Sign inSign up

ls250824/run-ai-toolkit

By ls250824

•Updated 2 months ago

Training Lora’s / models for ComfyUI with AI TOOLKIT user interface.

Image
Machine learning & AI
0

912

ls250824/run-ai-toolkit repository overview

⁠run-ai-toolkit

⁠Synopsis

A streamlined setup for running Ostris AI Toolkit for training models and LoRAs for ComfyUI.

  • UI interface
  • Authentication credentials can be set via secrets for:
    • Code server authentication.
    • Hugging Face token for model access.
    • ai-toolkit UI password.

⁠Hardware provisioning recommendation

⁠Z-Image Base, Flux-Klein 9B, Krea-2
  • GPU: RTX A5000, RTX 4090.
  • Pod volume: 40 GB (model)
  • Workspace: 20 GB (dataset and output)
⁠LTX-2.3
  • GPU: L40S.
  • Pod volume: 80 GB (model)
  • Workspace: 20 GB (dataset and output)

⁠Setup

ComponentVersion
OSUbuntu 24.04 x86_64
Python3.12.x
PyTorch2.10.0
CUDA12.8.x
Triton3.4.x
Code ServerLatest
ai-toolkitLatest

⁠Installed Attentions

⁠Wheels
PackageVersion
flash_attn2.8.3
sageattention2.2.0
⁠Build for
Processor exampleCompute CapabilitySM
RTX A50008.6sm_86
RTX 40908.9sm_89

⁠Environment Variables

⁠Authentication Tokens
TokenEnvironment Variable
Hugging FaceHF_TOKEN
Code ServerPASSWORD
ai-toolkitAI_TOOLKIT_AUTH

⁠Connection options

⁠Services
ServicePort
ai-toolkit UI8675 (HTTP)
Code Server9000 (HTTP)
SSH/SCP22 (TCP)

Tag summary

Content type

Image

Digest

sha256:ae0e7fe96…

Size

11 GB

Last updated

2 months ago

docker pull ls250824/run-ai-toolkit:27072026