v0.1.2

• Fix a 32-bit fseek truncation on Windows: dit-bf16.gguf and t5-bf16.gguf are 2.4-2.6GB, past the 2GB boundary where MSVC's 32-bit long truncates fseek's offset. This surfaced as "t5 load failed" on Windows (any backend) while working fine on Linux. Now uses a proper 64-bit seek (_fseeki64/fseeko).

v0.1.1

• Bundle CUDA runtime (libcudart/libcublas/libcublasLt) in the CUDA release tarball instead of requiring a system CUDA toolkit; add Blackwell (sm_120) kernel + PTX fallback
• Fix Windows CUDA 13 DLL location (moved to bin/x64)
• Windows: build with GGML_OPENMP=OFF (TheRock's Windows ROCm dist doesn't ship the LLVM OpenMP runtime; ggml's own thread pool covers it)

v0.1.0

Prebuilt backend binaries for Lemonade.