Welcome to TIR AI Platform Documentation!

TIR is a modern AI Development Platform designed to tackle the friction of training and serving large AI models.

We do so by using highly optimised GPU containers (NGC), pre-configured environments (pytorch, tensorflow, triton), automated API generation for model serving, shared notebook storage, and much more.

• Introduction
  • Introduction
    • Components of TIR Platform
    • Why AI Model Development is so hard?
    • Key Features of TIR Platform
• Getting Started
  • Notebooks
  • Datasets
  • Using Datasets with Notebooks
  • Model and Model Endpoints
• How-to Guides
  • Create notebooks instance with your own container
    • 1. Build a container image
      • 1.1 Patching using the Dockerfile
      • 1.2 Patching an instance of the container and Saving the Image
      • 1.3 Using Image Builder Utility
    • 2. Push the container image
      • 2.1 Setup Registry Integration
      • 2.2 Setup Registry Credential on your local system
      • 2.3 Push the image
    • 3. Create Notebook
  • List of TIR-provided Images
    • TIR-provided Images
    • Image Descriptions & Versions
      • Diffusers
      • Fast.AI
      • Jupyter
      • NVIDIA Cuda
      • NVIDIA Nemo-Speech
      • NVIDIA RAPIDS
      • PyTorch
      • SciPy
      • TensorFlow
      • Transformers
      • Ubuntu
• Projects
• Notebooks
  • Getting Started
  • Notebook options
  • Notebook Statuses
  • How to create Notebook ??
    • Notebook Details
      • Overview
      • Disk Size
      • Metrics
      • Associated Datasets
      • SSH Key
      • Update SSH Key
      • Add SSH Key After Notebook Creation
    • Notebook Actions
      • Launch Notebook
      • Stop
      • Update Notebook
      • Delete Notebook
    • Launch Notebook from Sidebar
• Committed Notebook
  • Creation of committed notebook
    • Getting Started
      • Auto-Renewal
      • Hourly Billing
      • Auto Deletion
  • Updating a hourly notebook to committed notebook
  • Updating a committed notebook to committed notebook
  • Delete Committed Notebook
• GPU H100 Plans
  • Notebook options
  • Steps to create GPU H100 notebook
    • Notebook Details
      • Overview
      • Disk Size
      • Metrics
      • Associated Datasets
      • SSH Key
    • Here are the Steps to verify the configuration of the GPU H100 notebook using the terminal
    • Or you can verify the configuration of the GPU H100 notebook using the Lab URL
    • Notebook Actions
      • Launch Notebook
      • Stop
      • Delete Notebook
• Datasets
  • How does it work?
  • Benefits
  • Usage
    • WebUI
    • SDK
  • Getting Started
    • Prequisites
    • Create a new dataset
    • Copy data into Dataset
  • Working with Datasets
    • Import Data from local machine
    • Import Data from Amazon S3
    • Import Data from GCS
    • Upload data from TIR Jupyter Notebook
    • Permenantly Migrate from other Cloud
  • How to Create Datasets ?
  • Delete DataSet
• Models (Storage)
  • Uploading weights to TIR Models
  • Downloading weights from TIR models
  • How to create Model in TIR dashboard?
• Model Deployments
  • Pre-built Containers
    • TorchServe
    • NVIDIA Triton
    • LLAMA2
    • CodeLLMA
    • Stable Diffusion
  • Custom Containers
• Fine-Tuning Models
  • Introduction
  • What is Fine-Tuning ?
  • How to Create a Fine-Tuning Job ?
  • How to define Dataset-Preparation ?
  • How to define a Hyperparameter Configuration ?
  • Viewing your Job parameters and Finetuned models ?
• Samples and Tutorials
  • Fine-tune LLMA with Multiple GPUs
    • Steps
    • Conclusion
  • Deploy Inference for LLMA 2
  • Deploy Inference for Codellama
    • Model Endpoint creation for Codellama-7b using prebuilt container
    • Creating Model endpoint with custom model weights
      • Supported Parameters
  • Deploy Inference for Stable Diffusion v2.1
    • A guide on Model Endpoint creation & Image generation
    • Creating Model endpoint with custom model weights
    • Supported parameters for image generation
  • Fine-tune Stable Diffusion model on TIR
    • About Training methods
      • Dreambooth
      • Textual Inversion
    • Fine tuning the model
      • Step-1: Launch a Notebook on TIR
      • Step-2: Initial Setup
      • Step-3: Settings for teaching the new concept
      • Step-4: Teach the model the new concept (Fine-tuning with the training method)
      • Step-5: Run Inference with the newly trained Model
      • Step-6: Save the newly created concept
      • Step-7: Create Inference Server against our newly trained model
    • Conclusion
  • Deploy Inference for MPT-7B-CHAT
    • Creating a ready-to-use Model endpoint
    • Creating Model endpoint with custom model weights
      • INFERENCE:
      • Supported parameters
  • Custom Containers in TIR
  • Fine Tuning Bloom
    • Introduction
    • What Is Fine-Tuning?
    • Why Fine-Tune BLOOM?
    • Benefits of Fine-Tuning BLOOM
    • Requirements
    • Python Libraries
    • Launch Your GPU-Backed Notebook
    • Packages and Libraries
    • BLOOM Model and Tokenizer
      • Training Data
      • Example
    • Training the BLOOM Model
      • Load the BLOOM model and tokenizer
      • Create the validation dataset
      • Define the training arguments
      • Create the trainer
    • Evaluation
      • Saving and Loading
      • Example 1:
      • Example 2:
      • Example 3:
    • Conclusion
  • Natural Language Queries to SQL wth Code Llama
    • Introduction
      • Technologies
      • LLama 2 architecture used for Code llama
      • Prerequisites
      • How to Use and Querying Examples
      • Challenges and Solutions
      • Conclusion
• API Tokens
• SSH Key
  • Introduction
  • Manage SSH Keys
    • Logging into E2E Networks ‘TIR AI PLATEFORM’
    • Navigate to SSH Keys
    • Add/Delete SSH Keys
    • ADD SSH Keys
    • Delete SSH Keys
    • Using an SSH Key
• Team Features
  • To create a Team click on create team.
  • How to add Team members in a team?
  • You can remove the member by clicking on Remove.
  • To delete the team, choose the team and then click on the delete icon.
• Settings
• Analytics
  • Activity
  • Resource usage
• Billing
  • Notebooks
  • Free-Tier Notebooks
  • Notebook Disks (PVC)
  • Datasets