Closed RFTFT closed 3 years ago
It seems like complaining your dataset_params
is None
, you can print it out to say what happens to it. That error line should be train.py#L157. maybe you changed something?
sorry, I clone this repo and did not change anything
Can you share more detail on your training script and error logs after clone? It's such a basic functional error that shouldn't happen, and I'm using this project every day...
I've tested on colab, and it should work fine:
Also, I changed centerloss
calculation from reduce_mean
to reduce_sum
long ago, means it's enlarged by multiply with batch_size
. So should use something like "centerloss": 0.01
"centerloss": 0.1
.
This is my code
This is my error log and I print dataset_params
I think it's data.prepare_dataset
returned None
. I just added more log info in this process.
This could happen if your data_path
directory is not exists, or it's empty.
Besides, may also remove the cache file [data_path]_shuffle.npz
to try again.
Thank you for your help I found that my datasets folder name was different