打开微信,使用扫一扫进入页面后,点击右上角菜单,
点击“发送给朋友”或“分享到朋友圈”完成分享
【寒武纪硬件产品型号】必填*:
必填项,例如:MLU370
MLU290
【MagicMind版本号】必填*:
magicmind_1.3.0-1_centos_7.6
【出错信息】必填*:
python3 predict_on_resnet50.py --path_mm_model ../cn_mm2/resnet50_parser.mm
Can not write to dir, file log disabled: /var/log
2023-06-25 19:38:30.990588: [cnrtError] [145311] [Card : 0] cnpapi__cnrtInvokeKernel: Found kernel(_Z25MLUTransposeKernel2DSmallIiLj1EEvPvS0_jjb) but not load(101315).
[2023-6-25 19:38:30] [CNNL] [Error]:Check failed: Found cncc does not register the module to runtime. after invoke kernel (MLUTransposeKernel2DSmall<T, REPEAT><<<kDim, kType, handle_->queue>>>( (void *)x, y, x_kernel[TR_N], x_kernel[TR_H], split_h))
2023-06-25 19:38:30.990938: [cnrtError] [145311] [Card : 0] cnpapi__cnrtInvokeKernel: Found kernel(_Z25quantizeParamGetMaxAndMinPvj14cnnlDataType_tiiS_) but not load(101315).
[2023-6-25 19:38:30] [CNNL] [Error]:Check failed: Found cncc does not register the module to runtime. after invoke kernel (quantizeParamGetMaxAndMin<<<kDim, kType, handle->queue>>>( (void *)input, dim_input, input_data_type, bitwidth, job_num, workspace))
2023-06-25 19:38:30.990999: ERROR: magicmind/runtime/kernel/cnnl_kernel/conv/float_conv.cc:115] Call cnnl function cnnlQuantizeParam(handle, CNNL_QUANTIZE_POSITION, x_desc, x_ptr, kVars::kBitWidthInt31, workspace, workspace_size, x_position_c, x_scale, x_offset) failed.
2023-06-25 19:38:30.991022: [cnrtError] [145311] [Card : 0] cnpapi__cnrtInvokeKernel: Found kernel(_Z25quantizeParamGetMaxAndMinPvj14cnnlDataType_tiiS_) but not load(101315).
[2023-6-25 19:38:30] [CNNL] [Error]:Check failed: Found cncc does not register the module to runtime. after invoke kernel (quantizeParamGetMaxAndMin<<<kDim, kType, handle->queue>>>( (void *)input, dim_input, input_data_type, bitwidth, job_num, workspace))
2023-06-25 19:38:30.991071: ERROR: magicmind/runtime/kernel/cnnl_kernel/conv/float_conv.cc:115] Call cnnl function cnnlQuantizeParam(handle, CNNL_QUANTIZE_POSITION, x_desc, x_ptr, kVars::kBitWidthInt31, workspace, workspace_size, x_position_c, x_scale, x_offset) failed.
2023-06-25 19:38:30.991154: ERROR: magicmind/runtime/core/mlu_device.cc:227]
Kernel Enqueue error:
Internal: cnnlFloatConvForward( handle_, conv_desc_, algo_, nullptr, input_, input_ptr_, weight_, weight_ptr_, bias_, bias_ptr_, workspace_, workspace_size_, nullptr, output_, output_ptr_, is_float_):Call cnnl function Failed, error code: CNNL_STATUS_EXECUTION_FAILED
In queue id 0 pointer 0x8427ec0
MLU device : MLU:0_0 in context ctx_id_
main/mm.conv2d/mm.conv2d/cnnl.conv : {
input {
cnnlTensorDe or {
Layout: NHWC
DataType: FLOAT32
DimNb: 4
OnchipDtype: INT31
Pos: 0
Scale: 1.000000
Offset: 0
1 256 256 3
}
ptr:
0x10080ffefc00000
}
weight {
cnnlTensorDe or {
Layout: NHWC
DataType: FLOAT32
DimNb: 4
OnchipDtype: INT31
Pos: 0
Scale: 1.000000
Offset: 0
64 7 7 3
}
ptr:
0x10080fff6140000
}
bias {
cnnlTensorDe or {
Layout: NHWC
DataType: FLOAT32
DimNb: 4
OnchipDtype: Unknown
Pos: 0
Scale: 1.000000
Offset: 0
1 1 1 64
}
ptr:
0x10080fff6163100
}
position_input {
ptr:
0
}
position_weight {
ptr:
0
}
output {
cnnlTensorDe or {
Layout: NHWC
DataType: FLOAT32
DimNb: 4
OnchipDtype: FLOAT32
Pos: 0
Scale: 1.000000
Offset: 0
1 128 128 64
}
ptr:
0x10080ffee800000
}
padding_ {
EXPLICIT
}
pad_ {
[3, 3, 3, 3]
}
stride_ {
[2, 2]
}
dilation_ {
[1, 1]
}
group_ {
1
}
layouts_ {
[NHWC, NHWC, ARRAY, NHWC]
}
runtime context {
workspace :0x10080ffef000000 with expected size 824096
with recorded size 4194304
input_zero_element_ {
0
}
compute_type_ {
FLOAT
}
onchip_type_ {
[]
}
2023-06-25 19:38:30.991725: ERROR: magicmind/runtime/executor/executor.cc:851] status == Status::OK() is false.
2023-06-25 19:38:30.991744: ERROR: magicmind/runtime/executor/function_runner.cc:484] Status::OK() == ret is false.
Traceback (most recent call last):
File "predict_on_resnet50.py", line 84, in <module>
main(args)
File "predict_on_resnet50.py", line 64, in main
assert context.enqueue(inputs, outputs, queue).ok()
Asserti
【操作步骤】选填:
【相关日志文档】选填
如有,可附件
【出错代码链接】选填:
github的或gitee的代码的链接,
热门帖子
精华帖子