Error When Use Tensorcore

Questions
#1

When I use TVM to infer the VIT network, the generated dense_tensorcore.cuda operator encounter errors when building CUDA code:error: incomplete type is not allowed. This is my error report：

屏幕截图 2023-11-05 201537
屏幕截图 2023-11-05 2015371684×946 106 KB
How can I solve this problem

#2

It’s better to provide the steps that reproduces the error.

#3 
with tvm.relay.quantize.qconfig(calibrate_mode="global_scale", global_scale=8.0, weight_scale = "power2", skip_dense_layer=False, skip_conv_layers=[0]):
    mod = tvm.relay.quantize.quantize(mod, params, dataset=dataset)


number = 20
repeat = 3
min_repeat_ms = 4  
timeout = 150  


runner = autotvm.LocalRunner(
    number=number,
    repeat=repeat,
    timeout=timeout,
    min_repeat_ms=min_repeat_ms,
    enable_cpu_cache_flush=True,
)

tuning_option = {
    "tuner": "xgb",
    "trials": 2000,
    "early_stopping": 600,
    "measure_option": autotvm.measure_option(
        builder=autotvm.LocalBuilder(build_func="default"), runner=runner
    ),
    "tuning_records": "vit_B32_224-autotuning_n.json",
}


tasks = autotvm.task.extract_from_program(mod["main"], target=target, params=params)

for i, task in enumerate(reversed(tasks)):
    prefix = "[Task %2d/%2d] " % (i + 1, len(tasks))
    tuner_obj = XGBTuner(task, loss_type="rank")
    tuner_obj.tune(
        n_trial=min(tuning_option["trials"], len(task.config_space)),
        early_stopping=tuning_option["early_stopping"],
        measure_option=tuning_option["measure_option"],
        callbacks=[
            autotvm.callback.progress_bar(tuning_option["trials"], prefix=prefix),
            autotvm.callback.log_to_file(tuning_option["tuning_records"]),
        ],
    )
#4

I tuned a quantified VIT net and encountered an error while running it

#5

I use test_quantization_accuracy_for_vit.py to run it