Skip to content

ai.md

Latest commit

 

History

History
291 lines (228 loc) · 9.73 KB

ai.md

File metadata and controls

291 lines (228 loc) · 9.73 KB
Raw

AI

LLMs, or Large Language Models, are advanced AI systems designed to understand and generate human-like
text based on vast amounts of data. They are trained on diverse text sources, enabling them to perform
a variety of tasks such as:

  • Natural Language Understanding: Comprehending and interpreting text to derive meaning.
  • Text Generation: Creating coherent and contextually appropriate text based on a given prompt.
  • Conversation: Engaging in human-like dialogue, answering questions, and carrying out conversations.
  • Summarization: Condensing long texts into brief summaries while retaining key information.
  • Translation: Converting text from one language to another with high accuracy.

LLMs have applications in chatbots, content creation, language translation, and much more,
revolutionizing the way we interact with technology.

Models

Name Date of Publishing Country of Origin Model Category
ChatGPT November 2022 USA General-purpose LLM
GPT-4 March 2023 USA Advanced multi-modal LLM
DeepSeek R1 January 2025 China General-purpose LLM
Mistral 7B September 2023 France Open-weight LLM
Claude 3.5 June 2024 USA Safety-focused LLM
Grok-1 November 2023 USA Open-source LLM
PaLM 2 May 2023 USA Multilingual LLM
Falcon 180B September 2023 UAE Open-weight LLM
Gemini 1.5 February 2024 USA Multimodal LLM
Llama 2 July 2023 USA Open-source LLM
Command R 2024 Canada Retrieval-augmented LLM
Phi-2 December 2023 USA Small-scale LLM
GPT-Neo March 2021 USA Open-source LLM
BERT October 2018 USA Bidirectional Encoder
LaMDA May 2021 USA Conversational LLM

Web interfaces

  • https://chatgpt.com - recent leader in AI
  • https://venice.ai - focus on privacy
  • https://playground.allenai.org/ - open source, academic LLM
  • https://openrouter.ai - connecting multiple LLM services, including free ones
  • https://www.perplexity.ai - deep research and instant answers
  • https://abacus.ai - one superasistant
  • https://console.groq.com - generous, free access to top LLMs
  • https://chat.deepseek.com - excellent Chinese model
  • https://chat.qwenlm.ai/ - Chinese model from Alibaba
  • https://github.com/cheahjs/free-llm-api-resources - free API resources

Simple chat

from ollama import chat

# Define the input message
message = {
    "model": "phi4",
    "messages": [{"role": "user", "content": "Is pluto a planet?"}]
}

# Use the Ollama chat function
response = chat(model=message["model"], messages=message["messages"])

# Print the response
print(response["message"]["content"])

Ollama vision

import ollama

response = ollama.chat(
    model='llama3.2-vision',
    messages=[{
        'role': 'user',
        'content': 'What is in this image?',
        'images': ['image2.jpg']
    }]
)

print(response)

List models

list local models

from ollama import list
from ollama import ListResponse

response: ListResponse = list()

for model in response.models:
    print('Name:', model.model)
    print('  Size (MB):', f'{(model.size.real / 1024 / 1024):.2f}')
    if model.details:
        print('  Format:', model.details.format)
        print('  Family:', model.details.family)
        print('  Parameter Size:', model.details.parameter_size)
        print('  Quantization Level:', model.details.quantization_level)
    print('\n')

Ollama via OpenAI API

Connect to the Ollama local models via OpenAI API, which became industry standard.

import openai

client = openai.OpenAI(
    base_url="http://localhost:11434/v1",
    api_key="nokeyneeded",
)

response = client.chat.completions.create(
    model="deepseek-r1",
    temperature=0.7,
    n=1,
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Write a haiku about a hungry cat"},
    ],
)

print("Response:")
print(response.choices[0].message.content)

Streaming example

from ollama import chat

prompt = "When was Alien movie released?"
for chunk in chat(model="deepseek-r1", messages=[{"role": "user", "content": prompt}], stream=True):
    print(chunk['message']['content'], end='', flush=True)

Sentiment analysis with HuggingFace

from transformers import AutoModelForSequenceClassification
from transformers import TFAutoModelForSequenceClassification
from transformers import AutoTokenizer, AutoConfig
import numpy as np
from scipy.special import softmax

# Preprocess text (username and link placeholders)


def preprocess(text):
    new_text = []
    for t in text.split(" "):
        t = '@user' if t.startswith('@') and len(t) > 1 else t
        t = 'http' if t.startswith('http') else t
        new_text.append(t)
    return " ".join(new_text)


