anotherjesse

package main

import (
	"fmt"
	"math/rand"
	"os"
	"os/signal"
)

const (

anotherjesse

seed: 0 output: The Complete Book of Modern Paisley (Paisley Collection)
seed: 1 output: The Complete Idiot's Guide to Doing Nothing
seed: 2 output: The Complete Guide to Cat Care
seed: 3 output: The Complete Idiot's Guide to Sewing, Stitching, Needlepoint, and More
seed: 4 output: The Complete Idiot's Guide to Tennis: Everything You Need to Know About Tennis, One Game at a Time
seed: 5 output: The Complete Guide to Building a Website (The Complete Guides)
seed: 6 output: The Complete Idiot's Guide to Baking
seed: 7 output: The Complete Guide to Bicycle Repair
seed: 8 output: The Complete Idiot's Guide to Gardening
seed: 9 output: Big Bird Singing Classic

anotherjesse

def generate(prompt, filename):
    # Simulate the generation process (you can replace this with your actual function)
    print(f"Generating '{prompt}' and saving as '{filename}'")

# Options for each category
expressions = ["happy", "sad", "crying", "angry", "open mouth"]
genders = ["male", "female", "non-binary"]
ages = ["kid", "teen", "young adult", "adult", "old"]
ethnicities = ["Caucasian", "Hispanic", "Black", "Middle-Eastern", "South Asian", "East Asian"]
styles = ["realistic", "cartoon", "comics"]

anotherjesse

A first pass at creating a @langchain vectorstore for each tweet I have "liked"

I already store them as a json blob on https://lets.m4ke.org/tweets

This ingests them documents - each looking like @twitter_user_name says tweet_text_here

Samples:

>>> qa.run("what is @jakedahn tweeting about")

' @jakedahn is tweeting about art and creativity.'

anotherjesse

polling

model = replicate.models.get("replicate/hello-world")
version = model.versions.get("5c7d5dc6dd8bf75c1acaa8565735e7986bc5b66206b55cca93cb72c9bf15ccaa")
inputs = {
  "text": "world",
}
output = version.predict(**inputs)

webhook

anotherjesse

import json
import random


db = {}


def get_data(kind, count=1):
    if kind not in db:
        with open("clip_interrogator/data/{}.txt".format(kind)) as f:

anotherjesse

big picture

Build your docker image in a way that works with replicate, while limiting image "layers" to less than 20GB.

Steps

Build COG without weights

First build a version of the cog without weights. To do this ensure remove extra files/weights from your cog directory and then build.

anotherjesse

!pip install diffusers==0.4.0
!pip install transformers  ftfy

from transformers import CLIPTokenizer, CLIPTextModel
import torch

torch_device = "cuda" if torch.cuda.is_available() else "cpu"  # or just let it be cpu

prompt = ["a photograph of an astronaut riding a horse"]

anotherjesse

import React, { useState } from "react";
import ReactDOM from "react-dom";
import { extend, Canvas } from "react-three-fiber";
import DatGui, { DatColor, DatNumber, DatSelect } from "react-dat-gui";
import { Text } from "troika-three-text";
import fonts from "./fonts";

import "react-dat-gui/build/react-dat-gui.css";
import "./styles.css";

anotherjesse

using golang to create image pipe for ffmpeg

If you want to stream a slideshow using ffmpeg, it seems like one easy way to do it is to create an imagepipe of the files.

This can then be streamed to twitch.tv or youtube ... https://trac.ffmpeg.org/wiki/StreamingGuide

This sample cycles between images on disk

usage