Generate pytorch models from scratch and the test data
- Add the python script to generate a python torch model. e.g.
single_conv.py
- Store format: {op}/{single|multi}_{op}.onnx
- Run the script
- Edit the
set
section insetting.ini
, specify- the model
- the location of the files to be store
- Modify
model_test.py
to change the input size if needed- the input data would be randomly generated
np_array = np.random.randn(1, 1, 32, 32).astype(np.float32)
- Run
model_test.py
to generateinput.pb
andoutput.pb
- you will get
{op}_input.pb
and{op}_output.pb
- you will get