Title: CUDA compatibility issue on Google Colab with L4 GPU

[MagellaX]

Description

Unable to run Luminal examples with CUDA support on Google Colab due to cudarc version incompatibility.

Environment

• Platform: Google Colab
• GPU: NVIDIA L4
• CUDA Version: 12.2 (nvcc: V12.2.140)
• Driver Version: 550.54.15
• Luminal commit: [latest main branch]

Error

thread 'main' panicked at /root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/cudarc-0.16.6/src/driver/sys/mod.rs:19260:18:
Expected symbol in library: DlSym { desc: "/usr/lib64-nvidia/libcuda.so: undefined symbol: cuCtxCreate_v4" }

Steps to Reproduce

1. Clone luminal on Google Colab
2. Run cargo run --release --features cuda in examples/llama
3. Error occurs during graph compilation

Suggested Fix

Update cudarc dependency to a version compatible with CUDA 12.2, or provide documentation on supported CUDA versions.