While running run-tensorrt-tests, facing below error on run_trtexec test

[PugalendhiMurugan]

[05/14/2022-20:52:43] [E] [TRT] CaffeParser: Could not open file //tensorrt-tests/mnist/mnist.caffemodel
[05/14/2022-20:52:43] [E] [TRT] CaffeParser: Could not parse model file
[05/14/2022-20:52:43] [E] Failed to parse caffe model or prototxt, tensors blob not found
[05/14/2022-20:52:43] [I] Finished parsing network model. Parse time: 0.0034991
[05/14/2022-20:52:43] [E] Parsing model failed
[05/14/2022-20:52:43] [E] Failed to create engine from model or file.
[05/14/2022-20:52:43] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v8602] # trtexec --deploy=//tensorrt-tests/mnist/mnist.prototxt --model=//tenst
=== FAIL: trtexec ===

[ichergui]

HI @PugalendhiMurugan
Could you please share the branch and machine you are using ?
I may need to update the test script for the L4T R36.2.0 DP.

[PugalendhiMurugan]

Hi ichergui,
I'm not sure about the branch and using agx orin machine.

[madisongh]

From the Could not open file //tensorrt-tests/mnist/mnist.caffemodel it looks like the model file didn't get copied over. If you look at that script, you'll see it's pretty simple-minded, and assumes that your current working directory is writable and has sufficient space to store the files needed to run the applications. Make sure you cd somewhere where that can happen before running the script.

[PugalendhiMurugan]

Will give it a try, Thank you

[PugalendhiMurugan]

I think below files are not available on the path-/usr/src/tensorrt/data/mnist

• copy_mnist_data
• rm -rf /test/tensorrt-tests/mnist
• cp -R /usr/src/tensorrt/data/mnist /test/tensorrt-tests/
• echo 'Running trtexec: A Simple mnist model from Caffe'
  Running trtexec: A Simple mnist model from Caffe
• trtexec --deploy=/test/tensorrt-tests/mnist/mnist.prototxt --model=/test/tensorrt-tests/mnist/mnist.caffemodel --out
  &&&& RUNNING TensorRT.trtexec [TensorRT v8602] # trtexec --deploy=/test/tensorrt-tests/mnist/mnist.prototxt --model=/t
  [05/15/2022-02:14:19] [I] === Model Options ===
  [05/15/2022-02:14:19] [I] Format: Caffe
  [05/15/2022-02:14:19] [I] Model: /test/tensorrt-tests/mnist/mnist.caffemodel
  [05/15/2022-02:14:19] [I] Prototxt: /test/tensorrt-tests/mnist/mnist.prototxt

---

[05/15/2022-02:14:25] [E] [TRT] CaffeParser: Could not open file /test/tensorrt-tests/mnist/mnist.caffemodel
[05/15/2022-02:14:25] [E] [TRT] CaffeParser: Could not parse model file
[05/15/2022-02:14:25] [E] Failed to parse caffe model or prototxt, tensors blob not found
[05/15/2022-02:14:25] [I] Finished parsing network model. Parse time: 0.00351136
[05/15/2022-02:14:25] [E] Parsing model failed
[05/15/2022-02:14:25] [E] Failed to create engine from model or file.
[05/15/2022-02:14:25] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v8602] # trtexec --deploy=/test/tensorrt-tests/mnist/mnist.prototxt --model=/tt

• set +x

[madisongh]

Then let's go back to @ichergui 's question... From the RUNNING TensorRT.trtexec [TensorRT v8602] message, it looks like you're using the work-in-progress branch for the JetPack 6 Developer Preview?

It looks like NVIDIA has removed the caffemodel files from the TensorRT samples package with the 8.6.2 version included there. The test script needs an update, and no one has done that yet.

[ichergui]

I have a fix for that. I will push updates to the wip-l4t-r36.2.0-dp
Both mnist.caffemodel and mnist.prototxt are removed

[ichergui]

Hi @PugalendhiMurugan

I just pushed updates. Please try the run-tensorrt-tests and let me know how it goes.