From the uDemy course on LLM engineering.
https://www.udemy.com/course/llm-engineering-master-ai-and-large-language-models
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
310 lines
11 KiB
310 lines
11 KiB
{ |
|
"cells": [ |
|
{ |
|
"cell_type": "code", |
|
"execution_count": null, |
|
"id": "a0adab93-e569-4af0-80f1-ce5b7a116507", |
|
"metadata": {}, |
|
"outputs": [], |
|
"source": [ |
|
"# imports\n", |
|
"\n", |
|
"import os" |
|
] |
|
}, |
|
{ |
|
"cell_type": "code", |
|
"execution_count": null, |
|
"id": "9f583520-3c49-4e79-84ae-02bfc57f1e49", |
|
"metadata": {}, |
|
"outputs": [], |
|
"source": [ |
|
"# Creating a set of classes to simplify LLM use\n", |
|
"\n", |
|
"from abc import ABC, abstractmethod\n", |
|
"from dotenv import load_dotenv\n", |
|
"# Imports for type definition\n", |
|
"from collections.abc import MutableSequence\n", |
|
"from typing import TypedDict\n", |
|
"\n", |
|
"class LLM_Wrapper(ABC):\n", |
|
" \"\"\"\n", |
|
" The parent (abstract) class to specific LLM classes, normalising and providing common \n", |
|
" and simplified ways to call LLMs while adding some level of abstraction on\n", |
|
" specifics\n", |
|
" \"\"\"\n", |
|
"\n", |
|
" MessageEntry = TypedDict('MessageEntry', {'role': str, 'content': str})\n", |
|
" \n", |
|
" system_prompt: str # The system prompt used for the LLM\n", |
|
" user_prompt: str # The user prompt\n", |
|
" __api_key: str # The (private) api key\n", |
|
" temperature: float = 0.5 # Default temperature\n", |
|
" __msg: MutableSequence[MessageEntry] # Message builder\n", |
|
"\n", |
|
" def __init__(self, system_prompt:str, user_prompt:str, env_apikey_var:str=None):\n", |
|
" \"\"\"\n", |
|
" env_apikey_var: str # The name of the env variable where to find the api_key\n", |
|
" # We store the retrieved api_key for future calls\n", |
|
" \"\"\"\n", |
|
" self.system_prompt = system_prompt\n", |
|
" self.user_prompt = user_prompt\n", |
|
" if env_apikey_var:\n", |
|
" load_dotenv(override=True)\n", |
|
" self.__api_key = os.getenv(env_apikey_var)\n", |
|
"\n", |
|
" # # API Key format check\n", |
|
" # if env_apikey_var and self.__api_key:\n", |
|
" # print(f\"API Key exists and begins {self.__api_key[:8]}\")\n", |
|
" # else:\n", |
|
" # print(\"API Key not set\")\n", |
|
" \n", |
|
" def setSystemPrompt(self, prompt:str):\n", |
|
" self.system_prompt = prompt\n", |
|
"\n", |
|
" def setUserPrompt(self, prompt:str):\n", |
|
" self.user_prompt = prompt\n", |
|
"\n", |
|
" def setTemperature(self, temp:float):\n", |
|
" self.temperature = temp\n", |
|
"\n", |
|
" def getKey(self) -> str:\n", |
|
" return self.__api_key\n", |
|
"\n", |
|
" def messageSet(self, message: MutableSequence[MessageEntry]):\n", |
|
" self.__msg = message\n", |
|
"\n", |
|
" def messageAppend(self, role: str, content: str):\n", |
|
" self.__msg.append(\n", |
|
" {\"role\": role, \"content\": content}\n", |
|
" )\n", |
|
"\n", |
|
" def messageGet(self) -> MutableSequence[MessageEntry]:\n", |
|
" return self.__msg\n", |
|
" \n", |
|
" @abstractmethod\n", |
|
" def getResult(self):\n", |
|
" pass\n", |
|
"\n" |
|
] |
|
}, |
|
{ |
|
"cell_type": "code", |
|
"execution_count": null, |
|
"id": "a707f3ef-8696-44a9-943e-cfbce24b9fde", |
|
