Missing training data

[I-Akrout]

Hello everyone, i'm having a problem with the missing data: when i run the following training command i get an error which says the number of steps is 0: BOARD_SIZE=19 python3 main.py train-dir \ gs://$BUCKET_NAME/data/training_chunks \ gs://$BUCKET_NAME/models/000001-somename \ --model_dir estimator_working_dir

after reading many part of the code i found that with this call

dual_net.train(training_file) #line 92 rl_loop.py

the default value of steps is -1 and in dual_net.py : if steps == -1:

some code here then

   steps = total_examples // effective_batch_size

going further : with gsutil ls gs://$BUCKET_NAME, there is no data folder or file:

gs://minigo-pub/v9-19x19/models/ gs://minigo-pub/v9-19x19/sgf/

does any one know how can i get the data

[brilee]

I've started copying over the ~500GB of training chunk data to gs://minigo-pub/v9-19x19/data/golden_chunks. Should be done in a few minutes.

[I-Akrout]

Thank you @brilee, that was hopeful

[fmscole]

@brilee Where can I get your data in China? Such as BaiduNetDisk. Because GS can not be used in China.Thank you.

[brilee]

Ah.. that's unfortunate. I don't know anything about the Chinese tech ecosystem so I'm afraid I can't help you out there.