This Python tool extracts _kMDItemPhotoEmbedding vectors from macOS photos and indexes them into FAISS for fast similarity search.
It's a private macOS Spotlight metadata attribute that contains a 256-dimensional float32 embedding (1024 bytes) generated by the Apple Neural Scene Analyzer (ANSA) - a MobileNetV3-based model trained with image-language contrastive learning. This embedding powers features like Visual Look Up and photo categorization.
| import numpy as np | |
| from scipy.integrate import solve_ivp | |
| import matplotlib.pyplot as plt | |
| from mpl_toolkits.mplot3d import Axes3D | |
| from itertools import permutations | |
| from collections import Counter | |
| # Define the Thomas attractor system | |
| def thomas_derivatives(t, u, b): | |
| x, y, z = u |
| #!/usr/bin/env bash | |
| # based off of https://gist.github.com/rauchg/c5f0b1dc245ad95c593de8336aa382ac?permalink_comment_id=4842642#gistcomment-4842642 | |
| # further adapted from https://gist.github.com/algal/5e815ffd371d57d3a6a37056db1bd281 | |
| if [ "$#" -eq 0 ]; then | |
| echo "Usage: $(basename $0) promt_to_send_to_perplexity" | |
| echo "" | |
| echo " Requirements: PERPLEXITY_API, defined; jq and curl, installed; bash, version 3 or higher." | |
| exit 1 | |
| fi |
| from distilabel.steps import LoadDataFromDicts | |
| from distilabel.steps.tasks import EvolInstruct | |
| from distilabel.pipeline import Pipeline | |
| from distilabel.llms import OllamaLLM | |
| import pprint | |
| # Set up the pipeline | |
| with Pipeline(name="evol-instruct-example") as pipeline: | |
| # Load initial instructions |
| import modal | |
| app = modal.App(name="moondream-label-datikz_v2") | |
| data_dict = modal.Dict.from_name("HF_DATASET", create_if_missing=True) | |
| def download_dataset(): | |
| from datasets import load_dataset | |
| data_dict["HF_DATASET"] = "nllg/datikz-v2" | |
| dataset = load_dataset(data_dict["HF_DATASET"]) |
| import torch | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| import xxhash | |
| from tqdm import tqdm | |
| # load moondream model | |
| model_id = "vikhyatk/moondream2" | |
| revision = "2024-05-20" | |
| model = AutoModelForCausalLM.from_pretrained( |
| import modal | |
| app = modal.App(name="pokemon-eval") | |
| moondream_image = modal.Image.micromamba( | |
| python_version="3.11" | |
| ).apt_install( | |
| "git" | |
| ).micromamba_install( | |
| "cudatoolkit", |
| podman run --rm --device nvidia.com/gpu=all --security-opt=label=disable --mount type=bind,src="${PWD}",target=/workspace/axol | |
| otl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface winglian/axolotl:main-py3.10-cu118-2.0.1 accelerate launch -m axolotl.cli.train examples/openllam | |
| a-3b/lora.yml |
| #!/bin/bash | |
| git clone https://github.com/continuedev/continue | |
| npm config set prefix ~/.local | |
| sh continue/install-dependencies.sh | |
| cd continue/extensions/vscode | |
| node scripts/prepackage.js | |
| npm install @vscode/vsce | |
| PATH=$PATH:~/.local/bin/ | |
| npm install yarn | |
| node scripts/package.js |
| import torch | |
| print("pytorch version: " + torch.__version__) | |
| print("CUDA available: " + str(torch.cuda.is_available())) | |
| print("device count: " + str(torch.cuda.device_count())) | |
| print("current device: " + str(torch.cuda.current_device())) | |
| print("device name: " + torch.cuda.get_device_name(torch.cuda.current_device())) | |
| print("backend:") | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| print(device) |