Browser test and local test have different result

#4
by ip - opened

Hi,

I'm playing with this model and noticed that in the sequence "US guys what do you think about this piece of shit?" (this is Reddit title) I get different result from browser test and from my local machine. this is the result:

Local machine: [ { "label":"anger", "score":0.30374875664711 }, { "label":"disgust", "score":0.4684724807739258 }, { "label":"fear", "score":0.089825838804245 }, { "label":"joy", "score":0.002520856214687228 }, { "label":"neutral", "score":0.10477810353040695 }, { "label":"sadness", "score":0.008448047563433647 }, { "label":"surprise", "score":0.02220587059855461 } ]

and from browser test:
Screenshot from 2022-10-22 12-02-35.png

I'm curious why there is such a discrepancy?

Thanks for bringing this up. I'll have to look into this. On Google Colab, I got the same results as on the Model Card (see attached).

Screenshot 2022-10-24 115438.png

I am facing the same issue. Also, the pipeline inference is somehow slower than the HF model implementation. Is the model used in hosted inference different from the model card?

Sign up or log in to comment