Skip to content

Instantly share code, notes, and snippets.

View atlury's full-sized avatar
💭
Working

Rahul Atlury atlury

💭
Working
View GitHub Profile
@atlury
atlury / sve2.md
Created April 17, 2025 07:03 — forked from zingaburga/sve2.md
ARM’s Scalable Vector Extensions: A Critical Look at SVE2 For Integer Workloads

ARM’s Scalable Vector Extensions: A Critical Look at SVE2 For Integer Workloads

Scalable Vector Extensions (SVE) is ARM’s latest SIMD extension to their instruction set, which was announced back in 2016. A follow-up SVE2 extension was announced in 2019, designed to incorporate all functionality from ARM’s current primary SIMD extension, NEON (aka ASIMD).

Despite being announced 5 years ago, there is currently no generally available CPU which supports any form of SVE (which excludes the [Fugaku supercomputer](https://www.fujitsu.com/global/about/innovation/

import os
import time
from pynput import keyboard
from datetime import datetime
import subprocess
import threading
import tkinter as tk
import queue
# ML imports
@atlury
atlury / Raspberry Pi 4 x 6 cluster 65B Llama runner gist
Created October 14, 2024 15:56 — forked from theycallmeloki/Raspberry Pi 4 x 6 cluster 65B Llama runner gist
an ansbile script that quickly bootstraps a Pi 4 cluster with tools required to run a 65B llama model
You will need the following two files, be sure to change the IP address in both sections, the `hostfile_contents`
as well as the ansible inventory section to reflect your configuration
Rest of the steps should run parallely across the different machines, for any updates to this script,
check this thread: https://github.com/ggerganov/llama.cpp/issues/2164#issuecomment-1639640846
If you want an automated way to do this in the future and would like an opionated setup procedure,
check this repo: https://github.com/theycallmeloki/edith-cli
Begin by enclosing all thoughts within <thinking> tags, exploring multiple angles and approaches.
Break down the solution into clear steps within <step> tags. Start with a 20-step budget, requesting more for complex problems if needed.
Use <count> tags after each step to show the remaining budget. Stop when reaching 0.
Continuously adjust your reasoning based on intermediate results and reflections, adapting your strategy as you progress.
Regularly evaluate progress using <reflection> tags. Be critical and honest about your reasoning process.
Assign a quality score between 0.0 and 1.0 using <reward> tags after each reflection. Use this to guide your approach:
0.8+: Continue current approach
0.5-0.7: Consider minor adjustments
Below 0.5: Seriously consider backtracking and trying a different approach
@atlury
atlury / qwen2vl.py
Created September 18, 2024 13:07 — forked from cavit99/qwen2vl.py
Qwen2-VL-7B-Instruct inference on Apple silicon
import torch
from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from PIL import Image
from pathlib import Path
import sys
# Toggle to switch between full response and extracted description
OUTPUT_FULL_RESPONSE = False
# Ensure we're using the MPS device if available
let rtcConnection = null;
let rtcLoopbackConnection = null;
let loopbackStream = new MediaStream(); // this is the stream you will read from for actual audio output
const offerOptions = {
offerVideo: true,
offerAudio: true,
offerToReceiveAudio: false,
offerToReceiveVideo: false,
};
@atlury
atlury / gist:0b397f444216fa7227ce0e1be79aa286
Created January 14, 2024 18:19 — forked from chrisdone/gist:02e165a0004be33734ac2334f215380e
Build and run minimal Linux / Busybox systems in Qemu

Common

export OPT=/opt
export BUILDS=/some/where/mini_linux
mkdir -p $BUILDS

Linux kernel

@atlury
atlury / saslauthd.conf
Created November 24, 2023 16:18 — forked from pbruna/saslauthd.conf
openLDAP Proxy with Meta Backend
# /etc/saslauthd.conf
ldap_servers: ldap://127.0.0.1
ldap_search_base: ou=%d,dc=local
ldap_timeout: 10
ldap_filter: (|(uid=%U)(SAMACCOUNTNAME=%U))
ldap_bind_dn: cn=Manager,dc=local
ldap_password: secret
ldap_deref: never
ldap_restart: yes
ldap_scope: sub
blueprint:
name: Frigate Notification (0.10.0)
description: |
## Frigate Mobile App Notification
This blueprint will send a notification to your device when a Frigate event for the selected camera is fired. The notification will initially include the thumbnail of the detection, but include an actionable notification allowing you to view the clip and snapshot.
With this blueprint, you may send the notification to multiple devices by leaving "Device" blank and instead use a [notification group][1].
### Software Version Requirements
@atlury
atlury / frigate_notification.yaml
Created July 16, 2022 17:23 — forked from hunterjm/frigate_notification.yaml
Frigate Notification Blueprint
blueprint:
name: Frigate Notification
description: |
## Frigate Mobile App Notification
This blueprint will send a notification to your device when a Frigate event for the selected camera is fired. The notification will initially include the thumbnail of the detection, but will update to include actionable notifications allowing you to view the saved clip/snapshot when available, or silence the notification for a configurable amount of time.
With this blueprint, you may send the notification to multiple devices by leaving "Device" blank and instead use a [notification group][1].
### Required entities: