pp_liteseg用paddle inference推理时，采用TensorRT加速，已开启use_trt_dynamic_shape

[gitleej]

问题确认 Search before asking

• 我已经搜索过问题，但是没有找到解答。I have searched the question and found no related answer.

请提出你的问题 Please ask your question

use_trt_dynamic_shape设置
if (FLAGS_dynamic_shape_path.empty()) {
std::map<std::string, std::vector> min_input_shape = {
{"x", {1, 3, 1, 1}},
{"image", {1, 3, 112, 112}},
{"bilinear_interp_v2_5.tmp_0", {1, 3, 1, 1}},
{"bilinear_interp_v2_0.tmp_0", {1, 3, 1, 1}},
{"bilinear_interp_v2_1.tmp_0", {1, 3, 1, 1}},
{"bilinear_interp_v2_2.tmp_0", {1, 3, 1, 1}},
{"max_4.tmp_0", {1, 3, 1, 1}},
{"max_5.tmp_0", {1, 3, 1, 1}},
{"mean_4.tmp_0", {1, 3, 1, 1}},
{"relu_36.tmp_0", {1, 3, 1, 1}},
{"bilinear_interp_v2_4.tmp_0", {1, 3, 1, 1}},
{"max_2.tmp_0", {1, 3, 1, 1}},
{"max_3.tmp_0", {1, 3, 1, 1}},
{"mean_2.tmp_0", {1, 3, 1, 1}},
{"relu_33.tmp_0", {1, 3, 1, 1}},
{"bilinear_interp_v2_3.tmp_0", {1, 3, 1, 1}},
{"max_0.tmp_0", {1, 3, 1, 1}},
{"max_1.tmp_0", {1, 3, 1, 1}},
{"mean_0.tmp_0", {1, 3, 1, 1}},
{"relu_30.tmp_0", {1, 3, 1, 1}}};
std::map<std::string, std::vector> max_input_shape = {
{"x", {1, 3, 1024, 2048}},
{"image", {1, 3, 1024, 2048}},
{"bilinear_interp_v2_5.tmp_0", {1, 3, 1024, 2048}},
{"bilinear_interp_v2_0.tmp_0", {1, 3, 1024, 2048}},
{"bilinear_interp_v2_1.tmp_0", {1, 3, 1024, 2048}},
{"bilinear_interp_v2_2.tmp_0", {1, 3, 1024, 2048}},
{"max_4.tmp_0", {1, 3, 1024, 2048}},
{"max_5.tmp_0", {1, 3, 1024, 2048}},
{"mean_4.tmp_0", {1, 3, 1024, 2048}},
{"relu_36.tmp_0", {1, 3, 1024, 2048}},
{"bilinear_interp_v2_4.tmp_0", {1, 3, 1024, 2048}},
{"max_2.tmp_0", {1, 3, 1024, 2048}},
{"max_3.tmp_0", {1, 3, 1024, 2048}},
{"mean_2.tmp_0", {1, 3, 1024, 2048}},
{"relu_33.tmp_0", {1, 3, 1024, 2048}},
{"bilinear_interp_v2_3.tmp_0", {1, 3, 1024, 2048}},
{"max_0.tmp_0", {1, 3, 1024, 2048}},
{"max_1.tmp_0", {1, 3, 1024, 2048}},
{"mean_0.tmp_0", {1, 3, 1024, 2048}},
{"relu_30.tmp_0", {1, 3, 1024, 2048}}};
std::map<std::string, std::vector> opt_input_shape = {
{"x", {1, 3, 512, 1024}},
{"image", {1, 3, 512, 1024}},
{"bilinear_interp_v2_5.tmp_0", {1, 3, 512, 1024}},
{"bilinear_interp_v2_0.tmp_0", {1, 3, 512, 1024}},
{"bilinear_interp_v2_1.tmp_0", {1, 3, 512, 1024}},
{"bilinear_interp_v2_2.tmp_0", {1, 3, 512, 1024}},
{"max_4.tmp_0", {1, 3, 512, 1024}},
{"max_5.tmp_0", {1, 3, 512, 1024}},
{"mean_4.tmp_0", {1, 3, 512, 1024}},
{"relu_36.tmp_0", {1, 3, 512, 1024}},
{"bilinear_interp_v2_4.tmp_0", {1, 3, 256, 512}},
{"max_2.tmp_0", {1, 3, 256, 512}},
{"max_3.tmp_0", {1, 3, 256, 512}},
{"mean_2.tmp_0", {1, 3, 256, 512}},
{"relu_33.tmp_0", {1, 3, 256, 512}},
{"bilinear_interp_v2_3.tmp_0", {1, 3, 256, 512}},
{"max_0.tmp_0", {1, 3, 256, 512}},
{"max_1.tmp_0", {1, 3, 256, 512}},
{"mean_0.tmp_0", {1, 3, 256, 512}},
{"relu_30.tmp_0", {1, 3, 256, 512}}};
infer_config.SetTRTDynamicShapeInfo(
min_input_shape, max_input_shape, opt_input_shape);

