UphamProjects

def create_look_ahead_mask(size, num_heads, special_token_indices, pronoun_special_tokens): mask = torch.triu(torch.ones((size, size)), diagonal=1).bool() mask = mask.unsqueeze(0).expand(num_heads, -1, -1) special_token_mask = torch.zeros(size, size, dtype=torch.bool) for idx in special_token_indices.values(): if idx != -1 and idx < size: # Check if the index is within…

import numpy as npimport scipy.fft as fftdef multi_head_attention_with_fft(Q, K, V, num_heads): # Split Q, K, and V into multiple heads Q_split = np.array_split(Q, num_heads, axis=-1) K_split = np.array_split(K, num_heads, axis=-1) V_split = np.array_split(V, num_heads, axis=-1) # Initialize lists to store results from each head head_results = [] for i in range(num_heads): # Apply FFT to split Q,…

You can find my colab here.

# -*- coding: utf-8 -*- """Copy of Copy of GAN and Stuff.ipynb Automatically generated by Colaboratory. Original file is located at https://colab.research.google.com/drive/12aatGMPZTv6CrAojSNXGnObFN1vd1KIN """ !pip install datasets import os import torch import torch.nn as nn import torch.optim as optim import torchvision.transforms as transforms from torchvision.datasets import MNIST from…

As I work through understanding llms I am creating various bots. I’ve just opened up the test server on discord. Feel free to meet and talk with some of my models. Only 1 is on right now and it’s smart as a bag of skittles. https://discord.gg/p3GXZJs8Ja

Meet my bot Evil Bob. Right now they can’t remember things too well and are wrong about things in general, but don’t let their poorly designed persona and model fool you. I’ll have multiple digital entities popping up on my site from time to time, these will be the Henchmen of Evil Bob or Sidekicks of Heroic Alice. I’m currently building their brains and have the tendency to lobotomize them…

This is an attempt at training a model. Its a start though. It does ok for summarizing things but is not up for general conversation. I’ll host it on discord once I get something resembling conversation going for it. Trying to understand what I’m doing with this still. The end goal is to get a lookup system across subjects, and feed blobs of predicate object to frame query from the original text.…

This is a simple image scraper that should work on most sites. Remember to check the sites robot text before implementing it. Made this after getting rejected for an identical task on Upwork. So here it is, as it is. import requests from bs4 import BeautifulSoup from urllib.parse import urlparse, urljoin from pathlib import Path def get_all_images(url, n,…

This is a simple script to create a new workbook and then tables from dictionary objects. First I define the dictionary objects as tableFields and then pass those to CreateNewWorkbookWithTables and iterate with a for loop nested twice. Sub Tables() Dim tableFields As Object Set tableFields = CreateObject("Scripting.Dictionary") tableFields("Clients") = Array("ID", "Name", "Address", "Phone",…

This is a start to a budgeting tool I started. This uses the python redisgraph library on a redis-stack server. The name == main block sets up the graph then demos Create Read Search Delete, and Sum but not update. I plan on expanding some schedule, and accrual methods. For good practice in methods add_line, remove_line, s_lines, and range_s_dates methods all use parameterized queries to prevent…

This is a start to a budgeting tool I have started. #LedgerInit.py import redis client = redis.Redis(host='localhost', port=6379, db=0) class Module: def __init__(self, name, properties, attributes, metadata): self.name = name self.properties = properties self.attributes = attributes self.metadata = metadata self.client = redis.Redis(host='localhost', port=6379, db=0) def…

Here is my first working attempt at ASR. I used the model wav2vec2-large-960h-lv60-self

This is my first working attempt at ASR. I didn’t train this model, it’s pre-trained and you can find it here. I had something else here that did not work, that sounded like a blender. The overall goal, which I have yet to reach, is live transcription, but at this point it’s a matter of repeating what I did below on a delay and loop. My next step though is making that weights not initialized…

This is part of a larger program I’m building. I’m using a discord server and bot as part of a productivity tool. I’ve had a few iterations of this so I don’t expect the final product to look anything like below. #vaa_bot.py import discord import EntityIO class DiscordBot: def __init__(self, token): self.token = token intents = discord.Intents.all() self.client =…

from graphviz import Digraph dot = Digraph(comment='Game Theory Model') dot.attr(rankdir='UD') dot.node('entity', label='Entities\n(Game Players)', shape='ellipse') dot.node('team', label='Teams\n(Parties)', shape='diamond') dot.node('objective', label='Objectives\n(Set of Actions towards a Desired/Forecast Goal/Outcome\nwith Results)', shape='box') dot.node('event', label='Events\n(Where…

Here is the lorem ipsum from reportlab.lib.pagesizes import letter from reportlab.pdfgen import canvas from reportlab.lib.utils import ImageReader from reportlab.lib.units import inch from lorem_text import lorem from PIL import Image, ImageDraw, ImageFont from PyPDF2 import PdfWriter, PdfReader from io import BytesIO def generate_pdf(flattened=False, num_pages=1,…

Tomorrow data type enforcement. #sap_schemas_template.py import redis redis_client = redis.StrictRedis(host='localhost', port=6379, db=0) data_structures_dict = { "PP": ["Material_Master", "Bill_of_Materials", "Work_Center", "Routing", "Production_Order"], "MM": ["Material_Master", "Vendor_Master", "Purchase_Order", "Goods_Receipt", "Inventory_Management", "Invoice_Verification"], "QM":…