[Bug] Error compiling WASM under Windows11: tvm.error.InternalError

[willo83417]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

1. Use Anaconda Prompt terminal > activate webllm
2. set TVM_SOURCE_DIR=C:\Users\Username\mlc-llm\3rdparty\tvm
3. set MLC_LLM_SOURCE_DIR=C:\Users\Username\mlc-llm
4. call C:\Users\Username\emsdk\emsdk_env.bat
5. cd C:\Users\Username\dist\models
6. git clone https://huggingface.co/mlc-ai/Qwen3-0.6B-q4f16_1-MLC
7. python -m mlc_llm compile dist\models\Qwen3-0.6B-q4f16_1-MLC\mlc-chat-config.json --overrides "context_window_size=4096;prefill_chunk_size=1024" --device webgpu -o dist\libs\Qwen3-0.6B-q4f16_1-webgpu.wasm

log.txt

Expected behavior

Environment

• Platform (e.g. WebGPU/Vulkan/IOS/Android/CUDA):
• Operating system (e.g. Ubuntu/Windows/MacOS/...):
• Device (e.g. iPhone 12 Pro, PC+RTX 3090, ...)
• How you installed MLC-LLM (conda, source):
• How you installed TVM (pip, source):
• Python version (e.g. 3.10):3.13
• GPU driver version (if applicable):AMD Radeon 680M (25.8.1)/ Nvidia Geforce RTX4060 Laptop GPU (580.97)
• CUDA/cuDNN version (if applicable): CUDA Version: 13.0
• TVM Hash Tag , applicable if you compile models): GIT_COMMIT_HASH: c6fb2be79f654588fd94727a74b9ca0754f63fa4
• Any other relevant information:

Additional context

Different from the official tutorial is
https://llm.mlc.ai/docs/install/emcc.html
Step 3 cannot use "./web/prep_emcc_deps.sh", use choco install make to directly use make for compilation within \mlc-llm\web.

The directory of \mlc-llm\web\dist\wasm

2,112 mlc_wasm_runtime.bc

57 mlc_wasm_runtime.d

195 mlc_wasm_runtime.wasm

Similarly, use make to compile within \mlc-llm\3rdparty\tvm\web
The directory of \mlc-llm\3rdparty\tvm\web\dist\wasm

80,935 tvmjs_runtime.js

4,577,631 tvmjs_runtime.wasm

173,764 tvmjs_support.bc

4,125 tvmjs_support.d

8,327,796 wasm_runtime.bc

11,530 wasm_runtime.d

233,464 webgpu_runtime.bc

5,047 webgpu_runtime.d