A LocalLLaMA user asked what models and tools people would archive in a hypothetical future where Chinese model weights are banned and US labs stop releasing open models.
Key points
The post asks which models to save for coding, security research, vision, embeddings and other tasks
It also asks about preserving fine-tuning and quantization tools, data and storage methods
The user says they currently use a Mac Studio with 192GB of memory
The post frames the scenario as a possibility, citing concerns about restrictions on Chinese models and fewer open releases from US labs. It asks readers what they are actually preserving, including model weights, training and quantization tools, calibration data, and copies stored on offline drives or through mirrors and torrents.