Skip to content

Ximilar-com/vlm-scripts

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Ximilar VLM Scripts

Example scripts for running your trained Vision-Language Models (VLMs) from the Ximilar Platform.

When you train a VLM on Ximilar, you can download the model and run it locally. This repository shows you how.

Frameworks

Run your models using Python with HuggingFace Transformers and PEFT (for LoRA adapters).

  • Simple run.py script per model
  • Supports LoRA adapters (.safetensors), full models (.safetensors), and PyTorch exports (.pt)
  • Auto-detects model format from directory contents
  • Works on NVIDIA GPU (CUDA), Apple Silicon (MPS), and CPU

Supported models: LiquidAI LFM2-VL-450M, LiquidAI LFM2.5-VL-1.6B, Google Gemma 3 4B, Qwen3-VL 2B, Qwen3-VL 4B

See the transformers/README.md for setup, usage, and troubleshooting.

Run quantized GGUF models locally using llama.cpp — full vision support, fast inference.

  • Single command with llama-mtmd-cli
  • Full image/vision support (unlike Ollama which doesn't support LFM2-VL vision yet)
  • GPU acceleration on Apple Silicon (Metal) and NVIDIA (CUDA)

Supported models: LiquidAI LFM2-VL-450M (GGUF quantized)

See the llama-cpp/README.md for setup, usage, and troubleshooting.

About

Scripts for running your trained VLM models from Ximilar Platform

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors