I tried to reproduce the results mentioned on this model card. Seems like my results do not match the claimed accuracy in the model card. I cannot figure out how to get the correct numbers, can you help me find my mistake?

• Claimed accuracy: 92.22%
• Received accuracy: 0.55%

Here are the details for my validation:

• I instantiate pre-trained model with transformers.pipeline() and use measure accuracy
• Evaluation was performed on CPU.
• Dataset was downloaded from locally from https://huggingface.co/datasets/emotion using load_dataset()