search

pulp-platform / dory

A tool to deploy Deep Neural Networks on PULP-based SoC's
Apache License 2.0
72 stars 22 forks source link

How to get the .json as well as .txt files needed to deploy the model ? #73

Open l1onog opened 4 months ago

l1onog commented 4 months ago

I have successfully quantized the model with the NEMO tool and generated the .onnx file, but when generating the network with the "network_generate.py" file, I get an error: "out_layer0.txt not found." How do I get these .txt files for deploying the model, and I was wondering if the .json file can be generated automatically? Or do I have to fill them in manually? I would be very grateful for your reply, the job is a great one. MV1-128

ccioflan commented 3 months ago

Generating the activations for the quantized layers is considered to be part of the quantization flow and is not covered by dory. You can find an example of how to generate them with Nemo here. Regarding the configuration .json, you can modify the example configuration file considering your requirements. It is worth mentioning that more recent quantization projects such as quantlib generate automatically the quantized model, the configuration file, and the intermediate activations.

l1onog commented 3 months ago

Thank you for your response, it's been incredibly helpful for my work. Additionally, I'm interested in learning whether this model can be deployed on an FPGA(pulp), such as ZCU102. Could you guide me on where to find tutorials or resources for this? Thanks again for your enthusiasm.