Failed to parse onnx file #4587
Description
&&&& RUNNING TensorRT.trtexec [TensorRT v101303] [b9] # ./trtexec --onnx=ISTD_NS.onnx --saveEngine=ISTD_N.trt --iterations=10 --precisionConstraints=obey
[09/28/2025-09:21:51] [I] TF32 is enabled by default. Add --noTF32 flag to further improve accuracy with some performance cost.
[09/28/2025-09:21:51] [W] When --precisionConstraints flag is set to "obey" or "prefer", please add --layerPrecision/--layerOutputTypes flags to set layer-wise precisions and output types.
[09/28/2025-09:21:51] [I] === Model Options ===
[09/28/2025-09:21:51] [I] Format: ONNX
[09/28/2025-09:21:51] [I] Model: ISTD_NS.onnx
[09/28/2025-09:21:51] [I] Output:
[09/28/2025-09:21:51] [I] === Build Options ===
[09/28/2025-09:21:51] [I] Memory Pools: workspace: default, dlaSRAM: default, dlaLocalDRAM: default, dlaGlobalDRAM: default, tacticSharedMem: default
[09/28/2025-09:21:51] [I] avgTiming: 8
[09/28/2025-09:21:51] [I] Precision: FP32 (obey precision constraints)
[09/28/2025-09:21:51] [I] LayerPrecisions:
[09/28/2025-09:21:51] [I] Layer Device Types:
[09/28/2025-09:21:51] [I] Calibration:
[09/28/2025-09:21:51] [I] Refit: Disabled
[09/28/2025-09:21:51] [I] Strip weights: Disabled
[09/28/2025-09:21:51] [I] Version Compatible: Disabled
[09/28/2025-09:21:51] [I] ONNX Plugin InstanceNorm: Disabled
[09/28/2025-09:21:51] [I] ONNX kENABLE_UINT8_AND_ASYMMETRIC_QUANTIZATION_DLA flag: Disabled
[09/28/2025-09:21:51] [I] TensorRT runtime: full
[09/28/2025-09:21:51] [I] Lean DLL Path:
[09/28/2025-09:21:51] [I] Tempfile Controls: { in_memory: allow, temporary: allow }
[09/28/2025-09:21:51] [I] Exclude Lean Runtime: Disabled
[09/28/2025-09:21:51] [I] Sparsity: Disabled
[09/28/2025-09:21:51] [I] Safe mode: Disabled
[09/28/2025-09:21:51] [I] Build DLA standalone loadable: Disabled
[09/28/2025-09:21:51] [I] Allow GPU fallback for DLA: Disabled
[09/28/2025-09:21:51] [I] DirectIO mode: Disabled
[09/28/2025-09:21:51] [I] Restricted mode: Disabled
[09/28/2025-09:21:51] [I] Skip inference: Disabled
[09/28/2025-09:21:51] [I] Save engine: ISTD_N.trt
[09/28/2025-09:21:51] [I] Load engine:
[09/28/2025-09:21:51] [I] Profiling verbosity: 0
[09/28/2025-09:21:51] [I] Tactic sources: Using default tactic sources
[09/28/2025-09:21:51] [I] timingCacheMode: local
[09/28/2025-09:21:51] [I] timingCacheFile:
[09/28/2025-09:21:51] [I] Enable Compilation Cache: Enabled
[09/28/2025-09:21:51] [I] Enable Monitor Memory: Disabled
[09/28/2025-09:21:51] [I] errorOnTimingCacheMiss: Disabled
[09/28/2025-09:21:51] [I] Preview Features: Use default preview flags.