"metadata": {}, |
|
"outputs": [], |
|
"source": [ |
|
"from openai import OpenAI\n", |
|
"\n", |
|
"class GPT_Wrapper(LLM_Wrapper):\n", |
|
"\n", |
|
" MODEL:str = 'gpt-4o-mini'\n", |
|
" llm:OpenAI\n", |
|
"\n", |
|
" def __init__(self, system_prompt:str, user_prompt:str):\n", |
|
" super().__init__(system_prompt, user_prompt, \"OPENAI_API_KEY\")\n", |
|
" self.llm = OpenAI()\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"system\", \"content\": self.system_prompt},\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
"\n", |
|
" def setSystemPrompt(self, prompt:str):\n", |
|
" super().setSystemPrompt(prompt)\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"system\", \"content\": self.system_prompt},\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
" def setUserPrompt(self, prompt:str):\n", |
|
" super().setUserPrompt(prompt)\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"system\", \"content\": self.system_prompt},\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
" def getResult(self, format=None):\n", |
|
" \"\"\"\n", |
|
" format is sent as an adittional parameter {\"type\", format}\n", |
|
" e.g. json_object\n", |
|
" \"\"\"\n", |
|
" if format:\n", |
|
" response = self.llm.chat.completions.create(\n", |
|
" model=self.MODEL,\n", |
|
" messages=super().messageGet(),\n", |
|
" temperature=self.temperature,\n", |
|
" response_format={\"type\": \"json_object\"}\n", |
|
" )\n", |
|
" if format == \"json_object\":\n", |
|
" result = json.loads(response.choices[0].message.content)\n", |
|
" else:\n", |
|
" result = response.choices[0].message.content\n", |
|
" else:\n", |
|
" response = self.llm.chat.completions.create(\n", |
|
" model=self.MODEL,\n", |
|
" messages=super().messageGet(),\n", |
|
" temperature=self.temperature\n", |
|
" )\n", |
|
" result = response.choices[0].message.content\n", |
|
" return result" |
|
] |
|
}, |
|
{ |
|
"cell_type": "code", |
|
"execution_count": null, |
|
"id": "a8529004-0d6a-480c-9634-7d51498255fe", |
|
"metadata": {}, |
|
"outputs": [], |
|
"source": [ |
|
"import ollama\n", |
|
"\n", |
|
"class Ollama_Wrapper(LLM_Wrapper):\n", |
|
"\n", |
|
" MODEL:str = 'llama3.2'\n", |
|
"\n", |
|
" def __init__(self, system_prompt:str, user_prompt:str):\n", |
|
" super().__init__(system_prompt, user_prompt, None)\n", |
|
" self.llm=ollama\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"system\", \"content\": self.system_prompt},\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
"\n", |
|
" def setSystemPrompt(self, prompt:str):\n", |
|
" super().setSystemPrompt(prompt)\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"system\", \"content\": self.system_prompt},\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
" def setUserPrompt(self, prompt:str):\n", |
|
" super().setUserPrompt(prompt)\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"system\", \"content\": self.system_prompt},\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
" def getResult(self, format=None):\n", |
|
" \"\"\"\n", |
|
" format is sent as an adittional parameter {\"type\", format}\n", |
|
" e.g. json_object\n", |
|
" \"\"\"\n", |
|
" response = self.llm.chat(\n", |
|
" model=self.MODEL, \n", |
|
" messages=super().messageGet()\n", |
