From fc9bd43ea03719d38270ceb5bff2ca03979ffcc8 Mon Sep 17 00:00:00 2001 From: dlimeng Date: Fri, 29 Mar 2024 19:32:38 +0800 Subject: [PATCH] =?UTF-8?q?=E5=A4=A7=E6=A8=A1=E5=9E=8B?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Llama/LlamaDemo3.py | 52 +++++++++++++++++++++++++++ Llama/LlamaDemo4.py | 62 ++++++++++++++++++++++++++++++++ ai-generated-daily/2024-03-11.md | 29 +++++++++++++++ ai-generated-daily/2024-03-12.md | 0 ollama/demo.py | 36 +++++++++++++++++++ ollama/demo2.py | 2 ++ openai/demo.py | 41 +++++++++++++++++++++ 7 files changed, 222 insertions(+) create mode 100644 Llama/LlamaDemo3.py create mode 100644 Llama/LlamaDemo4.py create mode 100644 ai-generated-daily/2024-03-11.md create mode 100644 ai-generated-daily/2024-03-12.md create mode 100644 ollama/demo.py create mode 100644 ollama/demo2.py create mode 100644 openai/demo.py diff --git a/Llama/LlamaDemo3.py b/Llama/LlamaDemo3.py new file mode 100644 index 0000000..11fa833 --- /dev/null +++ b/Llama/LlamaDemo3.py @@ -0,0 +1,52 @@ +import os +from transformers import AutoTokenizer, AutoModelForCausalLM +from sentence_transformers import SentenceTransformer, util +import torch + +# 加载预训练模型的分词器和模型 +tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf") +model = AutoModelForCausalLM.from_pretrained( + "meta-llama/Llama-2-7b-chat-hf", + device_map='auto') + +# 加载HuggingFace的Embeddings模型 +embedder = SentenceTransformer('all-MiniLM-L6-v2') + +# 指定知识库文件目录 +knowledge_base_dir = "path/to/your/knowledge/base" + +# 读取目录中的所有文件，并将其内容作为知识库条目 +knowledge_base = [] +for filename in os.listdir(knowledge_base_dir): + filepath = os.path.join(knowledge_base_dir, filename) + if os.path.isfile(filepath): + with open(filepath, 'r', encoding='utf-8') as file: + knowledge_base.append(file.read()) + +# 将知识库文本转换为embeddings +knowledge_embeddings = embedder.encode(knowledge_base, convert_to_tensor=True) + +# 定义一个用户的问题或提示 +prompt = "请给我讲个玫瑰的爱情故事?" + +# 将用户的问题转换为embedding +question_embedding = embedder.encode(prompt, convert_to_tensor=True) + +# 在知识库中寻找最相似的文本 +cos_scores = util.pytorch_cos_sim(question_embedding, knowledge_embeddings)[0] +top_result = torch.topk(cos_scores, k=1) + +# 使用找到的最相关的文本作为模型的输入 +input_text = knowledge_base[top_result.indices[0]] + +# 使用分词器将找到的最相关的文本转化为模型可以理解的格式 +inputs = tokenizer(input_text, return_tensors="pt").to("cuda") + +# 使用模型生成文本 +outputs = model.generate(inputs["input_ids"], max_new_tokens=2000) + +# 将生成的令牌解码成文本 +response = tokenizer.decode(outputs[0], skip_special_tokens=True) + +# 打印生成的响应 +print(response) diff --git a/Llama/LlamaDemo4.py b/Llama/LlamaDemo4.py new file mode 100644 index 0000000..b89298b --- /dev/null +++ b/Llama/LlamaDemo4.py @@ -0,0 +1,62 @@ +from openai import OpenAI +from typing_extensions import override +from openai import AssistantEventHandler + + +client = OpenAI(api_key="") + + +assistant = client.beta.assistants.create( + name="Math Tutor", + instructions="You are a personal math tutor. Write and run code to answer math questions.", + tools=[{"type": "code_interpreter"}], + model="gpt-4-turbo-preview", +) + +thread = client.beta.threads.create() + +message = client.beta.threads.messages.create( + thread_id=thread.id, + role="user", + content="I need to solve the equation `3x + 11 = 14`. Can you help me?" +) + + + +# First, we create a EventHandler class to define +# how we want to handle the events in the response stream. + +class EventHandler(AssistantEventHandler): + @override + def on_text_created(self, text) -> None: + print(f"\nassistant > ", end="", flush=True) + + @override + def on_text_delta(self, delta, snapshot): + print(delta.value, end="", flush=True) + + def on_tool_call_created(self, tool_call): + print(f"\nassistant > {tool_call.type}\n", flush=True) + + def on_tool_call_delta(self, delta, snapshot): + if delta.type == 'code_interpreter': + if delta.code_interpreter.input: + print(delta.code_interpreter.input, end="", flush=True) + if delta.code_interpreter.outputs: + print(f"\n\noutput >", flush=True) + for output in delta.code_interpreter.outputs: + if output.type == "logs": + print(f"\n{output.logs}", flush=True) + + +# Then, we use the `create_and_stream` SDK helper +# with the `EventHandler` class to create the Run +# and stream the response. + +with client.beta.threads.runs.create_and_stream( + thread_id=thread.id, + assistant_id=assistant.id, + instructions="Please address the user as Jane Doe. The user has a premium account.", + event_handler=EventHandler(), +) as stream: + stream.until_done() \ No newline at end of file diff --git a/ai-generated-daily/2024-03-11.md b/ai-generated-daily/2024-03-11.md new file mode 100644 index 0000000..e450fb6 --- /dev/null +++ b/ai-generated-daily/2024-03-11.md @@ -0,0 +1,29 @@ +1.prompt策略遇上分治算法 + +论文地址：https://arxiv.org/pdf/2402.05359.pdf + +2.不依赖token，字节级模型，直接处理二进制 + +论文：https://arxiv.org/abs/2402.19155 + +代码：https://github.com/sanderwood/bgpt + +模型：https://huggingface.co/sander-wood/bgpt + +项目主页：https://byte-gpt.github.io + +3.马斯克宣布本周开源Grok + +https://x.com/elonmusk/status/1767108624038449405?s=20 + +4.RAG搜索复杂PDF文档 + +https://x.com/llama_index/status/1766862130761125958?s=20 + +5.几行代码微调LLM Agent + +https://x.com/khoomeik/status/1766805213644800011?s=20 + +6.使用LoRA技术微调Gemma-2B模型 + +https://huggingface.co/blog/gemma-peft diff --git a/ai-generated-daily/2024-03-12.md b/ai-generated-daily/2024-03-12.md new file mode 100644 index 0000000..e69de29 diff --git a/ollama/demo.py b/ollama/demo.py new file mode 100644 index 0000000..0161477 --- /dev/null +++ b/ollama/demo.py @@ -0,0 +1,36 @@ +from vanna.ollama import Ollama +from vanna.chromadb.chromadb_vector import ChromaDB_VectorStore + +class MyVanna(ChromaDB_VectorStore, Ollama): + def __init__(self, config=None): + ChromaDB_VectorStore.__init__(self, config=config) + Ollama.__init__(self, config=config) + + +vn = MyVanna(config={'model': 'gemma:7b'}) + +vn.connect_to_sqlite('my-database.sqlite') + +df_ddl = vn.run_sql("SELECT type, sql FROM sqlite_master WHERE sql is not null") + +for ddl in df_ddl['sql'].to_list(): + vn.train(ddl=ddl) + +vn.train(ddl=""" + CREATE TABLE IF NOT EXISTS my-table ( + id INT PRIMARY KEY, + name VARCHAR(100), + age INT + ) +""") + +vn.train( + documentation="Our business defines OTIF score as the percentage of orders that are delivered on time and in full") + +vn.train(sql="SELECT * FROM my-table WHERE name = 'John Doe'") + +training_data = vn.get_training_data() + +from vanna.flask import VannaFlaskApp + +VannaFlaskApp(vn).run() diff --git a/ollama/demo2.py b/ollama/demo2.py new file mode 100644 index 0000000..7db4ff1 --- /dev/null +++ b/ollama/demo2.py @@ -0,0 +1,2 @@ +import vanna +print(dir(vanna)) diff --git a/openai/demo.py b/openai/demo.py new file mode 100644 index 0000000..910eace --- /dev/null +++ b/openai/demo.py @@ -0,0 +1,41 @@ +from openai import OpenAI + +client = OpenAI( + # defaults to os.environ.get("OPENAI_API_KEY") + api_key="", + base_url="https://api.chatanywhere.tech/v1" +) + + + +# 非流式响应 +def gpt_35_api(messages: list): + """为提供的对话消息创建新的回答 + + Args: + messages (list): 完整的对话消息 + """ + completion = client.chat.completions.create(model="gpt-3.5-turbo", messages=messages) + print(completion.choices[0].message.content) + +def gpt_35_api_stream(messages: list): + """为提供的对话消息创建新的回答 (流式传输) + + Args: + messages (list): 完整的对话消息 + """ + stream = client.chat.completions.create( + model='gpt-3.5-turbo', + messages=messages, + stream=True, + ) + for chunk in stream: + if chunk.choices[0].delta.content is not None: + print(chunk.choices[0].delta.content, end="") + +if __name__ == '__main__': + messages = [{'role': 'user','content': '鲁迅和周树人的关系'},] + # 非流式调用 + # gpt_35_api(messages) + # 流式调用 + gpt_35_api_stream(messages) \ No newline at end of file