I am seeing an error when using the SNPE tutorial setup script that converts the the inception_v3 TF model to DLC.
I am running the latest QC and Tensorflow tools (I think): SNPE version 1.61.0, Hexagon versoin 4.5.0.3, Tensorflow 2.6.2, Nvidia Toolbox latest.
Unfortunately, I cannot include the full output because the QC forum thinks I'm posting in an unallowed language if I include the whole thing. Here is the edited output. As you can see, it reports to have completed successfully, but there is a problem on this line:
2022-05-16 19:20:20.474169: I tensorflow/core/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory
Here is the trimmed script output:
$ snpe-tensorflow-to-dlc --input_network /home/jprofit/qc/snpe-1.61.0.3358/models/inception_v3/tensorflow/inception_v3_2016_08_28_frozen_opt.pb --input_dim input 1,299,299,3 --out_node InceptionV3/Predictions/Reshape_1 --output_path /home/jprofit/qc/snpe-1.61.0.3358/models/inception_v3/dlc/inception_v3.dlc2022-05-16 19:20:18.610432: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F FMATo enable them in other operations, rebuild TensorFlow with the appropriate compiler flags....2022-05-16 19:20:19.336455: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1510] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 2246 MB memory: -> device: 0, name: NVIDIA T500, pci bus id: 0000:01:00.0, compute capability: 7.52022-05-16 19:20:20.134022: I tensorflow/stream_executor/cuda/cuda_dnn.cc:369] Loaded cuDNN version 81002022-05-16 19:20:20.474169: I tensorflow/core/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory2022-05-16 19:20:22,339 - 214 - INFO - INFO_ALL_BUILDING_NETWORK:==============================================================Building Network==============================================================2022-05-16 19:20:22,667 - 214 - INFO - Resolving static sub-graphs in network...2022-05-16 19:20:22,761 - 214 - INFO - Resolving static sub-graphs in network, complete.2022-05-16 19:20:22,995 - 214 - INFO - INFO_DLC_SAVE_LOCATION: Saving model at /home/jprofit/qc/snpe-1.61.0.3358/models/inception_v3/dlc/inception_v3.dlc2022-05-16 19:20:23,745 - 214 - INFO - INFO_CONVERSION_SUCCESS: Conversion completed successfully
Hi,
I am also working on that SNPE tutorial setups only, Please try changing the versions.
Ubuntu:18.04
Tensorflow==1.15.0
Python==3.6.13
SNPE==1.54.2.2899
Thanks for the response.
Downgrading to Tensorflow 1.X is not an option for me. Are you saying the SNPE doesn't support Tensorflow 2? The documentation for SNPE suggests that Tensorflow 1.x is deprecated.
Any thoughts on the issue? I am just trying to run the tutorial straight from the instructions without success so I'm confused why I'm the only one having this problem.
Is there a known set of versions for OS, SNPE, Hexagon_SDK and Nvidia (using Tensorflow 2) that is known to work?
Thanks for any pointers.