WARNING: Logging before InitGoogleLogging() is written to STDERR I0710 15:28:49.381031 18480 main.cpp:94] Use GPU I0710 15:28:49.383030 18480 main.cpp:99] Use TRT I0710 15:28:49.383030 18480 main.cpp:100] trt_precision:fp32 I0710 15:28:49.383030 18480 main.cpp:121] Enable TRT dynamic shape WARNING: Logging before InitGoogleLogging() is written to STDERR I0710 15:28:52.924093 18480 analysis_predictor.cc:1099] TensorRT subgraph engine is enabled e[1me[35m--- Running analysis [ir_graph_build_pass]e[0m e[1me[35m--- Running analysis [ir_analysis_pass]e[0m e[32m--- Running IR pass [adaptive_pool2d_convert_global_pass]e[0m I0710 15:28:52.953110 18480 fuse_pass_base.cc:59] --- detected 1 subgraphs e[32m--- Running IR pass [shuffle_channel_detect_pass]e[0m e[32m--- Running IR pass [quant_conv2d_dequant_fuse_pass]e[0m e[32m--- Running IR pass [delete_fill_constant_op_pass]e[0m e[32m--- Running IR pass [delete_quant_dequant_op_pass]e[0m e[32m--- Running IR pass [delete_quant_dequant_filter_op_pass]e[0m e[32m--- Running IR pass [delete_weight_dequant_linear_op_pass]e[0m e[32m--- Running IR pass [delete_quant_dequant_linear_op_pass]e[0m e[32m--- Running IR pass [identity_scale_op_clean_pass]e[0m e[32m--- Running IR pass [add_support_int8_pass]e[0m I0710 15:28:52.994058 18480 fuse_pass_base.cc:59] --- detected 200 subgraphs e[32m--- Running IR pass [simplify_with_basic_ops_pass]e[0m e[32m--- Running IR pass [trt_embedding_eltwise_layernorm_fuse_pass]e[0m e[32m--- Running IR pass [preln_embedding_eltwise_layernorm_fuse_pass]e[0m e[32m--- Running IR pass [delete_c_identity_op_pass]e[0m e[32m--- Running IR pass [trt_multihead_matmul_fuse_pass_v2]e[0m e[32m--- Running IR pass [trt_multihead_matmul_fuse_pass_v3]e[0m e[32m--- Running IR pass [vit_attention_fuse_pass]e[0m e[32m--- Running IR pass [trt_skip_layernorm_fuse_pass]e[0m e[32m--- Running IR pass [preln_skip_layernorm_fuse_pass]e[0m e[32m--- Running IR pass [preln_residual_bias_fuse_pass]e[0m e[32m--- Running IR pass [layernorm_shift_partition_fuse_pass]e[0m e[32m--- Running IR pass [conv_bn_fuse_pass]e[0m I0710 15:28:53.060089 18480 fuse_pass_base.cc:59] --- detected 39 subgraphs e[32m--- Running IR pass [unsqueeze2_eltwise_fuse_pass]e[0m e[32m--- Running IR pass [trt_squeeze2_matmul_fuse_pass]e[0m e[32m--- Running IR pass [trt_flatten2_matmul_fuse_pass]e[0m e[32m--- Running IR pass [trt_map_matmul_v2_to_mul_pass]e[0m e[32m--- Running IR pass [trt_map_matmul_v2_to_matmul_pass]e[0m e[32m--- Running IR pass [trt_map_matmul_to_mul_pass]e[0m e[32m--- Running IR pass [fc_fuse_pass]e[0m e[32m--- Running IR pass [conv_elementwise_add_fuse_pass]e[0m I0710 15:28:53.081130 18480 fuse_pass_base.cc:59] --- detected 43 subgraphs e[32m--- Running IR pass [remove_padding_recover_padding_pass]e[0m e[32m--- Running IR pass [delete_remove_padding_recover_padding_pass]e[0m e[32m--- Running IR pass [dense_fc_to_sparse_pass]e[0m e[32m--- Running IR pass [dense_multihead_matmul_to_sparse_pass]e[0m e[32m--- Running IR pass [constant_folding_pass]e[0m e[32m--- Running IR pass [tensorrt_subgraph_pass]e[0m I0710 15:28:55.416604 18480 tensorrt_subgraph_pass.cc:244] --- detect a sub-graph with 15 nodes I0710 15:28:55.767565 18480 tensorrt_subgraph_pass.cc:560] Prepare TRT engine (Optimize model structure, Select OP kernel etc). This process may cost a lot of time. W0710 15:28:56.290634 18480 helper.h:110] Tensor DataType is determined at build time for tensors not marked as input or output. E0710 15:28:56.291595 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.291595 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.291595 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.291595 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.291595 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.292604 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.292604 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.292604 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.292604 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.293560 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.293560 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.293560 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.293560 