"This version of TensorRT does not support dynamic axes."

[squirrel-xs]

Description

Hi Developers,

When I transform image embedding onnx model to tensorrt engine, I meet this error
trtexec --onnx=embedding_onnx/sam_default_embedding.onnx --workspace=4096 --saveEngine=weights/sam_default_embedding.engine
ModelImporter.cpp:777: ERROR: builtin_op_importers.cpp:4493 In function importSlice: [8] Assertion failed: (axes.allValuesKnown()) && "This version of TensorRT does not support dynamic axes."

Environment

TensorRT version: 8.6.1
CUDA: 11.3
cudnn: 8.9.1.23
Operating System: Linux-x86_64
Python Version (if applicable): 3.8
PyTorch Version (if applicable): 1.12.1

The log are show in followers:
`(pytorch) root@crowd-max:/data/zhengwenqing/segment_anything_tensorrt# trtexec --onnx=embedding_onnx/sam_default_embedding.onnx --workspace=4096 --saveEngine=weights/sam_default_embedding.engine
&&&& RUNNING TensorRT.trtexec [TensorRT v8601] # trtexec --onnx=embedding_onnx/sam_default_embedding.onnx --workspace=4096 --saveEngine=weights/sam_default_embedding.engine
[06/19/2023-03:13:15] [W] --workspace flag has been deprecated by --memPoolSize flag.
[06/19/2023-03:13:15] [I] === Model Options ===
[06/19/2023-03:13:15] [I] Format: ONNX
[06/19/2023-03:13:15] [I] Model: embedding_onnx/sam_default_embedding.onnx
[06/19/2023-03:13:15] [I] Output:
[06/19/2023-03:13:15] [I] === Build Options ===
[06/19/2023-03:13:15] [I] Max batch: explicit batch
[06/19/2023-03:13:15] [I] Memory Pools: workspace: 4096 MiB, dlaSRAM: default, dlaLocalDRAM: default, dlaGlobalDRAM: default
[06/19/2023-03:13:15] [I] minTiming: 1
[06/19/2023-03:13:15] [I] avgTiming: 8
[06/19/2023-03:13:15] [I] Precision: FP32
[06/19/2023-03:13:15] [I] LayerPrecisions:
[06/19/2023-03:13:15] [I] Layer Device Types:
[06/19/2023-03:13:15] [I] Calibration:
[06/19/2023-03:13:15] [I] Refit: Disabled
[06/19/2023-03:13:15] [I] Version Compatible: Disabled
[06/19/2023-03:13:15] [I] TensorRT runtime: full
[06/19/2023-03:13:15] [I] Lean DLL Path:
[06/19/2023-03:13:15] [I] Tempfile Controls: { in_memory: allow, temporary: allow }
[06/19/2023-03:13:15] [I] Exclude Lean Runtime: Disabled
[06/19/2023-03:13:15] [I] Sparsity: Disabled
[06/19/2023-03:13:15] [I] Safe mode: Disabled
[06/19/2023-03:13:15] [I] Build DLA standalone loadable: Disabled
[06/19/2023-03:13:15] [I] Allow GPU fallback for DLA: Disabled
[06/19/2023-03:13:15] [I] DirectIO mode: Disabled
[06/19/2023-03:13:15] [I] Restricted mode: Disabled
[06/19/2023-03:13:15] [I] Skip inference: Disabled
[06/19/2023-03:13:15] [I] Save engine: weights/sam_default_embedding.engine
[06/19/2023-03:13:15] [I] Load engine:
[06/19/2023-03:13:15] [I] Profiling verbosity: 0
[06/19/2023-03:13:15] [I] Tactic sources: Using default tactic sources
[06/19/2023-03:13:15] [I] timingCacheMode: local
[06/19/2023-03:13:15] [I] timingCacheFile:
[06/19/2023-03:13:15] [I] Heuristic: Disabled
[06/19/2023-03:13:15] [I] Preview Features: Use default preview flags.
[06/19/2023-03:13:15] [I] MaxAuxStreams: -1
[06/19/2023-03:13:15] [I] BuilderOptimizationLevel: -1
[06/19/2023-03:13:15] [I] Input(s)s format: fp32:CHW
[06/19/2023-03:13:15] [I] Output(s)s format: fp32:CHW
[06/19/2023-03:13:15] [I] Input build shapes: model
[06/19/2023-03:13:15] [I] Input calibration shapes: model
[06/19/2023-03:13:15] [I] === System Options ===
[06/19/2023-03:13:15] [I] Device: 0
[06/19/2023-03:13:15] [I] DLACore:
[06/19/2023-03:13:15] [I] Plugins:
[06/19/2023-03:13:15] [I] setPluginsToSerialize:
[06/19/2023-03:13:15] [I] dynamicPlugins:
[06/19/2023-03:13:15] [I] ignoreParsedPluginLibs: 0
[06/19/2023-03:13:15] [I]
[06/19/2023-03:13:15] [I] === Inference Options ===
[06/19/2023-03:13:15] [I] Batch: Explicit
[06/19/2023-03:13:15] [I] Input inference shapes: model
[06/19/2023-03:13:15] [I] Iterations: 10
[06/19/2023-03:13:15] [I] Duration: 3s (+ 200ms warm up)
[06/19/2023-03:13:15] [I] Sleep time: 0ms
[06/19/2023-03:13:15] [I] Idle time: 0ms
[06/19/2023-03:13:15] [I] Inference Streams: 1
[06/19/2023-03:13:15] [I] ExposeDMA: Disabled
[06/19/2023-03:13:15] [I] Data transfers: Enabled
[06/19/2023-03:13:15] [I] Spin-wait: Disabled
[06/19/2023-03:13:15] [I] Multithreading: Disabled
[06/19/2023-03:13:15] [I] CUDA Graph: Disabled
[06/19/2023-03:13:15] [I] Separate profiling: Disabled
