Build

Tensorfuse

Tensorfuse lets you run serverless GPUs on your own infrastructure

Introduction

Deploy serverless GPU applications on your AWS account

Getting Started 

Environments help you isolate your resources and manage your deployments.

Environments

Declare flags in a config file and use it run tensorkube deploy

Config File

Use secrets in your deployments for sensitive information

Working with Secrets

Hot-reloading GPU-enabled containers for seamless testing and development

Dev Containers

Deploy your jobs and queue them programmatically with different parameters

Job Queues

Add custom domains to your ML endpoints with HTTPS

Custom domains with TLS

Finetune LoRA adapters for popular models using axolotl styled declarative configs

Finetuning

Finetune Llama 3 70B on your AWS account

Add members to your team and give them access to your cluster

Adding Other Members

How to create a tensorkube deployment with github actions

Using Tensorkube with Github Actions

Debug your applications by querying past logs

Debugging with Logs

Debugging Tensorfuse Deployments and Jobs with logs

Use volumes in your tensorkube deployments

Volumes

Understanding how Tensorfuse deals with node failures

Handling Node Failures

Guidelines and best practices for deploying applications on Tensorkube

Troubleshooting

Keep Tensorfuse runtime isolated and control Billing

Sub-accounts

Architecture of resources provisioned during configure

Configure

Serve

Jobs

Deploy Llama 4 Scout and Maverick from Meta models using Tensorfuse

Deploy production-ready Llama 4 on AWS

Deploy Llama 4 Models on your AWS account

Deploy a production ready LLM service using vLLM

Deploy Llama-3.3-70B-Instruct on Serverless GPUs

Deploy Qwen QwQ 32B service using vLLM

Deploy Qwen QwQ 32B on Serverless GPUs

Deploy Deepseek R1 671B param model using Tensorfuse

Deploy production ready DeepSeek R1 on AWS

Deploy a production-ready DeepSeek R1 service using vLLM

Deploy Deepseek R1 671B on Serverless GPUs

Deploy GGUF quants of Deepseek R1 671B param model using Tensorfuse

Deploy production ready GGUF quants of DeepSeek R1 on AWS

Deploy a production-ready llama.cpp service

Deploy GGUF quants of Deepseek R1 671B on Serverless GPUs

Deploying Pixtral-12B-Quantised-4-Bit on Serverless GPUs

Deploy Jina Embeddings model on your AWS

Fine-tune Qwen 7B for reasoning tasks on your AWS account using tensorfuse and unsloth with GRPO.

Transforming Qwen 7B into Your Own Reasoning Model

Deploy Mochi 1 Preview Video Model using ComfyUI on your AWS

Deploy Stable Diffusion XL model using ComfyUI on your AWS

Run SAM2 on your own AWS on Serverless GPUs

Deploying Stable Diffusion 3 Medium on Serverless GPUs

Deploying Resnet on serverless GPUs

Deploying FLUX.1-dev on Serverless GPUs

Deploying SpeechT5 on serverless GPUs

Deploy a serverless Sesame CSM 1B model on your AWS account

Deploying Sesame-CSM-1B on Serverless GPUs

Cheatsheet for Tensorkube users

Install AWS-CLI on Mac OS

Changelog

tensorkube account

tensorkube cluster

tensorkube configure

tensorkube datasets

tensorkube deploy

tensorkube deployment

tensorkube dev

tensorkube domain

tensorkube environment

tensorkube get-permissions-command

tensorkube get-principal-arn

tensorkube give-cluster-access

tensorkube install-prerequisites

tensorkube job

tensorkube list

tensorkube login

tensorkube reset

tensorkube secret

tensorkube sync

tensorkube teardown

tensorkube train

tensorkube upgrade

tensorkube version

tensorkube volume

AWS Permissions required for running `tensorkube configure`

Basics

Enterprise Setup

Architecture

Build