Check failed: kNumAttrs == attrs.size() (2 vs. 1) : ValueError: Incorrect kNumAttrs for instruction: Split

When I tried to build the model on my Mac, I got an error:

python3 build.py
Automatically configuring target: metal -keys=metal,gpu -max_function_args=31 -max_num_threads=256 -max_shared_memory_per_block=32768 -max_threads_per_block=1024 -thread_warp_size=32
Load cached module from dist/mod_cache_before_build.pkl and skip tracing. You can use --use-cache=0 to retrace
Traceback (most recent call last):
  File "/Users/b03/Desktop/old/build.py", line 195, in <module>
    build(mod, ARGS)
  File "/Users/b03/Desktop/old/build.py", line 130, in build
    db = ms.database.create(work_dir=args.db_path)
  File "/opt/anaconda3/envs/web2/lib/python3.10/site-packages/tvm/meta_schedule/database/database.py", line 417, in create
    return JSONDatabase(*args, **kwargs)
  File "/opt/anaconda3/envs/web2/lib/python3.10/site-packages/tvm/meta_schedule/database/json_database.py", line 86, in __init__
    self.__init_handle_by_constructor__(
  File "tvm/_ffi/_cython/./object.pxi", line 132, in tvm._ffi._cy3.core.ObjectBase.__init_handle_by_constructor__
  File "tvm/_ffi/_cython/./packed_func.pxi", line 288, in tvm._ffi._cy3.core.ConstructorCall
  File "tvm/_ffi/_cython/./packed_func.pxi", line 277, in tvm._ffi._cy3.core.FuncCall
  File "tvm/_ffi/_cython/./base.pxi", line 182, in tvm._ffi._cy3.core.CHECK_CALL
  File "/opt/anaconda3/envs/web2/lib/python3.10/site-packages/tvm/_ffi/base.py", line 481, in raise_last_ffi_error
    raise py_err
tvm._ffi.base.TVMError: Traceback (most recent call last):
  File "/Users/catalyst/Workspace/mlc-ai-package-self-runner/_work/package/package/tvm/src/support/parallel_for.cc", line 128
RuntimeError: parallel_for_dynamic error with [17:46:32] /Users/catalyst/Workspace/mlc-ai-package-self-runner/_work/package/package/tvm/src/meta_schedule/database/json_database.cc:198: ValueError: Unable to parse TuningRecord, on line 7 of file log_db/database_tuning_record.json. The workload is:
# from tvm.script import ir as I
# from tvm.script import tir as T

@I.ir_module
class Module:
    @T.prim_func
    def main(rxplaceholder: T.Buffer((T.int64(1), T.int64(4), T.int64(64), T.int64(64)), "float32"), T_multiply: T.Buffer((T.int64(1), T.int64(4), T.int64(64), T.int64(64)), "float32")):
        T.func_attr({"op_pattern": 0, "tir.noalias": T.bool(True)})
        # with T.block("root"):
        for ax0, ax1, ax2, ax3 in T.grid(T.int64(1), T.int64(4), T.int64(64), T.int64(64)):
            with T.block("T_multiply"):
                v_ax0, v_ax1, v_ax2, v_ax3 = T.axis.remap("SSSS", [ax0, ax1, ax2, ax3])
                T.reads(rxplaceholder[v_ax0, v_ax1, v_ax2, v_ax3])
                T.writes(T_multiply[v_ax0, v_ax1, v_ax2, v_ax3])
                T_multiply[v_ax0, v_ax1, v_ax2, v_ax3] = T.float32(0.041666667908430099) * rxplaceholder[v_ax0, v_ax1, v_ax2, v_ax3]
The JSONObject of TuningRecord is:
[T.int64(6), [[[["GetBlock", [], ["T_multiply", "main"], ["b0"]], ["GetLoops", ["b0"], [], ["l1", "l2", "l3", "l4"]], ["Fuse", ["l1", "l2", "l3", "l4"], [T.int64(1)], ["l5"]], ["SampleCategorical", [], [[T.int64(32), T.int64(64), T.int64(128), T.int64(256), T.int64(512), T.int64(1024)], [T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666)]], ["v6"]], ["Split", ["l5", "None", "v6"], [T.int64(1)], ["l7", "l8"]], ["Bind", ["l7"], ["blockIdx.x"], []], ["Bind", ["l8"], ["threadIdx.x"], []], ["EnterPostproc", [], [], []]], [[T.int64(3), T.int64(0)]]], [T.float64(2.767325769084694e-05)], {"thread_warp_size": T.int64(32), "host": {"mtriple": "arm64-apple-macos", "tag": "", "kind": "llvm", "mcpu": "apple-latest", "keys": ["arm_cpu", "cpu"]}, "max_threads_per_block": T.int64(1024), "max_function_args": T.int64(31), "max_num_threads": T.int64(256), "tag": "", "max_shared_memory_per_block": T.int64(32768), "kind": "metal", "keys": ["metal", "gpu"]}, [["TENSOR", "float32", [T.int64(1), T.int64(4), T.int64(64), T.int64(64)]], ["TENSOR", "float32", [T.int64(1), T.int64(4), T.int64(64), T.int64(64)]]]]]
The error message is:
[17:46:32] /Users/catalyst/Workspace/mlc-ai-package-self-runner/_work/package/package/tvm/src/meta_schedule/database/database.cc:167: ValueError: Unable to parse the JSON object: [[[["GetBlock", [], ["T_multiply", "main"], ["b0"]], ["GetLoops", ["b0"], [], ["l1", "l2", "l3", "l4"]], ["Fuse", ["l1", "l2", "l3", "l4"], [T.int64(1)], ["l5"]], ["SampleCategorical", [], [[T.int64(32), T.int64(64), T.int64(128), T.int64(256), T.int64(512), T.int64(1024)], [T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666), T.float64(0.16666666666666666)]], ["v6"]], ["Split", ["l5", "None", "v6"], [T.int64(1)], ["l7", "l8"]], ["Bind", ["l7"], ["blockIdx.x"], []], ["Bind", ["l8"], ["threadIdx.x"], []], ["EnterPostproc", [], [], []]], [[T.int64(3), T.int64(0)]]], [T.float64(2.767325769084694e-05)], {"thread_warp_size": T.int64(32), "host": {"mtriple": "arm64-apple-macos", "tag": "", "kind": "llvm", "mcpu": "apple-latest", "keys": ["arm_cpu", "cpu"]}, "max_threads_per_block": T.int64(1024), "max_function_args": T.int64(31), "max_num_threads": T.int64(256), "tag": "", "max_shared_memory_per_block": T.int64(32768), "kind": "metal", "keys": ["metal", "gpu"]}, [["TENSOR", "float32", [T.int64(1), T.int64(4), T.int64(64), T.int64(64)]], ["TENSOR", "float32", [T.int64(1), T.int64(4), T.int64(64), T.int64(64)]]]]
The error is: [17:46:32] /Users/catalyst/Workspace/mlc-ai-package-self-runner/_work/package/package/tvm/src/tir/schedule/primitive/.././instruction_traits.h:387: InternalError: Check failed: kNumAttrs == attrs.size() (2 vs. 1) : ValueError: Incorrect kNumAttrs for instruction: Split

Environments:
Mac M2
Python 3.10
mlc-ai-nightly 0.15.dev315

The report points to build.py:130 and log_db/database_tuning_record.json; begin by reproducing the failure with python3 build.py and inspecting how the cached TuningRecord is read. Done means the model build completes on the stated Mac M2, Python 3.10, and mlc-ai-nightly environment without the Split parsing error.