search

GreenWaves-Technologies / gap_sdk

SDK for Greenwaves Technologies' GAP8 IoT Application Processor
https://greenwaves-technologies.com/en/gap8-the-internet-of-things-iot-application-processor/
Apache License 2.0
137 stars 75 forks source link

Using dump with several inputs - NNTOOL #338

Closed finnfi closed 2 years ago

finnfi commented 2 years ago

Hi! I have a network that has three inputs: 3 pictures of different dimensions. I have a tflite graph, and I have quantized and converted it with the nntool (adjust, aquant, fusions --scale8). Now I want to verify, but it seems like none of the verification methods works with several inputs. When I try using dump (inside the nntool): 

NNT depth_estimation.tflite 0) dump ai_deck_code/input_samples/*/1_*.npy -T -S tensors_fp
dump - input file ['ai_deck_code/input_samples/im1/1_im1.npy', 'ai_deck_code/input_samples/im2/1_im2.npy', 'ai_deck_code/input_samples/state/1_state.npy']
graph_executer - execute uncached: quantization mode none
can only save dumps on one input to tensor store

What am I doing wrong?

Finn

finnfi commented 2 years ago

I kind of fixed it by saving it to a file instead with -p and then inspect with a python script.