Fasttext: How to to see the parameters of trained models

[seemamishra33]

I am trying to classify (./fasttext supervised -input train.txt -output model)my data sets and I want to see what are the parameters of my trained model I have got. I tried using model.args but it says command not found.

[EdouardGrave]

Hi @seemamishra33,

Have you tried using the dump subcommand?

To get the hyper-parameters of a trained model, you can do:

./fasttext dump MODEL_FILENAME args

Instead of args, you can use dict to get the vocabulary corresponding to the model, input to get the input embeddings or output to get the classifier weights (in case of a supervised model) or the output embeddings (in case of an unsupervised model).

Best

[seemamishra33]

Thanks a lot for reply
Well I just started using fasttext so, I am still learning. But I think
your response really helps.

Thanks

On Fri, Feb 16, 2018 at 9:28 AM, Edouard Grave [e-mail]
wrote:

Hi @seemamishra33 https://github.com/seemamishra33,

Have you tried using the dump subcommand?

To get the hyper-parameters of a trained model, you can do:

./fasttext dump MODEL_FILENAME args

Instead of args, you can use dict to get the vocabulary corresponding to
the model, input to get the input embeddings or output to get the
classifier weights (in case of a supervised model) or the output embeddings
(in case of an unsupervised model).

Best

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/facebookresearch/fastText/issues/418#issuecomment-366302469,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ARX6HOEgf03l0OCuZQQrk-gbY48A7RAoks5tVbrQgaJpZM4Ruvzx
.