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.293560 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.294612 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.294612 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.294612 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.294612 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.294612 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.294612 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.294612 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.295616 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.295616 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.295616 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.295616 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.295616 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.295616 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.296615 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.296615 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.296615 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.296615 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.296615 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.296615 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.297618 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.297618 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.297618 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.297618 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.297618 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.297618 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.297618 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.298615 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.298615 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) I0710 15:28:56.298615 18480 engine.cc:199] Run Paddle-TRT Dynamic Shape mode. E0710 15:28:56.298615 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.298615 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.299623 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.299623 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.299623 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.299623 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.299623 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.299623 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.299623 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.300616 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.300616 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.300616 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.300616 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.300616 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.300616 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.300616 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.301623 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.301623 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.301623 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.301623 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.301623 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.301623 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.302615 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.302615 18480 helper.h:114] 4: [convolutionNode.cpp::nvinfer1::builder::ConvolutionNode::computeOutputExtents::28] Error Code 4: Internal Error (conv2d (Output: batch_norm_42.tmp_2759): number of kernel weights does not match tensor dimensions) E0710 15:28:56.302615 18480 helper.h:114] 3: conv2d (Output: batch_norm_42.tmp_2759):kernel weights has count 72 but 180 was expected E0710 15:28:56.302615 18480 helper.h:114] 4: conv2d (Output: batch_norm_42.tmp_2759): count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1 = 180 E0710 15:28:56.302615 18480 helper.h:114] 4: [network.cpp::nvinfer1::Network::validate::2640] Error Code 4: Internal Error (Layer conv2d (Output: batch_norm_42.tmp_2759) failed validation) E0710 15:28:56.302615 18480 helper.h:114] 2: [builder.cpp::nvinfer1::builder::Builder::buildSerializedNetwork::417] Error Code 2: Internal Error (Assertion enginePtr != nullptr failed.)