[09/28/2025-09:21:51] [I] MaxAuxStreams: -1
[09/28/2025-09:21:51] [I] BuilderOptimizationLevel: -1
[09/28/2025-09:21:51] [I] MaxTactics: -1
[09/28/2025-09:21:51] [I] Calibration Profile Index: 0
[09/28/2025-09:21:51] [I] Weight Streaming: Disabled
[09/28/2025-09:21:51] [I] Runtime Platform: Same As Build
[09/28/2025-09:21:51] [I] Debug Tensors:
[09/28/2025-09:21:51] [I] Distributive Independence: Disabled
[09/28/2025-09:21:51] [I] Mark Unfused Tensors As Debug Tensors: Disabled
[09/28/2025-09:21:51] [I] Input(s)s format: fp32:CHW
[09/28/2025-09:21:51] [I] Output(s)s format: fp32:CHW
[09/28/2025-09:21:51] [I] Input build shapes: model
[09/28/2025-09:21:51] [I] Input calibration shapes: model
[09/28/2025-09:21:51] [I] === System Options ===
[09/28/2025-09:21:51] [I] Device: 0
[09/28/2025-09:21:51] [I] DLACore:
[09/28/2025-09:21:51] [I] Plugins:
[09/28/2025-09:21:51] [I] setPluginsToSerialize:
[09/28/2025-09:21:51] [I] dynamicPlugins:
[09/28/2025-09:21:51] [I] ignoreParsedPluginLibs: 0
[09/28/2025-09:21:51] [I]
[09/28/2025-09:21:51] [I] === Inference Options ===
[09/28/2025-09:21:51] [I] Batch: Explicit
[09/28/2025-09:21:51] [I] Input inference shapes: model
[09/28/2025-09:21:51] [I] Iterations: 10
[09/28/2025-09:21:51] [I] Duration: 3s (+ 200ms warm up)
[09/28/2025-09:21:51] [I] Sleep time: 0ms
[09/28/2025-09:21:51] [I] Idle time: 0ms
[09/28/2025-09:21:51] [I] Inference Streams: 1
[09/28/2025-09:21:51] [I] ExposeDMA: Disabled
[09/28/2025-09:21:51] [I] Data transfers: Enabled
[09/28/2025-09:21:51] [I] Spin-wait: Disabled
[09/28/2025-09:21:51] [I] Multithreading: Disabled
[09/28/2025-09:21:51] [I] CUDA Graph: Disabled
[09/28/2025-09:21:51] [I] Separate profiling: Disabled
[09/28/2025-09:21:51] [I] Time Deserialize: Disabled
[09/28/2025-09:21:51] [I] Time Refit: Disabled
[09/28/2025-09:21:51] [I] NVTX verbosity: 0
[09/28/2025-09:21:51] [I] Persistent Cache Ratio: 0
[09/28/2025-09:21:51] [I] Optimization Profile Index: 0
[09/28/2025-09:21:51] [I] Weight Streaming Budget: 100.000000%
[09/28/2025-09:21:51] [I] Inputs:
[09/28/2025-09:21:51] [I] Debug Tensor Save Destinations:
[09/28/2025-09:21:51] [I] Dump All Debug Tensor in Formats:
[09/28/2025-09:21:51] [I] === Reporting Options ===
[09/28/2025-09:21:51] [I] Verbose: Disabled
[09/28/2025-09:21:51] [I] Averages: 10 inferences
[09/28/2025-09:21:51] [I] Percentiles: 90,95,99
[09/28/2025-09:21:51] [I] Dump refittable layers:Disabled
[09/28/2025-09:21:51] [I] Dump output: Disabled
[09/28/2025-09:21:51] [I] Profile: Disabled
[09/28/2025-09:21:51] [I] Export timing to JSON file:
[09/28/2025-09:21:51] [I] Export output to JSON file:
[09/28/2025-09:21:51] [I] Export profile to JSON file:
[09/28/2025-09:21:51] [I]
[09/28/2025-09:21:51] [I] === Device Information ===
[09/28/2025-09:21:51] [I] Available Devices:
[09/28/2025-09:21:51] [I] Device 0: "NVIDIA GeForce RTX 3080" UUID: GPU-b1617a48-0d9b-8c1c-a994-d9339659f7f4
[09/28/2025-09:21:51] [I] Selected Device: NVIDIA GeForce RTX 3080
[09/28/2025-09:21:51] [I] Selected Device ID: 0
[09/28/2025-09:21:51] [I] Selected Device UUID: GPU-b1617a48-0d9b-8c1c-a994-d9339659f7f4
[09/28/2025-09:21:51] [I] Compute Capability: 8.6
[09/28/2025-09:21:51] [I] SMs: 70
[09/28/2025-09:21:51] [I] Device Global Memory: 11901 MiB
[09/28/2025-09:21:51] [I] Shared Memory per SM: 100 KiB
[09/28/2025-09:21:51] [I] Memory Bus Width: 384 bits (ECC disabled)
[09/28/2025-09:21:51] [I] Application Compute Clock Rate: 1.785 GHz
[09/28/2025-09:21:51] [I] Application Memory Clock Rate: 9.501 GHz
[09/28/2025-09:21:51] [I]
[09/28/2025-09:21:51] [I] Note: The application clock rates do not reflect the actual clock rates that the GPU is currently running at.
