George Chiramattel georgeck

Comparison of Nvidia GPUs for Fine-Tuning

Key Considerations for Fine-Tuning:

VRAM Capacity: Determines the maximum size of the model and the batch size you can use. Running out of VRAM is a common bottleneck.
Memory Bandwidth: How quickly data (model parameters, activations, gradients) can be moved between the GPU's memory and its compute units. High bandwidth is essential for keeping the powerful cores fed, especially with large models.
Compute Performance (FLOPS/TOPS): Raw processing power. Modern fine-tuning heavily relies on Tensor Cores for mixed-precision training (FP16, BF16, TF32, and increasingly FP8 on newer architectures like Hopper, Ada Lovelace, and Blackwell).
Architecture: Newer architectures (Blackwell > Hopper > Ada Lovelace) generally offer better performance per watt, more efficient Tensor Cores, and support for newer data formats (like FP8).
Target Environment: Data Center cards (B200, H200, H100, L40S, L4) are built for 24/7 operation, of

conjure-up

conjure-up -> Lists the available "spell" conjure-up lets you summon up a big-software stack as a "spell"

snap

A universal app store for Linux

snap list
sudo snap install hello
sudo snap refresh hello
udo snap remove hello

Java Data Time

In modern Java (after Java 8), we use the pacakage java.time.* to represent DataTime. This is much better than the java.util classes. For interoperability, we have methods in the java.util.* package, that can convert from the older java.util.Date to the newer types.

Instant

An instantaneous point on the time-line (with a precision of nano seconds). It models a single instantaneous point on the time-line. This might be used to record event time-stamps in the application. The epoch-seconds are measured from the standard Java epoch of 1970-01-01T00:00:00Z where instants after the epoch have positive values, and earlier instants have negative values.

Duration

This models a quantity or amount of time in terms of seconds and nanoseconds. It can be accessed using other duration-based units, such as minutes and hours. In addition, the DAYS unit can be used and is treated as exactly equal to 24 hours, thus ignoring daylight savings effects. Durations can also be negative.

Git cheat list

name of the current banch and nothing else (for automation)
```
git rev-parse --abbrev-ref HEAD
```
all commits that your branch have that are not yet in master
```
git log master..<HERE_COMES_YOUR_BRANCH_NAME>
```

	<!DOCTYPE html>
	<html lang="en">
	<head>
	<meta charset="UTF-8">
	<meta name="viewport" content="width=device-width, initial-scale=1.0">
	<meta name="description" content="A collection of my saved bookmarks">
	<title>My Bookmarks</title>
	<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/tailwindcss/2.2.19/tailwind.min.css" integrity="sha512-wnea99uKIC3TJF7v4eKk4Y+lMz2Mklv18+r4na2Gn1abDRPPOeef95xTzdwGD9e6zXJBteMIhZ1+68QC5byJZw==" crossorigin="anonymous" referrerpolicy="no-referrer" />
	<link rel="icon" href="data:image/svg+xml,<svg xmlns=%22http://www.w3.org/2000/svg%22 viewBox=%220 0 100 100%22><text y=%22.9em%22 font-size=%2290%22>📚</text></svg>">
	<style>

	// Uses the eventsource-parser library to parse the stream from OpenAI's chat completion API
	// Make sure to set the OPENAI_API_KEY environment variable to your OpenAI API key

	import {createParser} from 'eventsource-parser'

	function generatePrompt(prompt) {
	return [
	{"role": "system", "content": "The following is a conversation with an AI assistant. The assistant is helpful, creative, clever, and very friendly."},
	{"role": "user", "content": "Hello, who are you?"},
	{"role": "system", "content": `I am an AI created by OpenAI. How can I help you today?`},

	// Creating a promise
	let promise = new Promise (
	(resolve, reject) => {
	// Some function that takes time - in this case, retuen afer 1 sec
	setTimeout( () => {
	if ( Math.random() > .5) {
	resolve('😀');
	} else {
	reject('😡');
	}

	apiVersion: v1
	kind: ReplicationController
	metadata:
	name: hello-rc
	spec:
	replicas: 10
	selector:
	app: hello-world
	template:
	metadata:

	apiVersion: v1
	kind: Pod
	metadata:
	name: hello-pod
	spec:
	containers:
	- name: hello-ctr
	image: nigelpoulton/pluralsight-docker-ci:latest
	ports:
	- containerPort: 8080

	[core]
	editor = 'C:/Program Files (x86)/Notepad++/notepad++.exe' -multiInst -notabbar -nosession -noPlugin
	[alias]
	st = status
	ci = commit
	br = branch
	co = checkout
	df = diff
	dc = diff --cached
	lg = log -p