Optimized Coding Model Released for Small-GPU Devices

A developer released SharpSpark, a quantized Spark-X2.5-4B optimized for small-GPU and sub-16GB-RAM devices, able to solve real coding problems autonomously. Changes include chat-template fixes, a custom importance matrix and per-tensor bit allocation.