Blog
Notes on adapters, fine-tuning, and running models on your own hardware.
License Traps in LoRA Training and Distribution
License gotchas around base models, training data, and derived adapters.
2026-04-18
Base Checkpoint Management in a Registry
Keeping base model versions in sync with adapters that depend on them.
2026-03-05
Continuous Training Pipelines for Adapters
Setting up automated retraining of adapters as new data lands.
2026-02-12
Adapter Caching for a Self-Hosted Registry
Cache layout, ETag handling, and bandwidth planning for a private adapter registry.
2026-01-21
Hot-Swap Strategies for Adapter Serving
Loading and unloading LoRA adapters in a running server without dropping requests.
2025-12-08
LoRA on Apple Silicon with MLX
Training and running LoRA on M-series Macs using the MLX framework.
2025-11-15Loading LoRA Explicitly in the Diffusers Pipeline
Using load_lora_weights and set_adapters to control multiple LoRAs in Python code.
2025-10-20Using LoRA in Automatic1111
Loading a downloaded LoRA into the Stable Diffusion WebUI and controlling its weight.
2025-09-02
Multi-Adapter Fusion and Task Routing
Stacking several LoRA adapters on one base model, and routing requests to the right one.
2025-08-14Artifact Storage Layout for Self-Hosted Model Hubs
How to organize adapter files, checkpoints, and generated assets in a self-hosted registry so releases stay immutable and easy to verify.
2025-07-24