MODEL = f"cardiffnlp/twitter-roberta-base-sentiment-latest"
tokenizer = AutoTokenizer.from_pretrained(MODEL)
config = AutoConfig.from_pretrained(MODEL)
# PT
model = AutoModelForSequenceClassification.from_pretrained(MODEL)
# model.save_pretrained(MODEL)
# text = "Covid cases are increasing fast!"
text = "That was a trash movie."
text = preprocess(text)

encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)
scores = output[0][0].detach().numpy()
print(scores)
scores = softmax(scores)
print(scores)
# # TF
# model = TFAutoModelForSequenceClassification.from_pretrained(MODEL)
# model.save_pretrained(MODEL)
# text = "Covid cases are increasing fast!"
# encoded_input = tokenizer(text, return_tensors='tf')
# output = model(encoded_input)
# scores = output[0][0].numpy()
# scores = softmax(scores)
# Print labels and scores
ranking = np.argsort(scores)
ranking = ranking[::-1]

for i in range(scores.shape[0]):
    l = config.id2label[ranking[i]]
    s = scores[ranking[i]]
    print(f"{i+1}) {l} {np.round(float(s), 4)}")

OpenRouter

from openai import OpenAI
import os

client = OpenAI(
  base_url="https://openrouter.ai/api/v1",
  api_key=os.environ["OPENROUTER_API_KEY"],
)

completion = client.chat.completions.create(
  model="deepseek/deepseek-r1:free",
  messages=[
    {
      "role": "user",
      "content": "What is the meaning of life?"
    }
  ]
)

print(completion.choices[0].message.content)

Sentiment analysis

from openai import OpenAI

from pathlib import Path
import os
import time


client = OpenAI(
    base_url="https://api.deepseek.com",
    api_key=os.environ.get("DEEPSEEK_API_KEY"),
)

# movie_reviews = {
#     1: "The storyline was absolutely captivating, and the performances were brilliant. I couldn't look away for a second!",
#     2: "The pacing was excruciatingly slow, and the characters lacked depth. I was bored halfway through.",
#     3: "While the visuals were breathtaking, the plot felt predictable and uninspired.",
#     4: "This is a cinematic masterpiece that touched my heart. Every scene was perfection!",
#     5: "The dialogue was cringe-worthy, and the humor fell flat. Definitely not worth the hype.",
#     6: "It was an average film—not great, but not terrible either. I enjoyed some parts.",
#     7: "The chemistry between the leads was electric, and the soundtrack was phenomenal!",
#     8: "The movie started strong but completely fell apart in the second half. Such a disappointment.",
#     9: "A visually stunning film that combines action and emotion seamlessly. Highly recommend!",
#     10: "The premise was intriguing, but the execution left a lot to be desired. It just didn't click for me."
# }

slovak_movie_reviews = {
    1: "Príbeh bol úplne pútavý a herecké výkony brilantné. Nemohol som sa odtrhnúť ani na sekundu!",
    2: "Tempo bolo mimoriadne pomalé a postavy nemali žiadnu hĺbku. Nudil som sa už v polovici.",
    3: "Hoci vizuálne efekty boli ohromujúce, dej pôsobil predvídateľne a bez inšpirácie.",
    4: "Toto je filmové dielo, ktoré mi dojalo srdce. Každá scéna bola dokonalosť!",
    5: "Dialógy boli trápne a humor úplne zlyhal. Určite to nestojí za ten humbug.",
    6: "Bol to priemerný film – nie dobrý, ale ani úplná katastrofa. Niektoré časti ma bavili.",
    7: "Chemia medzi hlavnými postavami bola elektrizujúca a soundtrack fenomenálny!",
    8: "Film začal skvele, ale v druhej polovici sa úplne rozpadol. Veľké sklamanie.",
    9: "Vizualne ohromujúci film, ktorý dokonale spája akciu a emócie. Určite odporúčam!",
    10: "Premisa bola zaujímavá, ale realizácia bola slabá. Nedokázalo ma to zaujať."
}

for key, value in slovak_movie_reviews.items():

    # content = 'On a scale 0-1, figure out the sentiment of the the following movie review:'
    content = 'Na škále od 0 do 1, napíš sentiment nasledujúceho filmu:'
    content += value

    chat_completion = client.chat.completions.create(
        messages=[
            {
                "role": "user",
                "content": content,
            }
        ],
        temperature=0.7,
        top_p=0.9,
        model='deepseek-chat',
        max_completion_tokens=1000
    )

    # print(chat_completion.choices[0].message.content)
    output = chat_completion.choices[0].message.content
    print(key, value, output)