AIAnytime

import streamlit as st
import fastai
from deoldify.visualize import *

def image_colorizer(url):
    colorizer = get_image_colorizer(artistic=True)
    source_url = url
    render_factor = 35
    watermarked = True 

AIAnytime

# PyTube function for YouTube video
def save_audio(url):
    yt = YouTube(url)
    video = yt.streams.filter(only_audio=True).first()
    out_file = video.download()
    base, ext = os.path.splitext(out_file)
    file_name = base + '.mp3'
    try:
        os.rename(out_file, file_name)
    except WindowsError:

AIAnytime

# Assembly AI speech to text
def assemblyai_stt(audio_filename):
    with open(audio_filename , "rb") as f:
        response = requests.post(base_url + "/upload",
                                headers=headers,
                                data=f)

    upload_url = response.json()["upload_url"]
    data = {
        "audio_url": upload_url

AIAnytime

import streamlit as st
import os
from tortoise.models.classifier import AudioMiniEncoderWithClassifierHead
from glob import glob
import io
import librosa
import plotly.express as px
import torch
import torch.nn.functional as F
import torchaudio

AIAnytime

Nucleus sampling is a technique used in large language models to control the randomness and diversity of generated text. It works by sampling from only the most likely tokens in the model’s predicted distribution.

The key parameters are:

Temperature: Controls randomness, higher values increase diversity.

Top-p (nucleus): The cumulative probability cutoff for token selection. Lower values mean sampling from a smaller, more top-weighted nucleus.

Top-k: Sample from the k most likely next tokens at each step. Lower k focuses on higher probability tokens.

AIAnytime

import torch
from transformers import BertTokenizer, BertModel

tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
model = BertModel.from_pretrained('bert-base-uncased')

def encode(text):
    tokens = tokenizer.tokenize(text)
    tokens = ['[CLS]'] + tokens + ['[SEP]']
    return tokenizer.convert_tokens_to_ids(tokens)

AIAnytime

Hello everyone, I am Sneha Roy, working as a full-stack AI Engineer and recently I have been working on Generative AI projects. 
Today, I will be showcasing a solution that we have developed.........."GCC KB Chatbot". 
it is an intelligent assistant tailored to simplify how call centers operate, in this case.. the GCC team.

GCC faces a significant challenge managing a high volume of tickets daily. 
Each of these tickets is needing assistance, and each minute spent resolving one is critical. 
Traditionally, our colleagues at GCC relied on a comprehensive but cumbersome Knowledge Base (KB) to find solutions, 
which significantly contributed to longer turnaround times.

That's where our solution comes in...  it drastically reduces the time taken to resolve tickets. Acting as a virtual agent, 

AIAnytime

import os
#set you runpod key as a environment variable
os.environ['RUNPOD_API_KEY'] = "your_runpod_api_key"

import runpod
from IPython.display import display, Markdown

runpod.api_key = os.getenv("RUNPOD_API_KEY", "your_runpod_api_key")

if runpod.api_key == "your_runpod_api_key":

AIAnytime

## With Evaluation Harness
!pip install git+https://github.com/EleutherAI/lm-evaluation-harness.git
!pip install bitsandbytes
!pip install --upgrade transformers
!pip install auto-gptq optimum autoawq

!lm_eval --model hf --model_args pretrained=google/gemma-7b --tasks winogrande,hellaswag,arc_challenge --device cuda:0 --num_fewshot 1 --batch_size 8 --output_path ./eval_harness/gemma-7b
  
!lm_eval --model hf --model_args pretrained=google/gemma-7b --tasks winogrande,hellaswag,arc_challenge --device cuda:0 --num_fewshot 5 --batch_size 8 --output_path ./eval_harness/gemma-7b-5shot
  

AIAnytime

training_arguments = TrainingArguments(
        output_dir="./results",
        per_device_train_batch_size=4,
        per_device_eval_batch_size=4,
        gradient_accumulation_steps=2,
        optim="adamw_8bit",
        logging_steps=50,
        learning_rate=1e-4,
        evaluation_strategy="steps",
        do_eval=True,