fpgasystems / spooNN

FPGA-based neural network inference project with an end-to-end approach (from training to implementation to deployment)
GNU Affero General Public License v3.0
259 stars 74 forks source link

deorefa and evaluate import errors #3

Closed xilinxJAM closed 6 years ago

xilinxJAM commented 6 years ago

Hi I am trying to run your example flow for the Pynq.

When I run the training script I get errors for the dorefa and evaluate imports.

from dorefa import get_dorefa from evaluate import *

DoReFa I found and took from https://github.com/tensorpack/tensorpack/tree/master/examples/DoReFa-Net

Is this correct and where should I get evaluate from?

kaankara commented 6 years ago

Hi, 1) I would not recommend running the training on a PYNQ. This will be extremely slow. I recommend using an x86 machine, ideally with a GPU attached. 2) The modified dorefa script is already included in spooNN (hls-nn-lib/training). Please execute "$ export PYTHONPATH=/path/to/spooNN/hls-nn-lib/training:$PYTHONPATH", as described in the steps for training.

xilinxJAM commented 6 years ago

Thanks, that works, I missed that line in the setup. I am running the training on a x86, with a GPU. I reworded my question as misleading. Thanks again & congratulations on DAC2018

kaankara commented 6 years ago

Awesome, I am glad it works.