Check Out GenAI for Text-to-Image with a Stable Diffusion Intel Tiber Developer Cloud Workshop.
What is Intel Tiber Developer Cloud?
With access to state-of-the-art Intel hardware and software solutions, developers, AI/ML researchers, ecosystem partners, AI startups, and enterprise customers can build, test, run, and optimize AI and High-Performance Computing applications at a low cost and overhead thanks to the Intel Tiber Developer Cloud, a cloud-based platform. With access to AI-optimized software like oneAPI, the Intel Tiber Developer Cloud offers developers a simple way to create with small or large workloads on Intel CPUs, GPUs, and the AI PC.
Developers and enterprise clients have the option to use free shared workspaces and Jupyter notebooks to explore the possibilities of the platform and hardware and discover what Intel can accomplish.
Text-to-Image
This article will guide you through a workshop that uses the Stable Diffusion model practically to produce visuals in response to a written challenge. You will discover how to conduct inference using the Stable Diffusion text-to-image generation model using PyTorch and Intel Gaudi AI Accelerators. Additionally, you will see how the Intel Tiber Developer Cloud can assist you in creating and implementing generative AI workloads.
Text To Image AI Generator
AI Generation and Steady Diffusion
Industry-wide, generative artificial intelligence (GenAI) is quickly taking off, revolutionizing content creation and offering fresh approaches to problem-solving and creative expression. One prominent GenAI application is text-to-image generation, which uses an understanding of the context and meaning of a user-provided description to generate images based on text prompts. To learn correlations between words and visual attributes, the model is trained on massive datasets of photos linked with associated textual descriptions.
A well-liked GenAI deep learning model called Stable Diffusion uses text-to-image synthesis to produce images. Diffusion models work by progressively transforming random noise into a visually significant result. Due to its efficiency, scalability, and open-source nature, stable diffusion is widely used in a variety of creative applications.
The Stable Diffusion model in this training is run using PyTorch and the Intel Gaudi AI Accelerator. The Intel Extension for PyTorch, which maximizes deep learning training and inference performance on Intel CPUs for a variety of applications, including large language models (LLMs) and Generative AI (GenAI), is another option for GPU support and improved performance.
Stable Diffusion
To access the Training page once on the platform, click the Menu icon in the upper left corner.
The Intel Tiber Developer Cloud‘s Training website features a number of JupyterLab workshops that you may try out, including as those in AI, AI with Intel Gaudi 2 Accelerators, C++ SYCL, Gen AI, and the Rendering Toolkit.
Workshop on Inference Using Stable Diffusion
Thwy will look at the Inference with Stable Diffusion v2.1 workshop and browse to the AI with Intel Gaudi 2 Accelerator course in this tutorial.
Make that Python 3 (ipykernel) is selected in the upper right corner of the Jupyter notebook training window once it launches. To see an example of inference using stable diffusion and creating an image from your prompt, run the cells and adhere to the notebook’s instructions. An expanded description of the procedures listed in the training notebook can be found below.
Note: the Jupyter notebook contains the complete code; the cells shown here are merely for reference and lack important lines that are necessary for proper operation.
Configuring the Environment
Installing all the Python package prerequisites and cloning the Habana Model-References repository branch to this docker will come first. Additionally, They are going to download the Hugging Face model checkpoint.
%cd ~/Gaudi-tutorials/PyTorch/Single_card_tutorials
!git clone -b 1.15.1 https://github.com/habanaai/Model-References
%cd Model-References/PyTorch/generative_models/stable-diffusion-v-2-1
!pip install -q -r requirements.txt
!wget https://huggingface.co/stabilityai/stable-diffusion-2-1-base/resolve/main/
v2-1_512-ema-pruned.ckpt
Executing the Inference
prompt = input("Enter a prompt for image generation: ")The prompt field is created by the aforementioned line of code, from which the model generates the image. To generate an image, you can enter any text; in this tutorial, for instance, they’ll use the prompt “cat wearing a hat.”
cmd = f'python3 scripts/txt2img.py --prompt "{prompt}" 1
--ckpt v2-1_512-ema-pruned.ckpt \
--config configs/stable-diffusion/v2-inference.yaml \
--H 512 --W 512 \
--n_samples 1 \
--n_iter 2 --steps 35 \
--k_sampler dpmpp_2m \
--use_hpu_graph'
print(cmd)
import os
os.system(cmd)Examining the Outcomes
Stable Diffusion will be used to produce their image, and Intel can verify the outcome. To view the created image, you can either run the cells in the notebook or navigate to the output folder using the File Browser on the left-hand panel:
/Gaudi-tutorials/PyTorch/Single_card_tutorials/Model-References
/PyTorch/generative_models/stable-diffusion-v-2-1/outputs/txt2img-samples/
Once you locate the outputs folder and locate your image, grid-0000.png, you may examine the resulting image. This is the image that resulted from the prompt in this tutorial:
You will have effectively been introduced to the capabilities of GenAI and Stable Diffusion on Intel Gaudi AI Accelerators, including PyTorch, model inference, and quick engineering, after completing the tasks in the notebook.
