Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Model used inemotion downstream#426

bagustris started this conversation inGeneral
Oct 21, 2022· 3 comments· 1 reply
Discussion options

Hi@leo19941227

I would ask/clarify the model used foremotion downstream. In thisissue, you mentioned thater task only used linear layers with average pooling. Checking these lines, we'll see that these configs (linear layer + pooling) are shown in line 37-41.

modelrc:
projector_dim:256
select:UtteranceLevel
UtteranceLevel:
pooling:MeanPooling
DeepModel:
model_type:CNNSelfAttention
hidden_dim:80
kernel_size:5
padding:2
pooling:5
dropout:0.4

My questions are:

  1. Is this true:er used a fully connected network/linear layer with a single layer and 256 units/nodes (two layers if the output layer with 4 nodes is counted)?
  2. What is the role ofDeepModel and its configuration (model_type, etc) in the config above? Is it used (or not used?) when I runer downstream using the default config?
You must be logged in to vote

Replies: 3 comments 1 reply

Comment options

Hi@bagustris ,

  1. Yes your are correct.
  2. The DeepModel was used to boost the performance on IEMOCAP when using fbank feature. It is an available model for emotion downstream however is is not used for the config you paste. You need to change theselect field toDeepModel to use the DeepModel
You must be logged in to vote
0 replies
Comment options

Hello,
Could you give me the directions to correct define theconfig.yaml to use models likeHuBERT?

You must be logged in to vote
0 replies
Comment options

@celsofranssa

The upstream (this is what you meant by models like HuBERT) is not defined inconfig.yaml, but when running the program, i.e.,

python3 run_downstream.py -m train -n iemocap-hubert-large -u hubert_large_ll60k -d emotion

The argument-u define what upstream is used, you can name the experiment with-n with any name you want.

You must be logged in to vote
1 reply
@celsofranssa
Comment options

great, thank you.

Sign up for freeto join this conversation on GitHub. Already have an account?Sign in to comment
Category
General
Labels
None yet
3 participants
@bagustris@celsofranssa@leo19941227
[8]ページ先頭
©2009-2026 Movatter.jp