How to use angular softmax loss

[rameshkunasi]

Hi,

Can you please give a brief idea about how to configure angular softmax as a loss function in the script?

Training steps I am not able to understand

[rameshkunasi]

I understood the procedure to run. But while running I am getting Could not find enough idle GPUs error. How to resolve this issue

[mycrazycracy]

Hi,

First, make sure your tensorflow can find your gpu (launch python and follow the steps in tf website).

Then, check utils/parallel/limit_num_gpus.sh and you can find which step it fails.

As to the asoftmax, find the corresponding config file in nnet_conf/ to see whether you can understand the configuration.

2020年5月6日 下午6:03，Kunasi Ramesh notifications@github.com 写道：

I understood the procedure to run. But while running I am getting Could not find enough idle GPUs error. How to resolve this issue

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/mycrazycracy/tf-kaldi-speaker/issues/12#issuecomment-624556340, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG6X2XRRLMYURI5QWN24TO3RQEYYNANCNFSM4M2FPC7Q.

[rameshkunasi]

@mycrazycracy ,

I have installed tensorflow-gpu==1.12 and verified whether tensorflow able to detect gpu or not, it detected properly.

I am using single gpu for training. In utils/parallel/limit_num_gpus.sh num_gpus_assigned=0 so it is failing in that step. why the gpu should be idle? i am running another training script in the same gpu, will it be problem?

[rameshkunasi]

I am using python3.6 below error occuring in the log file after resolving GPU issue

File "/root/data_common3tb/Ramesh/tf-kaldi-speaker/dataset/data_loader.py", line 291, in batch_random rd.jumpahead(seed) AttributeError: 'Random' object has no attribute 'jumpahead'

[mycrazycracy]

Yes, in python3, jumpahead is removed. Please refer to https://github.com/mycrazycracy/tf-kaldi-speaker/issues/10 https://github.com/mycrazycracy/tf-kaldi-speaker/issues/10

2020年5月6日 下午6:42，Kunasi Ramesh notifications@github.com 写道：

I am using python3.6 below error occuring in the log file after resolving GPU issue

File "/root/data_common3tb/Ramesh/tf-kaldi-speaker/dataset/data_loader.py", line 291, in batch_random rd.jumpahead(seed) AttributeError: 'Random' object has no attribute 'jumpahead'

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mycrazycracy/tf-kaldi-speaker/issues/12#issuecomment-624573851, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG6X2XUPJ54ZVEJFTVLFNGTRQE5HXANCNFSM4M2FPC7Q.

[mycrazycracy]

If you want to run several programs in a single GPU, change the script for yourself please.

[mycrazycracy]

BTW, I have some copyright problems currently. Would set the repository private temporarily.