What's Changed

• feature: GGML backend close functions by @deadprogram in #96
• feature: download the cudart files needed for Windows CUDA install by @deadprogram in #97
• feature: add remaining LoRA functions bindings from llama.cpp by @deadprogram in #99
• benchmarks: improvements by @deadprogram in #100
• Fixing Issue 95 with downloading llamacpp by @ardan-bkennedy in #98
• fix: use singleton for model progress loader callback by @deadprogram in #101
• feature: usability improvements by @deadprogram in #103
• fix: use mtmd_helper_log_set instead of mtmd_log_set to control logging by @deadprogram in #104
• examples: cleanup logs to avoid extra noise by @deadprogram in #105

Full Changelog: v1.0.0-beta1...v1.0.0-beta2