|
" )\n", |
|
" result = response['message']['content']\n", |
|
" return result" |
|
] |
|
}, |
|
{ |
|
"cell_type": "code", |
|
"execution_count": null, |
|
"id": "f25ffb7e-0132-46cb-ad5b-18a300a7eb51", |
|
"metadata": {}, |
|
"outputs": [], |
|
"source": [ |
|
"import anthropic\n", |
|
"\n", |
|
"class Claude_Wrapper(LLM_Wrapper):\n", |
|
"\n", |
|
" MODEL:str = 'claude-3-5-haiku-20241022'\n", |
|
" MAX_TOKENS:int = 200\n", |
|
" llm:anthropic.Anthropic\n", |
|
"\n", |
|
" def __init__(self, system_prompt:str, user_prompt:str):\n", |
|
" super().__init__(system_prompt, user_prompt, \"ANTHROPIC_API_KEY\")\n", |
|
" self.llm = anthropic.Anthropic()\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
" def setSystemPrompt(self, prompt:str):\n", |
|
" super().setSystemPrompt(prompt)\n", |
|
"\n", |
|
" def setUserPrompt(self, prompt:str):\n", |
|
" super().setUserPrompt(prompt)\n", |
|
" super().messageSet([\n", |
|
" {\"role\": \"user\", \"content\": self.user_prompt}\n", |
|
" ])\n", |
|
"\n", |
|
" def getResult(self, format=None):\n", |
|
" \"\"\"\n", |
|
" format is sent as an adittional parameter {\"type\", format}\n", |
|
" e.g. json_object\n", |
|
" \"\"\"\n", |
|
" response = self.llm.messages.create(\n", |
|
" model=self.MODEL,\n", |
|
" max_tokens=self.MAX_TOKENS,\n", |
|
" temperature=self.temperature,\n", |
|
" system=self.system_prompt,\n", |
|
" messages=super().messageGet()\n", |
|
" )\n", |
|
" result = response.content[0].text\n", |
|
" return result" |
|
] |
|
}, |
|
{ |
|
"cell_type": "code", |
|
"execution_count": null, |
|
"id": "4379f1c0-6eeb-4611-8f34-a7303546ab71", |
|
"metadata": {}, |
|
"outputs": [], |
|
"source": [ |
|
"import google.generativeai\n", |
|
"\n", |
|
"class Gemini_Wrapper(LLM_Wrapper):\n", |
|
"\n", |
|
" MODEL:str = 'gemini-1.5-flash'\n", |
|
" llm:google.generativeai.GenerativeModel\n", |
|
"\n", |
|
" def __init__(self, system_prompt:str, user_prompt:str):\n", |
|
" super().__init__(system_prompt, user_prompt, \"GOOGLE_API_KEY\")\n", |
|
" self.llm = google.generativeai.GenerativeModel(\n", |
|
" model_name=self.MODEL,\n", |
|
" system_instruction=self.system_prompt\n", |
|
" )\n", |
|
" google.generativeai.configure(api_key=super().getKey())\n", |
|
"\n", |
|
" def setSystemPrompt(self, prompt:str):\n", |
|
" super().setSystemPrompt(prompt)\n", |
|
"\n", |
|
" def setUserPrompt(self, prompt:str):\n", |
|
" super().setUserPrompt(prompt)\n", |
|
"\n", |
|
" def getResult(self, format=None):\n", |
|
" \"\"\"\n", |
|
" format is sent as an adittional parameter {\"type\", format}\n", |
|
" e.g. json_object\n", |
|
" \"\"\"\n", |
|
" response = self.llm.generate_content(self.user_prompt)\n", |
|
" result = response.text\n", |
|
" return result" |
|
] |
|
} |
|
], |
|
"metadata": { |
|
"kernelspec": { |
|
"display_name": "Python 3 (ipykernel)", |
|
"language": "python", |
|
"name": "python3" |
|
}, |
|
"language_info": { |
|
"codemirror_mode": { |
|
"name": "ipython", |
|
"version": 3 |
|
}, |
|
"file_extension": ".py", |
|
"mimetype": "text/x-python", |
|
"name": "python", |
|
"nbconvert_exporter": "python", |
|
"pygments_lexer": "ipython3", |
|
"version": "3.11.11" |
|
} |
|
}, |
|
"nbformat": 4, |
|
"nbformat_minor": 5 |
|
}
|
|
|