Checkpoint does not match model architecture
#1
by TomSchimansky - opened
When running the GPU inference example, I get:
Some weights of SERWhisperECAPAForAudioClassification were not initialized from the model checkpoint at MERaLiON/MERaLiON-SER-v1 and are newly initialized: ['dim_layer.1.bias', 'emotion_classification_layer.0.bias']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
Hi Tom,
I just verified that the GPU version is working well without any error. I have attached the recent download screenshot. May be you can check the environment. I tested with transformers version 4.46.3, however, it should work on any new transformer lib package as well.
