So let's say I take 3 different sentences, and tokenize them in GPT Neo :
"My name is Noel"
"My name is Bob"
"My name is Haley"
Results:
{"tokens":[3220,1416,310,642,293]}
{"tokens":[3220,1416,310,8679]}
{"tokens":[3220,1416,310,388,25406]}
Judging by the numbers, it means 3220=My, 1416=name...