[Asthestarsfalll]

能否有更清晰的报错提示和运行命令等？

[gitleej]

能否有更清晰的报错提示和运行命令等？

上面就是所有报错信息了，程序源码就是PaddleSeg套件里的cpp推理源码，命令--use_trt --use_dynamic_shape

[Asthestarsfalll]

number of kernel weights does not match tensor dimensions，count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1，请检查一下通道数

[gitleej]

number of kernel weights does not match tensor dimensions，count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1，请检查一下通道数

是模型的通道数和trt 加速max shape、min shape、opt shape要统一吗？或者max、min、opt shape的参数应该怎么设置？

[Asthestarsfalll]

number of kernel weights does not match tensor dimensions，count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1，请检查一下通道数

是模型的通道数和trt 加速max shape、min shape、opt shape要统一吗？或者max、min、opt shape的参数应该怎么设置？

请给出复现步骤，包括但不限于paddleseg版本、使用模型、命令等，我来看一下

[gitleej]

number of kernel weights does not match tensor dimensions，count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1，请检查一下通道数

是模型的通道数和trt 加速max shape、min shape、opt shape要统一吗？或者max、min、opt shape的参数应该怎么设置？

请给出复现步骤，包括但不限于paddleseg版本、使用模型、命令等，我来看一下

• paddleseg release/2.8
• model：pp_liteseg
• 命令：
  • 训练：

  python tools/train.py --config configs/pp_liteseg/pp_liteseg_stdc1_camvid_960x720_10k_for_test.yml --save_dir output/pp_liteseg_stdc1_camvid_960x720_10k_for_test --save_interval 1000 --num_workers 2 --do_eval --use_vdl

  • 导出模型

  python tools/export.py --config configs/pp_liteseg/pp_liteseg_stdc1_camvid_960x720_10k_for_test.yml --model_path output/pp_liteseg_stdc1_camvid_960x720_10k_for_test/best_model/model.pdparams --save_dir output/infer_model/pp_liteseg_stdc1_camvid_960x720_10k_for_test

[gitleej]

number of kernel weights does not match tensor dimensions，count of 72 weights in kernel, but kernel dimensions (3,3) with 10 input channels, 2 output channels and 1 groups were specified. Expected Weights count is 10 * 3*3 * 2 / 1，请检查一下通道数

是模型的通道数和trt 加速max shape、min shape、opt shape要统一吗？或者max、min、opt shape的参数应该怎么设置？

请给出复现步骤，包括但不限于paddleseg版本、使用模型、命令等，我来看一下

• paddleseg release/2.8

• model：pp_liteseg

• 命令：

  • 训练：

  python tools/train.py --config configs/pp_liteseg/pp_liteseg_stdc1_camvid_960x720_10k_for_test.yml --save_dir output/pp_liteseg_stdc1_camvid_960x720_10k_for_test --save_interval 1000 --num_workers 2 --do_eval --use_vdl

  • 导出模型

  python tools/export.py --config configs/pp_liteseg/pp_liteseg_stdc1_camvid_960x720_10k_for_test.yml --model_path output/pp_liteseg_stdc1_camvid_960x720_10k_for_test/best_model/model.pdparams --save_dir output/infer_model/pp_liteseg_stdc1_camvid_960x720_10k_for_test

问题解决了：
根据这里#2679 (comment) 导出dynamic_shape.pbtxt即可解决问题。

#导出模型
python tools/export.py --config configs/pp_liteseg/pp_liteseg_stdc1_medicine_board_840x450_10k_for_test.yml --model_path output/pp_liteseg_stdc1_medicine_board_840x450_10k_for_test_lr_01/best_model/model.pdparams --save_dir output/infer_model/pp_liteseg_stdc1_medicine_board_840x450_10k_for_test_lr_01
#导出dynamic_shape.pbtxt
python deploy/python/collect_dynamic_shape.py --config output/infer_model/pp_liteseg_stdc1_medicine_board_840x450_10k_for_test_lr_01/deploy.yaml --image_path data/02-Medicine_Board/JPEGImages/10.bmp --dynamic_shape_path output/infer_model/pp_liteseg_stdc1_medicine_board_840x450_10k_for_test_lr_01/dynamic_shape.pbtxt

C++推理时，命令行参数加上

--use_trt --use_dynamic_shape --dynamic_shape_path=path/to/dynamic_shape.pbtxt

[ToddBear]

以上回答已经充分解答了问题，如果有新的问题欢迎随时提交issue，或者在此条issue下继续回复～
我们开启了飞桨套件的ISSUE攻关活动，欢迎感兴趣的开发者参加：PaddlePaddle/PaddleOCR#10223