Quantcast
Channel: TAO Toolkit - NVIDIA Developer Forums
Viewing all articles
Browse latest Browse all 497

ONNX to Tensorrt Fails for deployable Retail Object detection model (Fan-DINO)

$
0
0

I am trying to convert retail_object_detection_binary_v2.2.2.3.onnx to a tensorrt model.

following is the command I use inside tritonserver:21.12 docker image.

trtexec --onnx=/etc/models/retail-detect/dino-fan/model.onnx --verbose --maxShapes=input:16x3x540x960 --optShapes=input:16x3x540x960 --minShapes=input:1x3x540x960 --saveEngine=model_repository/retail-detect_dino-fan/1/model.plan --fp16

And I am getting the following error

[01/10/2025-08:06:25] [V] [TRT] Registering layer: /model/backbone/backbone.0/Unsqueeze for ONNX node: /model/backbone/backbone.0/Unsqueeze
[01/10/2025-08:06:25] [V] [TRT] Registering tensor: /model/backbone/backbone.0/Unsqueeze_output_0 for ONNX tensor: /model/backbone/backbone.0/Unsqueeze_output_0
[01/10/2025-08:06:25] [V] [TRT] /model/backbone/backbone.0/Unsqueeze [Unsqueeze] outputs: [/model/backbone/backbone.0/Unsqueeze_output_0 → (1)[INT32]],
[01/10/2025-08:06:25] [V] [TRT] Parsing node: /model/backbone/backbone.0/body/patch_embed/stem/stem.1/LayerNormalization [LayerNormalization]
[01/10/2025-08:06:25] [V] [TRT] Searching for input: /model/backbone/backbone.0/body/patch_embed/stem/stem.1/Transpose_output_0
[01/10/2025-08:06:25] [V] [TRT] Searching for input: model.backbone.0.body.patch_embed.backbone.stem.1.weight
[01/10/2025-08:06:25] [V] [TRT] Searching for input: model.backbone.0.body.patch_embed.backbone.stem.1.bias
[01/10/2025-08:06:25] [V] [TRT] /model/backbone/backbone.0/body/patch_embed/stem/stem.1/LayerNormalization [LayerNormalization] inputs: [/model/backbone/backbone.0/body/patch_embed/stem/stem.1/Transpose_output_0 → (-1, 135, 240, 128)[FLOAT]], [model.backbone.0.body.patch_embed.backbone.stem.1.weight → (128)[FLOAT]], [model.backbone.0.body.patch_embed.backbone.stem.1.bias → (128)[FLOAT]],
[01/10/2025-08:06:25] [I] [TRT] No importer registered for op: LayerNormalization. Attempting to import as plugin.
[01/10/2025-08:06:25] [I] [TRT] Searching for plugin: LayerNormalization, plugin_version: 1, plugin_namespace:
[01/10/2025-08:06:25] [E] [TRT] ModelImporter.cpp:773: While parsing node number 7 [LayerNormalization → “/model/backbone/backbone.0/body/patch_embed/stem/stem.1/LayerNormalization_output_0”]:
[01/10/2025-08:06:25] [E] [TRT] ModelImporter.cpp:774: — Begin node —
[01/10/2025-08:06:25] [E] [TRT] ModelImporter.cpp:775: input: “/model/backbone/backbone.0/body/patch_embed/stem/stem.1/Transpose_output_0”
input: “model.backbone.0.body.patch_embed.backbone.stem.1.weight”
input: “model.backbone.0.body.patch_embed.backbone.stem.1.bias”
output: “/model/backbone/backbone.0/body/patch_embed/stem/stem.1/LayerNormalization_output_0”
name: “/model/backbone/backbone.0/body/patch_embed/stem/stem.1/LayerNormalization”
op_type: “LayerNormalization”
attribute {
name: “axis”
i: -1
type: INT
}
attribute {
name: “epsilon”
f: 1e-06
type: FLOAT
}

[01/10/2025-08:06:25] [E] [TRT] ModelImporter.cpp:776: — End node —
[01/10/2025-08:06:25] [E] [TRT] ModelImporter.cpp:779: ERROR: builtin_op_importers.cpp:4870 In function importFallbackPluginImporter:
[8] Assertion failed: creator && “Plugin not found, are the plugin name, version, and namespace correct?”
[01/10/2025-08:06:26] [E] Failed to parse onnx file
[01/10/2025-08:06:26] [I] Finish parsing network model
[01/10/2025-08:06:26] [E] Parsing model failed
&&&& FAILED TensorRT.trtexec [TensorRT v8201] # trtexec --onnx=/etc/models/retail-detect/dino-fan/model.onnx --verbose --maxShapes=input:16x3x540x960 --optShapes=input:16x3x540x960 --minShapes=input:1x3x540x960 --saveEngine=model_repository/retail-detect_dino-fan/1/model.plan --fp16
[01/10/2025-08:06:26] [E] Failed to create engine from model.
[01/10/2025-08:06:26] [E] Engine set up failed

3 posts - 2 participants

Read full topic


Viewing all articles
Browse latest Browse all 497

Trending Articles