Sentiment analysis using RoBERTa model

I checked with two sentences running on demo console first and then running the python code shown in console. When I ran the first example sentence “a very well-made, funny and entertaining picture.”, console said positive and python output was ‘logits’: [5.192181587219238, -3.854187250137329],
‘probs’: [0.9998822212219238, 0.00011780422937590629]. From which I inferred that the first classification probability is for positive and second element is for negative. Then when I ran the second sentence, “I feel like now is a good time to start day drinking on this social distancing Sunday” the console said 65% positive but when I run the python code, output was: ‘logits’: [-1.1452772617340088, 0.4947904050350189],
‘probs’: [0.16245585680007935, 0.8375442028045654]

which shows it is 83% negative. I am new to this, I am getting different results for the same sentence. Please correct me if I am wrong. Thank you.