VESSL RunLaunch Generative AI models in seconds
VESSL Run

Run AI models with
a single YAML

VESSL Run brings a unified YAML interface for traning, fine-tuning, and scaling
your own Stable Diffusion, LLaMA, and more — on any cloud, with a single command.

pip install vessl
vessl hello
Run AI model section



vessl run result image

Run any models
in seconds

VESSL’s unified YAML definition lets you run any models from Generative AI models to award-winning academic papers all with a single command.

Dreambooth on Stable Diffusion

Dreambooth on Stable Diffusion

Finetune Stable Diffusion in the fastest way.

View demo →
Segment Anything

Segment Anything

Segment ‘Anything’ using FAIR’s SAM.

LangChain QA Chatbot

LangChain QA Chatbot

Generate your own assistant trained with your data.

Thin-Plate-Spline Motion Model

Thin-Plate-Spline Motion Model

Animate your own image in the desired way.

MobileNeRF

MobileNeRF

Image rendering that can even run on your phone.

nanoGPT

nanoGPT

The fastest way to build your own storyteller.

ImageBind

ImageBind

Compare similarities across different modalities.

Visual Programming

Visual Programming

Let LLM solve your vision task without training.

Top Down Visual Attention

Top Down Visual Attention

Use Top-Down attention to improve Vision-Language tasks.

The latest open-source models
that just works

VESSL Run simplifies the manual Python dependency and CUDA configuration into a single YAML definition so you can spend more time iterating the models. Start training by pointing to GitHub repositories, our custom Docker images, and your datasets.

lastest model
stable-diffusion.yamlmain.py
1
2
3
4
5
6
7
8
9
10
11
12

name: stable-diffusion

resources:
cluster: aws
accelerators: V100:4
run:
- workdir: /root/Stable-Diffusion
command: |
python scripts/stable_txt2img.py

Launch projects from Git repos

Launch projects from Git repos

Use public, open-source models as a starting point for your projects simply by pointing to a GitHub repository.

Bring your own datasets

Bring your own datasets

Mount your datasets from the cloud or on-prem storage and fine-tune your model with the same YAML definition.

Custom Docker Images

Custom Docker Images

Our pre-built Docker images remove the need to manually install and configure environment dependencies.

Ready for scale

Fine-tune and deploy models at scale on any cloud without worrying
about the complex compute backends and system details.

stable-diffusion.yamlmain.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
name: stable-diffusion
volumes:
/input/dataset: s3://public-bucket/dreambooth/stable_diffusion/
/root/Stable-Diffusion: git://github.com/XavierXiao/ Dreambooth-Stable-Diffusion
image: vessl-ai/ngc-20.10:v1

run:
- workdir: /root/Stable-Diffusion
command: |
conda env create -f environment.yaml
conda activate ldm
mkdir data/
env:
class_word: vessl_logo
prompt: "a photo of vessl logo"

ready for scale

Train on multiple clouds

Launch your Run on any cloud with managed spot instances, elastic scaling, and resource monitoring.

AWSAWS
GCPGCP
AzureAzure
Oracle CloudOracle Cloud

Bring your own GPUs

Set up unlimited number of Kubernetes-backed on-prem GPU clusters with a single command.

vessl cluster create --name 'dgx-a100'

Automatic scaling

  • GPU Optimization
  • Batch scheduling
  • Distributed training
  • Termination protection
  • Pay by the second

Go beyond training

Add snippets to the existing YAML definition to deploy your fine-tuned model as a micro app or define them as individual steps for a CI/CD pipeline.

stable-diffusion.yamlmain.py
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

name : stable-diffusion
resources:
cluster: aws-apne2-prod1
accelerators: V100:1
image: quay.io/vessl-ai/ngc-pytorch-kernel:22.12-py3-202301160809
volumes:
/root/examples: git://github.com/vessl-ai/examples

ready for scale

Create your own micro AI app

Open up a port to your Run and integrate tools like Streamlit and Gradio to create interactive apps.

Automate with CI/CD

Orchestrate every step of ML from data ingestion to model deployment with a single YAML file.

Sign up for private beta

Get started in
under a minute

Get started right from your terminal and train your first nanoGPT with free VESSL Run credits.

pip install vessl
vessl hello

Sign up today and receive 5 hours of high-performance GPU instances.

nanoGPT