googlecolab-colabtools

Install the necessary libraries for image generation and move the pipeline to the GPU (cuda).

Required packages:

• diffusers
• accelerate
• mediapy

This example uses the stabilityai/sdxl-turbo model with torch.float16 for optimized performance.

%pip install --quiet --upgrade diffusers accelerate mediapy

import mediapy as media, random, sys, torch
from diffusers import AutoPipelineForText2Image

pipe = AutoPipelineForText2Image.from_pretrained(
    "stabilityai/sdxl-turbo",
    torch_dtype=torch.float16,
    use_safetensors=True,
    variant="fp16",
    )

pipe = pipe.to("cuda")

Stable Diffusion requires a GPU to run efficiently. Before starting, verify that a GPU is attached to your Colab runtime by checking the output of nvidia-smi. If it fails, navigate to Runtime > Change runtime type and select a GPU-enabled hardware accelerator.

import os

if os.system('nvidia-smi'):
  raise Exception("No GPU found. Access a GPU through Runtime > Change runtime type and try again.")

If you need help using Colab, you can:

• General Questions: Submit questions on StackOverflow using the google-colaboratory tag.
• Product Issues: Submit an issue via GitHub Issues or use the "Help" -> "Send Feedback" menu within the Colab interface.

Models in google.colab.ai are categorized by their intended use case:

• Pro: Most capable models, ideal for complex reasoning, creative tasks, and detailed analysis.
• Flash: Optimized for high speed and efficiency; best for summarization, chat applications, and rapid responses.
• Gemma: Lightweight, open-weight models suitable for various text generation tasks and experimentation.

Colab breaks your notebook into slides based on cell types and hierarchy:

• Standard Cells: Generally, each individual cell (text or code) becomes its own slide.
• Code Cells: Code cells are shown as slides and remain executable within slideshow mode.
• Collapsible Sections (Special Case): If you use Colab's collapsible section feature to create a hierarchy, a section consisting of exactly one text cell (explaining a concept) and one code cell (demonstrating the concept) will be grouped together into a single slide.
• Slide Titles:
  • By default, the notebook title is used.
  • If using collapsible sections, the slide title becomes the header of the parent section (e.g., an H2 header will use its parent H1 header as the title).

You can enter slideshow mode to present your notebook as a series of slides. You can start a slideshow from the beginning or from the currently focused cell.

Ways to start from the beginning:

• Use the menu: View > Start slideshow from beginning
• Use the command palette (Ctrl + Shift + P) and search for Start notebook slideshow from beginning
• Use the keyboard shortcut: Alt + Shift + V
• Append #slideshowMode=true to the end of any Colab notebook URL.

Ways to start from the current cell:

• Use the menu: View > Start slideshow
• Use the command palette (Ctrl + Shift + P) and search for Start notebook slideshow
• Use the keyboard shortcut: Alt + V

To use OpenAI models in Colab, follow these steps:

1. Create an API key at https://platform.openai.com/api-keys.
2. In Colab, click the key icon (Secrets) on the left sidebar.
3. Add a new secret named OPENAI_API_KEY and paste your key.
4. Grant the notebook access to the secret.

Note: It is recommended to install llmx before openai to avoid installation errors.

Required packages:

• llmx
• openai

!pip install -q llmx
!pip install -q openai
from openai import OpenAI
from google.colab import userdata

openai_api_secret_name = 'OPENAI_API_KEY'

try:
  OPENAI_API_KEY=userdata.get(openai_api_secret_name)
  client = OpenAI(
    api_key=OPENAI_API_KEY
  )
except userdata.SecretNotFoundError as e:
   # Handle missing secret
   raise e
except userdata.NotebookAccessError as e:
  # Handle missing access permissions
  raise e