Rishabh Sinha 11ec2397fe [Cache] Store cached kernel params as JSON instead of cloudpickle (#3143) ...

The kernel cache and the autotuner result cache loaded params.pkl with
cloudpickle.load(), so a writable cache directory allowed arbitrary code
execution on the next cache hit. Kernel params are plain data, so store
them as JSON via TVM's reflection (tvm.ir.save_json / load_json), which
rebuilds IR nodes without executing code. Entries without params.json are
treated as incomplete and rebuilt.

Fixes #2817

2026-09-06 16:06:20 +08:00

..

test_tilelang_cuda_binary_cache.py

[Cache] Store cached kernel params as JSON instead of cloudpickle (#3143)

2026-09-06 16:06:20 +08:00

test_tilelang_kernel_cache_atomic_save.py

[Cache] Store cached kernel params as JSON instead of cloudpickle (#3143)

2026-09-06 16:06:20 +08:00

test_tilelang_kernel_cache_integrity.py

[Cache] Store cached kernel params as JSON instead of cloudpickle (#3143)

2026-09-06 16:06:20 +08:00

test_tilelang_kernel_cache.py

[Runtime][Cache] Export libraries after disk-cache hits (#3116)

2026-08-31 19:22:37 +08:00