I trained in Colab and get models, but how do I test these models ? about light-sernet HOT 4 CLOSED

@AryaAftab
Hi, please help for test models

from light-sernet.

@kananmammad
Hi, kananmammad
The best model in 10 fold-cross validation was saved in TFLite format. You can use the TensorFlow lite tutorial in this site to learn how to deploy your trained TensorFlow Lite model on edge devices such as mobile and Raspberry Pi. If you have more questions feel free and ask.

from light-sernet.

@AryaAftab I don't have Rasperry PI, I use Ubuntu 20.04.3 LTS
There is train model with code in your project, but there isn't test the model with code

python train.py -dn "IEMOCAP" \
                -id 7 \
                -at "impro" \
                -ln "cross_entropy" \
                -v 1 \
                -it "mfcc"

Please, tell me how do I test this model with code ?

from light-sernet.

Inside the train.py file, the desired model is evaluated in each fold after training, and confusion matrices were saved in result folder. For a separate test, you can use the functions prepared in tflite_evaluate.py to evaluate your model:

Example:

from tflite_evaluate import evaluate_model

BuffX = ... # Your mfccs (size : (batch_size, n_frame, n_mfcc, 1))
BuffY = ... # Your Labels (size : (batch_size, ))

evaluate_model(best_modelname_float32, model_type="float32", test_audios=BuffX, test_labels=BuffY)

from light-sernet.