fixed an issue that prevented gguf files from loading

gui functionality

Add GGUF models directory and update CMakeLists files

implementation of gui

Phase 6: Moonshot Expansions - Hybrid ray-traced attention for 1M+ tokens, multimodal video decode prototype, adaptive precision switching, federated multi-GPU external memory, WASM/Docker/browser ecosystem

Phase 5: Ultimate Performance Polish - Optimize ray query MoE with dynamic thresholds/coop matrices, full bindless overhaul, async enhancements, dequant fusion expansion, benchmark suite upgrades

Phase 4: Verify & Stabilize - Stress tests for sparse MoE, GEMM benchmarks, multi-GPU scaling validation, numerical stability checks, portability audit

Phase 3: Out-of-the-Box Moonshots - Sparse MoE attention with ray queries, mixed-precision support, async compute queues, Vulkan video extensions, CI/GitHub Actions, comparison table

Phase 2: Performance Polish - GEMM shader optimizations, kernel fusion, paged KV cache, bindless descriptors, benchmark enhancements

Phase 1: Verification & Hardening - Replace all mock test code with real model loading and generation, add speculative decoding flag, implement GPU profiling with timestamps

Phase 1: Verification & Hardening - Add runtime Vulkan extension checks with fallbacks, implement per-layer GPU timestamp queries for matmul/attention/transfer breakdown, expand tests with integration tests for speculative decoding, multi-GPU, LoRA, verify FlashAttention causality on long prompts, tune dynamic workgroup sizes

Phase 3: Maximum Fusion & Optimizations - fused FFN and RMSNorm kernels

Phase 2: True FlashAttention-2 Mastery - complete tiled causal attention

Phase 1: Perfect GEMM Suite - optimized tiled matmul kernels

feat: Phase 1 - Rewrite GEMM shaders with modern optimizations