[09/28/2025-09:21:51] [I]
[09/28/2025-09:21:51] [I] TensorRT version: 10.13.3
[09/28/2025-09:21:51] [I] Loading standard plugins
[09/28/2025-09:21:51] [I] [TRT] [MemUsageChange] Init CUDA: CPU +2, GPU +0, now: CPU 25, GPU 887 (MiB)
[09/28/2025-09:21:53] [I] [TRT] [MemUsageChange] Init builder kernel library: CPU +2134, GPU +396, now: CPU 2315, GPU 1283 (MiB)
[09/28/2025-09:21:53] [I] Start parsing network model.
[09/28/2025-09:21:53] [I] [TRT] ----------------------------------------------------------------
[09/28/2025-09:21:53] [I] [TRT] Input filename: ISTD_NS.onnx
[09/28/2025-09:21:53] [I] [TRT] ONNX IR version: 0.0.8
[09/28/2025-09:21:53] [I] [TRT] Opset version: 17
[09/28/2025-09:21:53] [I] [TRT] Producer name: pytorch
[09/28/2025-09:21:53] [I] [TRT] Producer version: 2.8.0
[09/28/2025-09:21:53] [I] [TRT] Domain:
[09/28/2025-09:21:53] [I] [TRT] Model version: 0
[09/28/2025-09:21:53] [I] [TRT] Doc string:
[09/28/2025-09:21:53] [I] [TRT] ----------------------------------------------------------------
[09/28/2025-09:21:53] [E] Error[3]: INetworkDefinition::addPoolingNd: Error Code 3: API Usage Error (Parameter check failed, condition: allDimsGtEq(windowSize, 1) && volume(windowSize) < (int64_t(std::pow(100000.0F, (nbSpatialDims) *0.5F))). )
[09/28/2025-09:21:53] [E] [TRT] ModelImporter.cpp:949: While parsing node number 1176 [AveragePool -> "/up_decoder2/coatt/AveragePool_output_0"]:
[09/28/2025-09:21:53] [E] [TRT] ModelImporter.cpp:950: --- Begin node ---
input: "/Add_1_output_0"
output: "/up_decoder2/coatt/AveragePool_output_0"
name: "/up_decoder2/coatt/AveragePool"
op_type: "AveragePool"
attribute {
name: "ceil_mode"
i: 0
type: INT
}
attribute {
name: "count_include_pad"
i: 1
type: INT
}
attribute {
name: "kernel_shape"
ints: 384
ints: 384
type: INTS
}
attribute {
name: "pads"
ints: 0
ints: 0
ints: 0
ints: 0
type: INTS
}
attribute {
name: "strides"
ints: 384
ints: 384
type: INTS
}
[09/28/2025-09:21:53] [E] [TRT] ModelImporter.cpp:951: --- End node ---
[09/28/2025-09:21:53] [E] [TRT] ModelImporter.cpp:954: ERROR: ModelImporter.cpp:177 In function parseNode:
[1] Exception occurred in - /up_decoder2/coatt/AveragePool
Internal Error!
[09/28/2025-09:21:53] [E] Failed to parse onnx file
[09/28/2025-09:21:53] [I] Finished parsing network model. Parse time: 0.0665538
[09/28/2025-09:21:53] [E] Parsing model failed
[09/28/2025-09:21:53] [E] Failed to create engine from model or file.
[09/28/2025-09:21:53] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v101303] [b9] # ./trtexec --onnx=ISTD_NS.onnx --saveEngine=ISTD_N.trt --iterations=10 --precisionConstraints=obey