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
硬件平台为MLU370 s4,使用mmyolo中的easydeploy导出onnx,之后使用onnx_build工具时报错,以下是所有输出,不知是否与nms也被转为onnx有关:
2023-07-28 15:32:33.674027: INFO: /home/kyxz/magicmind/mm_build/main.cc:13]
==================== Parameter Information
--precision<Str> :
--calibration<Bool> : false
--rpc_server<Str> :
--calibration_algo<Str> : linear
--file_list<Vec<Str>> :
--calibration_data_path<Str> :
--random_calib_range<Vec<Float>>:
--batch_size<Vec<Int>> :
--input_dims<Vec<Vec<Int>>> :
--input_layout<Vec<Str>> :
--input_dtypes<Vec<Str>> : FLOAT
--dynamic_shape<Bool> : true
--vars<Vec<Vec<Float>>> :
--means<Vec<Vec<Float>>> :
--output_layout<Vec<Str>> :
--output_dtypes<Vec<Str>> :
--mlu_arch<Vec<Str>> :
--mlu_product_name<Vec<Str>> :
--plugin<Vec<Str>> :
--magicmind_model<Str> : ./model
--build_config<Str> :
--toolchain_path<Str> : /usr/local/gcc-linaro-6.2.1-2016.11-x86_64_aarch64-linux-gnu/
--rgb2bgr<Bool> : false
--cluster_num<Vec<Vec<Int>>> :
--onnx<Str> : ./rtmdet.onnx
MagicMind: 1.5.0
2023-07-28 15:32:35.296165: WARNING: magicmind/conversion/parser/onnx2mm/lower/tensor_ops.cc:316] Attributes: cubic_coeff_a, exclude_outside and extrapolation_value are not supported in magicmind now!
2023-07-28 15:32:35.296208: WARNING: magicmind/conversion/parser/onnx2mm/lower/tensor_ops.cc:322] ROI value is used for mode: tf_crop_and resize, but mode: tf_crop_and_resize is not supported in MagicMind now!
2023-07-28 15:32:35.299926: WARNING: magicmind/conversion/parser/onnx2mm/lower/tensor_ops.cc:316] Attributes: cubic_coeff_a, exclude_outside and extrapolation_value are not supported in magicmind now!
2023-07-28 15:32:35.299964: WARNING: magicmind/conversion/parser/onnx2mm/lower/tensor_ops.cc:322] ROI value is used for mode: tf_crop_and resize, but mode: tf_crop_and_resize is not supported in MagicMind now!
2023-07-28 15:32:35.706265: INFO: /home/kyxz/magicmind/mm_build/main_process.cc:385] Reset input dtypes.
2023-07-28 15:32:35.707928: INFO: magicmind/dialect/common/dim_range_utils.cc:1267] Input tensor(0)'s shape ([1, 3, 640, 640]) will be overwritten by dim range set on build config ([-1, -1, -1, -1])
loc("main/mm.neg/cnnl.neg"): error: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:35.894233: ERROR: magicmind/dialect/mm/transforms/mm_const_fold.cc:251] Run pass pipeline failed!
loc("main/Neg_631/mm.neg"): error: Const fold failed.
2023-07-28 15:32:35.894436: INFO: magicmind/conversion/pass_manager.cc:173] Pipeline dumping .//compile_graph/process_shape_infer_and_fold2/0_5_MMConstFold
2023-07-28 15:32:37.571327: ERROR: magicmind/dialect/mm/transforms/mm_shape_infer_and_fold.cc:56] Failed to run pass pipeline: MMConstFold: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:37.571363: ERROR: magicmind/dialect/mm/transforms/mm_shape_infer_and_fold.cc:56] Failed to run pass pipeline: MMShapeInferAndFold: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:37.571373: ERROR: magicmind/dialect/mm/transforms/mm_shape_infer_and_fold.cc:56] Failed to run pass pipeline: Verify: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:37.571380: ERROR: magicmind/dialect/mm/transforms/mm_shape_infer_and_fold.cc:56] Failed to run pass pipeline: MMConstFold: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.mm.neg
2023-07-28 15:32:37.571386: ERROR: magicmind/dialect/mm/transforms/mm_shape_infer_and_fold.cc:56] Failed to run pass pipeline: MMShapeInferAndFold: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.mm.neg
2023-07-28 15:32:37.571391: ERROR: magicmind/dialect/mm/transforms/mm_shape_infer_and_fold.cc:56] Failed to run pass pipeline: Verify: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.
loc("builtin.module"): error: Failures have been detected while processing an MLIR pass pipeline
2023-07-28 15:32:40.862648: INFO: magicmind/conversion/pass_manager.cc:173] Pipeline func.func(per-axis-qcast-conversion) dumping .//compile_graph/process_main_process_optimize1/2_4_MMShapeInferAndFold
2023-07-28 15:32:42.467898: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: MMConstFold: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:42.467935: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: MMShapeInferAndFold: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:42.467942: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: Verify: 'cnnl.neg' op Input type must be Int32 or FloatType
2023-07-28 15:32:42.467949: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: MMConstFold: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.mm.neg
2023-07-28 15:32:42.467955: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: MMShapeInferAndFold: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.mm.neg
2023-07-28 15:32:42.467961: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: Verify: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.
2023-07-28 15:32:42.467966: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: MMShapeInferAndFold: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.Failures have been detected while processing an MLIR pass pipelinebuiltin.module
2023-07-28 15:32:42.467973: ERROR: magicmind/cc/implement/processor.cc:211] Failed to run pass pipeline: Verify: 'cnnl.neg' op Input type must be Int32 or FloatTypeConst fold failed.Failures have been detected while processing an MLIR pass pipeline
2023-07-28 15:32:42.469622: ERROR: magicmind/cc/implement/processor.cc:445] status(Status) error:Internal: Run pass pipeline failed!.
2023-07-28 15:32:42.469642: ERROR: magicmind/cc/implement/interface_builder_impl.cc:118] Build model failed!
2023-07-28 15:32:42.470435: ERROR: /home/kyxz/magicmind/mm_build/main_process.cc:417] BuildModel failed
2023-07-28 15:32:42.470445: ERROR: /home/kyxz/magicmind/mm_build/../mm_build/main_process.h:37] BuildAndSerialize(net, config, param) is null or 0
已放弃 (核心已转储)
请问应该如何解决?
热门帖子
精华帖子