[Update] Update MLC course materials to TVM latest API

[tlopex]

Overview

This PR updates the MLC course materials to be compatible with TVM latest API changes. This is the first step in modernizing the course content, focusing on API updates. Additional content updates will follow in subsequent PRs.

Changes

API Updates

1. Compilation API

   • tvm.build() → tvm.compile() for TIR modules
   • relax.build() → tvm.compile() for Relax modules
   • lib.time_evaluator() → lib.mod.time_evaluator() (tvm.compile returns Executable with mod attribute)

2. Runtime API

   • tvm.nd.array() → tvm.runtime.tensor()
   • tvm.nd.empty() → tvm.runtime.tensor(np.empty(...))
   • Type annotations: tvm.nd.NDArray → tvm.runtime.Tensor

3. Function Registration

   • @tvm.register_func() → @tvm.register_global_func()

Code Fixes

• Fixed syntax errors in code blocks
• Updated installation commands (mlc-ai-nightly → mlc-ai-nightly-cpu)
• Fixed pseudo-code syntax issues

Next Steps

In the following PR, I will update repo mlc-zn as well and then include updating outdated usage
(like PyTorch FX->exported_program).

[tlopex]

cc @tqchen @MasterJH5574

[MasterJH5574]

@tlopex Hey Shushi, could you also update this line to -cpu so the github action can run? Right now it fails due to this reason.

mlc-en/static/build.yml

Line 6 in 02262bf

- mlc-ai-nightly

[tlopex]

cc @MasterJH5574