[06/19/2023-03:13:15] [I] Time Deserialize: Disabled
[06/19/2023-03:13:15] [I] Time Refit: Disabled
[06/19/2023-03:13:15] [I] NVTX verbosity: 0
[06/19/2023-03:13:15] [I] Persistent Cache Ratio: 0
[06/19/2023-03:13:15] [I] Inputs:
[06/19/2023-03:13:15] [I] === Reporting Options ===
[06/19/2023-03:13:15] [I] Verbose: Disabled
[06/19/2023-03:13:15] [I] Averages: 10 inferences
[06/19/2023-03:13:15] [I] Percentiles: 90,95,99
[06/19/2023-03:13:15] [I] Dump refittable layers:Disabled
[06/19/2023-03:13:15] [I] Dump output: Disabled
[06/19/2023-03:13:15] [I] Profile: Disabled
[06/19/2023-03:13:15] [I] Export timing to JSON file:
[06/19/2023-03:13:15] [I] Export output to JSON file:
[06/19/2023-03:13:15] [I] Export profile to JSON file:
[06/19/2023-03:13:15] [I]
[06/19/2023-03:13:15] [I] === Device Information ===
[06/19/2023-03:13:15] [I] Selected Device: Tesla T4
[06/19/2023-03:13:15] [I] Compute Capability: 7.5
[06/19/2023-03:13:15] [I] SMs: 40
[06/19/2023-03:13:15] [I] Device Global Memory: 15109 MiB
[06/19/2023-03:13:15] [I] Shared Memory per SM: 64 KiB
[06/19/2023-03:13:15] [I] Memory Bus Width: 256 bits (ECC enabled)
[06/19/2023-03:13:15] [I] Application Compute Clock Rate: 1.59 GHz
[06/19/2023-03:13:15] [I] Application Memory Clock Rate: 5.001 GHz
[06/19/2023-03:13:15] [I]
[06/19/2023-03:13:15] [I] Note: The application clock rates do not reflect the actual clock rates that the GPU is currently running at.
[06/19/2023-03:13:15] [I]
[06/19/2023-03:13:15] [I] TensorRT version: 8.6.1
[06/19/2023-03:13:15] [I] Loading standard plugins
[06/19/2023-03:13:17] [I] [TRT] [MemUsageChange] Init CUDA: CPU +207, GPU +0, now: CPU 211, GPU 2199 (MiB)
[06/19/2023-03:13:29] [I] [TRT] [MemUsageChange] Init builder kernel library: CPU +737, GPU +172, now: CPU 1025, GPU 2371 (MiB)
[06/19/2023-03:13:29] [W] [TRT] CUDA lazy loading is not enabled. Enabling it can significantly reduce device memory usage and speed up TensorRT initialization. See "Lazy Loading" section of CUDA documentation https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#lazy-loading
[06/19/2023-03:13:29] [I] Start parsing network model.
[06/19/2023-03:13:30] [I] [TRT] ----------------------------------------------------------------
[06/19/2023-03:13:30] [I] [TRT] Input filename: embedding_onnx/sam_default_embedding.onnx
[06/19/2023-03:13:30] [I] [TRT] ONNX IR version: 0.0.7
[06/19/2023-03:13:30] [I] [TRT] Opset version: 14
[06/19/2023-03:13:30] [I] [TRT] Producer name: pytorch
[06/19/2023-03:13:30] [I] [TRT] Producer version: 1.12.1
[06/19/2023-03:13:30] [I] [TRT] Domain:
[06/19/2023-03:13:30] [I] [TRT] Model version: 0
[06/19/2023-03:13:30] [I] [TRT] Doc string:
[06/19/2023-03:13:30] [I] [TRT] ----------------------------------------------------------------
[06/19/2023-03:13:30] [W] [TRT] onnx2trt_utils.cpp:374: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[06/19/2023-03:13:45] [W] [TRT] onnx2trt_utils.cpp:400: One or more weights outside the range of INT32 was clamped
[06/19/2023-03:13:45] [E] [TRT] ModelImporter.cpp:771: While parsing node number 1494 [Slice -> "onnx::Slice_1269"]:
[06/19/2023-03:13:45] [E] [TRT] ModelImporter.cpp:772: --- Begin node ---
[06/19/2023-03:13:45] [E] [TRT] ModelImporter.cpp:773: input: "onnx::Slice_1259"
input: "onnx::Slice_13652"
input: "onnx::Slice_1265"
input: "onnx::Slice_13653"
input: "onnx::Slice_1268"
output: "onnx::Slice_1269"
name: "Slice_1494"
op_type: "Slice"

[06/19/2023-03:13:45] [E] [TRT] ModelImporter.cpp:774: --- End node ---
[06/19/2023-03:13:45] [E] [TRT] ModelImporter.cpp:777: ERROR: builtin_op_importers.cpp:4493 In function importSlice:
[8] Assertion failed: (axes.allValuesKnown()) && "This version of TensorRT does not support dynamic axes."
[06/19/2023-03:13:45] [E] Failed to parse onnx file
[06/19/2023-03:13:46] [I] Finished parsing network model. Parse time: 16.2497
[06/19/2023-03:13:46] [E] Parsing model failed
[06/19/2023-03:13:46] [E] Failed to create engine from model or file.
[06/19/2023-03:13:46] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v8601] # trtexec --onnx=embedding_onnx/sam_default_embedding.onnx --workspace=4096 --saveEngine=weights/sam_default_embedding.engine`

[BooHwang]

It seems like you haven't specified an input dimension