妙算3 能否支持 docker 或者升级 jetpack

[jkl375]

妙算3 目前只能用 dji 用户，用不了 root 权限，用 python3 install.py apt docker 后，发现 docker 并没有真正安装，而且自带的 tensorrt 版本过老，只有8.5.2，很多算子不支持，能否告知如何安装 并使用 docker

trtexec --onnx=turbo-lora-bs1.onnx --saveEngine=OmDetV2T_base_CXT_B_n104-fp16.trt --useCudaGraph
&&&& RUNNING TensorRT.trtexec [TensorRT v8502] # trtexec --onnx=turbo-lora-bs1.onnx --saveEngine=OmDetV2T_base_CXT_B_n104-fp16.trt --useCudaGraph
[01/04/2026-07:33:43] [I] === Model Options ===
[01/04/2026-07:33:43] [I] Format: ONNX
[01/04/2026-07:33:43] [I] Model: turbo-lora-bs1.onnx
[01/04/2026-07:33:43] [I] Output:
[01/04/2026-07:33:43] [I] === Build Options ===
[01/04/2026-07:33:43] [I] Max batch: explicit batch
[01/04/2026-07:33:43] [I] Memory Pools: workspace: default, dlaSRAM: default, dlaLocalDRAM: default, dlaGlobalDRAM: default
[01/04/2026-07:33:43] [I] minTiming: 1
[01/04/2026-07:33:43] [I] avgTiming: 8
[01/04/2026-07:33:43] [I] Precision: FP32
[01/04/2026-07:33:43] [I] LayerPrecisions: 
[01/04/2026-07:33:43] [I] Calibration: 
[01/04/2026-07:33:43] [I] Refit: Disabled
[01/04/2026-07:33:43] [I] Sparsity: Disabled
[01/04/2026-07:33:43] [I] Safe mode: Disabled
[01/04/2026-07:33:43] [I] DirectIO mode: Disabled
[01/04/2026-07:33:43] [I] Restricted mode: Disabled
[01/04/2026-07:33:43] [I] Build only: Disabled
[01/04/2026-07:33:43] [I] Save engine: OmDetV2T_base_CXT_B_n104-fp16.trt
[01/04/2026-07:33:43] [I] Load engine: 
[01/04/2026-07:33:43] [I] Profiling verbosity: 0
[01/04/2026-07:33:43] [I] Tactic sources: Using default tactic sources
[01/04/2026-07:33:43] [I] timingCacheMode: local
[01/04/2026-07:33:43] [I] timingCacheFile: 
[01/04/2026-07:33:43] [I] Heuristic: Disabled
[01/04/2026-07:33:43] [I] Preview Features: Use default preview flags.
[01/04/2026-07:33:43] [I] Input(s)s format: fp32:CHW
[01/04/2026-07:33:43] [I] Output(s)s format: fp32:CHW
[01/04/2026-07:33:43] [I] Input build shapes: model
[01/04/2026-07:33:43] [I] Input calibration shapes: model
[01/04/2026-07:33:43] [I] === System Options ===
[01/04/2026-07:33:43] [I] Device: 0
[01/04/2026-07:33:43] [I] DLACore: 
[01/04/2026-07:33:43] [I] Plugins:
[01/04/2026-07:33:43] [I] === Inference Options ===
[01/04/2026-07:33:43] [I] Batch: Explicit
[01/04/2026-07:33:43] [I] Input inference shapes: model
[01/04/2026-07:33:43] [I] Iterations: 10
[01/04/2026-07:33:43] [I] Duration: 3s (+ 200ms warm up)
[01/04/2026-07:33:43] [I] Sleep time: 0ms
[01/04/2026-07:33:43] [I] Idle time: 0ms
[01/04/2026-07:33:43] [I] Streams: 1
[01/04/2026-07:33:43] [I] ExposeDMA: Disabled
[01/04/2026-07:33:43] [I] Data transfers: Enabled
[01/04/2026-07:33:43] [I] Spin-wait: Disabled
[01/04/2026-07:33:43] [I] Multithreading: Disabled
