Blog
Notes on adapters, fine-tuning, and running models on your own hardware.
Training Your First LoRA
A practical first pass at building a LoRA adapter from a small, well-cleaned dataset.
2025-05-12Prompt Engineering for Adapter-Tuned Models
Prompt patterns that actually change behavior after a LoRA is applied.
2025-05-05Running LoRA Adapters in llama.cpp
Loading a safetensors LoRA into a GGUF base model for CPU inference.
2025-05-05Designing a Small Adapter Registry
How to structure a lightweight registry for LoRA adapters, metadata, versions, downloads, and model cards without heavy infrastructure.
2025-04-29Choosing Precision and Format
How to pick between fp16, bf16, int8, GGUF, and safetensors without breaking compatibility.
2025-04-23Secure Handling of Private Training Data
A practical guide to handling private uploads, temporary datasets, logs, and retention without leaking sensitive training data.
2025-04-15LoRA vs Full Fine-Tuning
A practical comparison of adapter training and full-model fine-tuning for real production workflows.
2025-04-11Writing a Model Card That Prevents Mistakes
Model card fields that actually matter for LoRA distribution and reuse.
2025-04-11Checksums and Reproducible Downloads
How to make model downloads verifiable, immutable, and safe to cache across a small adapter registry.
2025-04-03Serving Text Adapters with vLLM
Loading multiple LoRA adapters into a single vLLM instance and routing requests.
2025-03-19