Warnings when running `fpAnalyserCompilerIpevaluator_520.py`
We have a MobileNetV1 models (convert from TensorFlow frozen graph to TFLite then to ONNX). When running fpAnalyserCompilerIpevaluator_520.py
in toolchain, it shows lots of `We find some weight which are not reasonable!!!` in different layers, excerpt as follows:
# python /workspace/scripts/fpAnalyserCompilerIpevaluator_520.py -t 8 input = /workspace/.tmp/updater.json We find some weight which are not reasonable!!! MobilenetV1/Logits/Conv2d_1c_1x1/BiasAdd_bias We find some weight which are not reasonable!!! MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6_weight We find some weight which are not reasonable!!! MobilenetV1/MobilenetV1/Conv2d_13_depthwise/Relu6_weight We find some weight which are not reasonable!!! MobilenetV1/MobilenetV1/Conv2d_12_pointwise/Relu6_bias We find some weight which are not reasonable!!! MobilenetV1/MobilenetV1/Conv2d_12_pointwise/Relu6_weight ... ...
But in the end, it shows success:
Done! [piano][warning][graph_gen.cc:92][GenerateGraph] Model [/data1/fpAnalyser/char_mobilenet_1.0_128_20180515_opt.quan.wqbi.bie] is BIE, skip optimization config
Is there anything I should worry about for the warnings?
The discussion has been closed due to inactivity. To continue with the topic, please feel free to post a new discussion.
Comments
This warning is typically when the weight size does not match the expectation based on the onnx node. For example, the conv weight size is smaller than what the conv layer needs. Could you please attach the tensorflow file and onnx file (dummy weight is okay, just need to check the structure)?
The zip file contains:
output_graph.pb
: The original Tensorflow frozen graphtest_mobilenet_1.0_128.tflite
: The TFLite version convert from frozen graph with command:test_mobilenet_1.0_128.onnx
: ONNX file convert from kneron toolchain with command:python /workspace/libs/ONNX_Convertor/tflite-onnx/onnx_tflite/tflite2onnx.py -tflite path_of_input_tflite_model -save_path path_of_output_onnx_file -release_mode True
Hi Phidias,
Here is what you need to do to resolve the issue.
Please take a try and let us know no matter whether you get good results or need more support.
Thanks for the help!
This time I use command:
to convert the file, and it can compile without any warnings.
However, in hardware verification step it failed, with message:
/data1/c_sim/node_0000_final_output.txt
:/data1/c_sim/node_0000_final_output_float.txt
:/data1/c_sim/node_0000_final_output_matrix.txt
:Hi Phidias,
Could you please find "simulator" and "csim" folder, pack them and attach here? We can take a look.
Thanks,
Bike
This result was generated by using the testing model uploaded earlier, please have a look, thank you in advance!
Hi Phidias,
We found and fixed some simulation bug. It will be released soon in the next tool chain release. For now, you can ignore Csim output and try port to hardware and check performance directly.
Good to hear that, thank you for the support!
Thanks for supporting us and helping us to improve our tool chain.