Gpt4allloraquantizedbin+repack Upd

The LoRA adapters were incorrectly fused into the base model. This happens with sloppy repacks. Fix: Download a different repack from a trusted quantizer (e.g., "MaziyarPanahi" or "TheBloke" archives).

A low-rank approximation of a ghost. LoRA fine-tune of GPT4All-XL-v2. Quantized with optimal rounding. Repacked to decouple inference from attention dimension constraints. Also: there is a wasp nest in your attic. Northeast corner. gpt4allloraquantizedbin+repack

The step merges the LoRA adapter into the base model, then quantizes the combined result. Benefits: The LoRA adapters were incorrectly fused into the base model