-
Notifications
You must be signed in to change notification settings - Fork 5.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
change --frame argument #1389
change --frame argument #1389
Conversation
Interesting. I'm not sure what run_tdnn_py.sh was modified from; anyway it is not following the correct pattern of these filenames; get Yiming to explain the system to you. |
|
yes it should use the xconfigs, that tdnn sript in tedlium was out of date
…On Tue, Jan 31, 2017 at 10:27 PM, LvHang ***@***.***> wrote:
1. The "run_tdnn_py.sh" was modified from "tedlium/s5_r2/local/nnet3/run_tdnn.sh".
In the 12th stage of "run_tdnn.sh", it call the shell
script--"steps/nnet3/tdnn/train.sh". But in #1370
<#1370> , I think you let me
modify the "train_dnn.py". So I rewrite it. I asked Yiming and move them to
tuning directory.
2. I will rerun the baseline right now.
3. You said the config-generating scripts is out-of-date. However, I
check the "master" and "shortcut" branch in your repository, all of them
use the "steps/nnet3/tdnn/make_configs.py". I guess you mean we need
to write xconfig and use "steps/nnet3/xconfig_to_configs.py", right?
Hang
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1389 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/ADJVu88NYe-us5Vgk1alah2LS_MgB-9aks5rX_uOgaJpZM4Lzb1->
.
|
Hi Dan, |
I think those differences are just the normal differences we see when we rerun. |
Ok. I need to modify it on "master" branch or "shortcut" branch? Thanks |
shortcut
…On Thu, Feb 2, 2017 at 8:11 PM, LvHang ***@***.***> wrote:
Ok. I need to modify it on "master" branch or "shortcut" branch? Thanks
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1389 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/ADJVuwhJp9dKIQzp0c9aezv4i_9SJJfaks5rYn6sgaJpZM4Lzb1->
.
|
Update. |
… training (kaldi-asr#1389) ... makes it vary on each iteration, not in big chunks of time.
@danpovey
Related to #1370 . Attention, this is in "shortcut" branch, not "master" branch.
Hi Dan,
| after-change | ori
decode-dev | 11.7 | 12.6
decode_dev_rescore | 10.9 | 11.5
decode_test | 11.8 | 11.9
decode_test_rescore | 11.2 | 10.9
(The result in "ori" column is extracted from "RESULT" file in tedlium/s5_r2)
Hang