Visual Product Recommender
🛒
1
Find similar products by uploading a product image
lawrence
Lawrence-okolo
AI & ML interests
Enterprise AI systems, LLM architecture, edge AI deployment, MLOps, robotics intelligence, real-time inference systems, privacy-preserving machine learning, and production-grade AI pipelines.
Recent Activity
liked
a Space 2 days ago
Lawrence-okolo/product-recommendation updated
a dataset 6 days ago
Lawrence-okolo/abo-visual-search-index published
a dataset 8 days ago
Lawrence-okolo/abo-visual-search-index Organizations
None yet
liked a Space 2 days ago
updated a dataset 6 days ago
published a dataset 8 days ago
updated a Space 8 days ago
published a Space 8 days ago
commented on Deploying Open Source Vision Language Models (VLM) on Jetson 11 days ago
Running Cosmos-Reason2-2B on Jetson Orin 8GB requires careful memory management before launching.
You must stop the desktop environment (sudo systemctl isolate multi-user.target), disable swap (sudo swapoff -a), and flush OS cache (sudo sysctl -w vm.drop_caches=3) to free enough unified memory.
The default --max-model-len 256 works for text-only but is too small for image inputs .. use at least --max-model-len 1024 with --gpu-memory-utilization 0.55 for vision tasks.
With these changes the model loads successfully and the API responds correctly, however the 8GB unified memory leaves no headroom for concurrent robotics applications such as ROS2 .. a dual-Jetson setup or upgrading to a higher memory variant is recommended for production deployments.
https://vasi.ca