[01/04/2026-07:33:43] [I] CUDA Graph: Enabled
[01/04/2026-07:33:43] [I] Separate profiling: Disabled
[01/04/2026-07:33:43] [I] Time Deserialize: Disabled
[01/04/2026-07:33:43] [I] Time Refit: Disabled
[01/04/2026-07:33:43] [I] NVTX verbosity: 0
[01/04/2026-07:33:43] [I] Persistent Cache Ratio: 0
[01/04/2026-07:33:43] [I] Inputs:
[01/04/2026-07:33:43] [I] === Reporting Options ===
[01/04/2026-07:33:43] [I] Verbose: Disabled
[01/04/2026-07:33:43] [I] Averages: 10 inferences
[01/04/2026-07:33:43] [I] Percentiles: 90,95,99
[01/04/2026-07:33:43] [I] Dump refittable layers:Disabled
[01/04/2026-07:33:43] [I] Dump output: Disabled
[01/04/2026-07:33:43] [I] Profile: Disabled
[01/04/2026-07:33:43] [I] Export timing to JSON file: 
[01/04/2026-07:33:43] [I] Export output to JSON file: 
[01/04/2026-07:33:43] [I] Export profile to JSON file: 
[01/04/2026-07:33:43] [I] 
[01/04/2026-07:33:43] [I] === Device Information ===
[01/04/2026-07:33:43] [I] Selected Device: Orin
[01/04/2026-07:33:43] [I] Compute Capability: 8.7
[01/04/2026-07:33:43] [I] SMs: 8
[01/04/2026-07:33:43] [I] Compute Clock Rate: 0.918 GHz
[01/04/2026-07:33:43] [I] Device Global Memory: 14834 MiB
[01/04/2026-07:33:43] [I] Shared Memory per SM: 164 KiB
[01/04/2026-07:33:43] [I] Memory Bus Width: 256 bits (ECC disabled)
[01/04/2026-07:33:43] [I] Memory Clock Rate: 0.918 GHz
[01/04/2026-07:33:43] [I] 
[01/04/2026-07:33:43] [I] TensorRT version: 8.5.2
[01/04/2026-07:33:43] [I] [TRT] [MemUsageChange] Init CUDA: CPU +220, GPU +0, now: CPU 249, GPU 3426 (MiB)
[01/04/2026-07:33:46] [I] [TRT] [MemUsageChange] Init builder kernel library: CPU +302, GPU +432, now: CPU 574, GPU 3877 (MiB)
[01/04/2026-07:33:46] [I] Start parsing network model
[01/04/2026-07:33:46] [I] [TRT] ----------------------------------------------------------------
[01/04/2026-07:33:46] [I] [TRT] Input filename:   turbo-lora-bs1.onnx
[01/04/2026-07:33:46] [I] [TRT] ONNX IR version:  0.0.8
[01/04/2026-07:33:46] [I] [TRT] Opset version:    17
[01/04/2026-07:33:46] [I] [TRT] Producer name:    pytorch
[01/04/2026-07:33:46] [I] [TRT] Producer version: 2.2.0
[01/04/2026-07:33:46] [I] [TRT] Domain:           
[01/04/2026-07:33:46] [I] [TRT] Model version:    0
[01/04/2026-07:33:46] [I] [TRT] Doc string:       
[01/04/2026-07:33:46] [I] [TRT] ----------------------------------------------------------------
[01/04/2026-07:33:47] [W] [TRT] onnx2trt_utils.cpp:375: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[01/04/2026-07:33:47] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[01/04/2026-07:33:47] [I] [TRT] No importer registered for op: LayerNormalization. Attempting to import as plugin.
[01/04/2026-07:33:47] [I] [TRT] Searching for plugin: LayerNormalization, plugin_version: 1, plugin_namespace: 
[01/04/2026-07:33:47] [E] [TRT] ModelImporter.cpp:726: While parsing node number 83 [LayerNormalization -> "/model/backbone/stages.0.0/norm/LayerNormalization_output_0"]:
[01/04/2026-07:33:47] [E] [TRT] ModelImporter.cpp:727: --- Begin node ---
[01/04/2026-07:33:47] [E] [TRT] ModelImporter.cpp:728: input: "/model/backbone/stages.0.0/Transpose_output_0"
input: "model.backbone.stages.0.0.norm.weight"
input: "model.backbone.stages.0.0.norm.bias"
output: "/model/backbone/stages.0.0/norm/LayerNormalization_output_0"
name: "/model/backbone/stages.0.0/norm/LayerNormalization"
op_type: "LayerNormalization"
attribute {
  name: "axis"
  i: -1
  type: INT
}
attribute {
  name: "epsilon"
  f: 1e-06
  type: FLOAT
}
doc_string: "/usr/local/lib/python3.10/dist-packages/torch/nn/functional.py(2548): layer_norm\n/workspace/omdet/modeling/backbone/convnext.py(99): forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1500): _slow_forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1519): _call_impl\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1510): _wrapped_call_impl\n/workspace/omdet/modeling/backbone/convnext.py(64): forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1500): _slow_forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1519): _call_impl\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1510): _wrapped_call_impl\n/workspace/omdet/modeling/backbone/convnext.py(253): forward_features\n/workspace/omdet/modeling/backbone/convnext.py(413): forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1500): _slow_forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1519): _call_impl\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1510): _wrapped_call_impl\n/workspace/omdet/omdet_v2_turbo/detector.py(677): forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1500): _slow_forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1519): _call_impl\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1510): _wrapped_call_impl\n/workspace/detectron2/detectron2/export/flatten.py(259): <lambda>\n/workspace/detectron2/detectron2/export/flatten.py(294): forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1500): _slow_forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1519): _call_impl\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1510): _wrapped_call_impl\n/usr/local/lib/python3.10/dist-packages/torch/jit/_trace.py(129): wrapper\n/usr/local/lib/python3.10/dist-packages/torch/jit/_trace.py(138): forward\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1519): _call_impl\n/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py(1510): _wrapped_call_impl\n/usr/local/lib/python3.10/dist-packages/torch/jit/_trace.py(1291): _get_trace_graph\n/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py(17): inner\n/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py(410): _fn\n/usr/local/lib/python3.10/dist-packages/torch/onnx/utils.py(915): _trace_and_get_graph_from_model\n/usr/local/lib/python3.10/dist-packages/torch/onnx/utils.py(1011): _create_jit_graph\n/usr/local/lib/python3.10/dist-packages/torch/onnx/utils.py(1135): _model_to_graph\n/usr/local/lib/python3.10/dist-packages/torch/onnx/utils.py(1613): _export\n/usr/local/lib/python3.10/dist-packages/torch/onnx/utils.py(516): export\n/workspace/omdet/infernece/det_engine.py(304): export_onnx\n/workspace/export.py(158): <module>\n"

[01/04/2026-07:33:47] [E] [TRT] ModelImporter.cpp:729: --- End node ---
[01/04/2026-07:33:47] [E] [TRT] ModelImporter.cpp:731: ERROR: builtin_op_importers.cpp:5427 In function importFallbackPluginImporter:
[8] Assertion failed: creator && "Plugin not found, are the plugin name, version, and namespace correct?"
[01/04/2026-07:33:47] [E] Failed to parse onnx file
[01/04/2026-07:33:47] [I] Finish parsing network model
[01/04/2026-07:33:47] [E] Parsing model failed
[01/04/2026-07:33:47] [E] Failed to create engine from model or file.
[01/04/2026-07:33:47] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v8502] # trtexec --onnx=turbo-lora-bs1.onnx --saveEngine=OmDetV2T_base_CXT_B_n104-fp16.trt --useCudaGraph