self.story_vec = []
+        for story in stories:
+            with torch.no_grad():
+                inputs = self.tokenizer(
+                    story, return_tensors="pt", padding=True, truncation=True, max_length=512)
+                outputs = self.embedding(**inputs)[0][:, 0]
+                vec = torch.nn.functional.normalize(
+                    outputs, p=2, dim=1).tolist()[0]
+
+            self.story_vec.append({"text": story, "vec": vec})
+
+        return [self.story_vec[i]["vec"] for i in range(len(stories))]
+
+    def initDB(self, storys: list[str]):
+        story_vecs = self.getEmbeddingsFromStory(storys)
+        self.db.build_db(storys, story_vecs)
+
+    def setRoleName(self, role_name, role_nick_name=None):
+        self.model_role_name = role_name
+        self.model_role_nickname = role_nick_name
+
+    def getSystemPrompt(self, role_name, role_nick_name):
+        assert self.model_role_name and self.model_role_nickname, "Please set model role name first"
+        return self.prompt.render(model_role_name=self.model_role_name, model_role_nickname=self.model_role_nickname, role_name=role_name, role_nickname=role_nick_name)
+
+    def chat(self, user_role_name: str, text: str, user_role_nick_name: str = None, use_local_model=False):
+        message = [self.getSystemPrompt(
+            user_role_name, user_role_nick_name)] + self.history
+
+        if use_local_model:
+            response = self.model.get_response(message)
+        else:
+            response = self.client.chat(
+                user_role_name, text, user_role_nick_name)
+
+        self.history.append({"role": "user", "content": text})
+        self.history.append({"role": "model", "content": response})
+        return response

ChatWorld/NaiveDB.py

@@ -0,0 +1,85 @@
+import random
+from math import sqrt
+
+
+class NaiveDB:
+    def __init__(self):
+        self.verbose = False
+        self.init_db()
+
+    def init_db(self):
+        if self.verbose:
+            print("call init_db")
+        self.stories = []
+        self.norms = []
+        self.vecs = []
+        self.flags = []  # 用于标记每个story是否可以被搜索
+        self.metas = []  # 用于存储每个story的meta信息
+        self.last_search_ids = []  # 用于存储上一次搜索的结果
+
+    def build_db(self, stories, vecs, flags=None, metas=None):
+        self.stories = stories
+        self.vecs = vecs
+        self.flags = flags if flags else [True for _ in self.stories]
+        self.metas = metas if metas else [{} for _ in self.stories]
+        self.recompute_norm()
+
+    def save(self, file_path):
+        print(
+            "warning! directly save folder from dbtype NaiveDB has not been implemented yet, try use role_from_hf to load role instead")
+
+    def load(self, file_path):
+        print(
+            "warning! directly load folder from dbtype NaiveDB has not been implemented yet, try use role_from_hf to load role instead")
+
+    def recompute_norm(self):
+        # 补全这部分代码，self.norms 分别存储每个vector的l2 norm
+        # 计算每个向量的L2范数
+        self.norms = [sqrt(sum([x ** 2 for x in vec])) for vec in self.vecs]
+
+    def get_stories_with_id(self, ids):
+        return [self.stories[i] for i in ids]
+
+    def clean_flag(self):
+        self.flags = [True for _ in self.stories]
+
+    def disable_story_with_ids(self, close_ids):
+        for id in close_ids:
+            self.flags[id] = False
+
+    def close_last_search(self):
+        for id in self.last_search_ids:
+            self.flags[id] = False
+
+    def search(self, query_vector, n_results):
+
+        if self.verbose:
+            print("call search")
+
+        if len(self.norms) != len(self.vecs):
+            self.recompute_norm()
+
+        # 计算查询向量的范数
+        query_norm = sqrt(sum([x ** 2 for x in query_vector]))
+
+        idxs = list(range(len(self.vecs)))
+
+        # 计算余弦相似度
+        similarities = []
+        for vec, norm, idx in zip(self.vecs, self.norms, idxs):
+            if len(self.flags) == len(self.vecs) and not self.flags[idx]:
+                continue
+
+            dot_product = sum(q * v for q, v in zip(query_vector, vec))
+            if query_norm < 1e-20:
+                similarities.append((random.random(), idx))
+                continue
+            cosine_similarity = dot_product / (query_norm * norm)
+            similarities.append((cosine_similarity, idx))
+
+        # 获取最相似的n_results个结果， 使用第0个字段进行排序
+        similarities.sort(key=lambda x: x[0], reverse=True)
+        self.last_search_ids = [x[1] for x in similarities[:n_results]]
+
+        top_indices = [x[1] for x in similarities[:n_results]]
+        return top_indices

