basavyr

This code provides a helper for retrieving standard datasets from Torchvision, such as MNIST, CIFAR10, etc. It moreover provides the denormalization step, which is required before graphical representations of the image tensors.

from torch.utils.data import Dataset, DataLoader
from torchvision.transforms import Compose, ToTensor, Grayscale, Normalize, RandomResizedCrop
from torchvision.datasets import MNIST, CIFAR10, CIFAR100
import torch

from typing import Tuple
import os

basavyr

How to set plot annotations via plt.text() using percentages for the x-y location within the figure.

fig, ax = plt.subplots()

    # plt.plot(phi_1s, T1s, '-ro', label=r'$T_1-1axis$')
    plt.plot(phi_1s, T2s(x1_const), '-bo', label=r'$T_2-1axis$')
    plt.plot(phi_1s, T2s(5), '-ro', label=r'$T_2-1axis$')
    plt.plot(phi_1s, T2s(10), '-ko', label=r'$T_2-1axis$')
 plt.xlabel(r'$\varphi_1$')

basavyr

Use the code below to download and process the smaller variant of the Imagenet dataset.

Credit: https://github.com/basavyr/insta_exp/blob/main/src/utils.py#L204

def download_and_prepare_tiny_imagenet():
    base_name = "tiny-imagenet-200"
    url = f"http://cs231n.stanford.edu/{base_name}.zip"
    zip_path = os.path.join(DATASETS_DIR, f"{base_name}.zip")
    extract_path = os.path.join(DATASETS_DIR, base_name)

basavyr

Resources on GPU architecture

• A youtube channel (found on geohotz videos)
  • 14 GPU Architecture 1 - YouTube
• Documentation on CUDA architecture from Modal AI:
  • https://modal.com/
  • Full glossary with useful GPU specific terminology
  • What is a CUDA Device Architecture? | GPU Glossary
  • From George Hotz | how do GPUs work? (noob) + paper reading (not noob) | tinycorp.myshopify.com - YouTube
  • Warp scheduler: https://modal.com/gpu-glossary/device-hardware/warp-scheduler

basavyr

Formally, the mixing ratio can be regarded as the relative strength of the E2 component vs. the M1 component of the electroamgnetic transitions.

In other words, the mixing ratio will show the magnitude and character of the transition: $\delta=\frac{\langle I_f||E_2|| I_i \rangle}{\langle I_f||M_1|| I_i \rangle}$

Thus, this dimensionless quantity shows that if $|\delta | <1$ then the transition will have an M1 character, while an electrice quadrupole E2 character will exhibit when $|\delta|>1$.

For the context of wobbling motion, the mixing ratio will tell:

• how much of the wobbling angular momentum is perpendicular to the rotational axis (E2 component).
• how much is along the main rotation axis (M1 component)
• Theory predicts a large E2 component in the ΔI = 1 inter-band transitions.

basavyr

sudo apt update && sudo apt install -y neofetch htop tree vim nano tldr unzip nmap curl wget net-tools iputils-ping tmux zoxide ripgrep fzf

sudo apt update && sudo apt install -y zsh tmux tree jq bat fd-find ripgrep htop iotop net-tools nmon sysstat curl wget nmap traceroute mtr whois build-essential software-properties-common apt-transport-https ca-certificates git make cmake pkg-config zip unzip python3 python3-pip

sudo apt update && sudo apt install -y neofetch screenfetch inxi htop btop cmatrix cowsay fortune-mod lolcat figlet sysvbanner colorized-logs progress byobu ranger mc tldr zoxide fzf


for pyenv:

sudo apt update; sudo apt install build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev curl libncursesw5-dev xz-utils tk-dev libxml2-dev libxmlsec1-dev libffi-dev liblzma-dev

basavyr

import torch

import triton
import triton.language as tl

def unpack_weights(packed: torch.Tensor, bits: int = 2) -> torch.Tensor:
    values_per_item = 8 // bits
    packed_shape = packed.shape

    if len(packed_shape) == 1:

basavyr

import torch

import triton
import triton.language as tl

def unpack_weights(packed: torch.Tensor, bits: int = 2) -> torch.Tensor:
    values_per_item = 8 // bits
    packed_shape = packed.shape

    if len(packed_shape) == 1:

basavyr

Initial setup for a developer oriented macOS

In this guide we will focus on the necessary steps, settings, and tools that are required in order to have an environment that is streamlined for development, research, and overal productivity; all while keeping an emphasiz on distraction-free and minimalism

OS level

• We will always try to keep the operating system at the latest available version (stable releases only), so the very first step would be to make sure that our system has all the OS updates installed
• At the time of writing this guide, my macOS version is Version 14.5 (23F79)

Categories

basavyr

Executing Mathematica Scripts Within the Command-Line

This guide will show how to run Wolfram Mathematica scripts right from the command-line, without using the Frontend.

Setting up the console environment (MacOS)

Technically, a Mathematica script can be executed from the command line using the Mathematica Kernel.

After installation of the Wolfram Mathematica application, default location of that executable kernel should be at /Applications/Mathematica.app/Contents/MacOS: