LLM number 4

!pip install -q ipdb
!pip install -q tqdm

!pip install -q sentencepiece

!pip install -q wandb

and also pytorch, but this library is already istalled if we use Colab.

after we installed requirments, we need change runtime setting in colab. change from CPU to T4GPU

and open new notebook and import necessary library
#import library

import os, sys

import ipdb

import tqdm

from datetime import datetime

import platform, shutil

import requests, zipfile, io

#pytorch

import torch

import torch.nn as nn

from torch.nn import functional as F

#tokenizer

import sentencepiece as spm

#these improve performance for Ampere architecture

torch.backends.cuda.matmul.allow_tf32 = True

torch.backends.cudnn.allow_tf32 = True

#empty GPU cache memeory

torch.cuda.empty_cache()