ChatWorld/__init__.py

@@ -0,0 +1 @@
+from .ChatWorld import ChatWorld

ChatWorld/models.py

@@ -0,0 +1,11 @@
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+
+class qwen_model:
+    def __init__(self, model_name):
+        self.tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+        self.model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", trust_remote_code=True).eval()
+
+    def get_response(self, message):
+        self.tokenizer.apply_chat_template(message, tokenize=False, add_generation_prompt=True)
+        return "test"

ChatWorld/utils.py

@@ -0,0 +1,9 @@
+from transformers import AutoModel, AutoTokenizer
+
+
+def initEmbedding(model_name="BAAI/bge-small-zh-v1.5", **model_wargs):
+    return AutoModel.from_pretrained(model_name, **model_wargs)
+
+
+def initTokenizer(model_name="BAAI/bge-small-zh-v1.5", **model_wargs):
+    return AutoTokenizer.from_pretrained(model_name)

README.md

@@ -1,13 +1,6 @@
 ---
 title: ChatWorld
-emoji: 😻
-colorFrom: indigo
-colorTo: yellow
-sdk: gradio
-sdk_version: 4.18.0
 app_file: app.py
-pinned: false
-license: apache-2.0
+sdk: gradio
+sdk_version: 3.50.2
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,65 @@
+import logging
+import os
+
+import gradio as gr
+
+from ChatWorld import ChatWorld
+
+logging.basicConfig(level=logging.INFO, filename="demo.log", filemode="w",
+                    format="%(asctime)s - %(name)s - %(levelname)-9s - %(filename)-8s : %(lineno)s line - %(message)s",
+                    datefmt="%Y-%m-%d %H:%M:%S")
+
+os.environ["HF_ENDPOINT"] = "https://hf-mirror.com"
+
+chatWorld = ChatWorld()
+
+
+def getContent(input_file):
+    # 读取文件内容
+    with open(input_file.name, 'r', encoding='utf-8') as f:
+        logging.info(f"read file {input_file.name}")
+        input_text = f.read()
+        logging.info(f"file content: {input_text}")
+
+    # 保存文件内容
+    input_text_list = input_text.split("\n")
+    chatWorld.initDB(input_text_list)
+    role_name_set = set()
+
+    # 读取角色名
+    for line in input_text_list:
+        role_name_set.add(line.split(":")[0])
+
+    role_name_list = [i for i in role_name_set if i != ""]
+    logging.info(f"role_name_list: {role_name_list}")
+
+    return gr.Radio(choices=role_name_list, interactive=True, value=role_name_list[0]), gr.Radio(choices=role_name_list, interactive=True, value=role_name_list[-1])
+
+
+def submit_message(message, history, model_role_name, role_name, model_role_nickname, role_nickname):
+    chatWorld.setRoleName(model_role_name, model_role_nickname)
+    response = chatWorld.chat(
+        role_name, message, role_nickname, use_local_model=True)
+    return response
+
+
+with gr.Blocks() as demo:
+
+    upload_c = gr.File(label="上传文档文件")
+
+    with gr.Row():
+        model_role_name = gr.Radio([], label="模型角色名")
+        model_role_nickname = gr.Textbox(label="模型角色昵称")
+
+    with gr.Row():
+        role_name = gr.Radio([], label="角色名")
+        role_nickname = gr.Textbox(label="角色昵称")
+
+    upload_c.upload(fn=getContent, inputs=upload_c,
+                    outputs=[model_role_name, role_name])
+
+    chatBox = gr.ChatInterface(
+        submit_message, chatbot=gr.Chatbot(height=400, render=False), additional_inputs=[model_role_name, role_name, model_role_nickname, role_nickname])
+
+
+demo.launch(share=True, debug=True, server_name="0.0.0.0")

requirements.txt

@@ -0,0 +1,4 @@
+gradio==4.19.2
+Jinja2==3.1.3
+torch==2.2.0
+transformers==4.38.1

run_gradio.sh

@@ -0,0 +1,5 @@
+export CUDA_VISIBLE_DEVICES=0
+export HF_HOME="/workspace/jyh/.cache/huggingface"
+
+# Start the gradio server
+/workspace/jyh/miniconda3/envs/ChatWorld/bin/python /workspace/jyh/Zero-Haruhi/app.py