attention_entropies.jsonl

{"layer":0,"head":0,"avg_attn_entropy":6.1600648546}
{"layer":0,"head":1,"avg_attn_entropy":6.1584506788}
{"layer":0,"head":2,"avg_attn_entropy":6.1870868464}
{"layer":0,"head":3,"avg_attn_entropy":6.0735976958}
{"layer":0,"head":4,"avg_attn_entropy":6.1231322632}
{"layer":0,"head":5,"avg_attn_entropy":6.1623747745}
{"layer":0,"head":6,"avg_attn_entropy":6.1639594169}
{"layer":0,"head":7,"avg_attn_entropy":6.146125165}
{"layer":0,"head":8,"avg_attn_entropy":6.1052541461}
{"layer":0,"head":9,"avg_attn_entropy":6.1374521623}
{"layer":0,"head":10,"avg_attn_entropy":6.118777566}
{"layer":0,"head":11,"avg_attn_entropy":6.1617808399}
{"layer":1,"head":0,"avg_attn_entropy":6.1494196639}
{"layer":1,"head":1,"avg_attn_entropy":6.0499497714}
{"layer":1,"head":2,"avg_attn_entropy":6.0912573204}
{"layer":1,"head":3,"avg_attn_entropy":5.9151432519}
{"layer":1,"head":4,"avg_attn_entropy":6.1039670753}
{"layer":1,"head":5,"avg_attn_entropy":6.1303987584}
{"layer":1,"head":6,"avg_attn_entropy":6.1378704433}
{"layer":1,"head":7,"avg_attn_entropy":6.1732876511}
{"layer":1,"head":8,"avg_attn_entropy":6.0844343901}
{"layer":1,"head":9,"avg_attn_entropy":6.0224072151}
{"layer":1,"head":10,"avg_attn_entropy":6.1575298805}
{"layer":1,"head":11,"avg_attn_entropy":6.0931021695}
{"layer":2,"head":0,"avg_attn_entropy":5.8973599391}
{"layer":2,"head":1,"avg_attn_entropy":6.0091449938}
{"layer":2,"head":2,"avg_attn_entropy":6.0747294588}
{"layer":2,"head":3,"avg_attn_entropy":6.1213797913}
{"layer":2,"head":4,"avg_attn_entropy":6.1024243207}
{"layer":2,"head":5,"avg_attn_entropy":6.1493336368}
{"layer":2,"head":6,"avg_attn_entropy":6.1045174026}
{"layer":2,"head":7,"avg_attn_entropy":5.9014663444}
{"layer":2,"head":8,"avg_attn_entropy":6.1138490577}
{"layer":2,"head":9,"avg_attn_entropy":6.063114234}
{"layer":2,"head":10,"avg_attn_entropy":5.9123973575}
{"layer":2,"head":11,"avg_attn_entropy":6.0291404247}
{"layer":3,"head":0,"avg_attn_entropy":6.0902112989}
{"layer":3,"head":1,"avg_attn_entropy":5.957479712}
{"layer":3,"head":2,"avg_attn_entropy":6.0550741978}
{"layer":3,"head":3,"avg_attn_entropy":5.8173761544}
{"layer":3,"head":4,"avg_attn_entropy":5.8348281069}
{"layer":3,"head":5,"avg_attn_entropy":6.1080144997}
{"layer":3,"head":6,"avg_attn_entropy":5.7191191387}
{"layer":3,"head":7,"avg_attn_entropy":6.065710546}
{"layer":3,"head":8,"avg_attn_entropy":5.7916838818}
{"layer":3,"head":9,"avg_attn_entropy":6.0804467278}
{"layer":3,"head":10,"avg_attn_entropy":5.7831644416}
{"layer":3,"head":11,"avg_attn_entropy":5.8379818053}
{"layer":4,"head":0,"avg_attn_entropy":5.8622342019}
{"layer":4,"head":1,"avg_attn_entropy":6.0836263485}
{"layer":4,"head":2,"avg_attn_entropy":6.0946244893}
{"layer":4,"head":3,"avg_attn_entropy":5.9619548087}
{"layer":4,"head":4,"avg_attn_entropy":6.1173812771}
{"layer":4,"head":5,"avg_attn_entropy":6.1273642006}
{"layer":4,"head":6,"avg_attn_entropy":5.6209671931}
{"layer":4,"head":7,"avg_attn_entropy":5.7213881693}
{"layer":4,"head":8,"avg_attn_entropy":5.760812582}
{"layer":4,"head":9,"avg_attn_entropy":6.0715985699}
{"layer":4,"head":10,"avg_attn_entropy":6.0334770617}
{"layer":4,"head":11,"avg_attn_entropy":5.7757602422}
{"layer":5,"head":0,"avg_attn_entropy":6.0019089279}
{"layer":5,"head":1,"avg_attn_entropy":5.8313751068}
{"layer":5,"head":2,"avg_attn_entropy":5.989206809}
{"layer":5,"head":3,"avg_attn_entropy":6.0627046595}
{"layer":5,"head":4,"avg_attn_entropy":5.7283513236}
{"layer":5,"head":5,"avg_attn_entropy":5.9491969538}
{"layer":5,"head":6,"avg_attn_entropy":5.9316726594}
{"layer":5,"head":7,"avg_attn_entropy":5.9880246181}
{"layer":5,"head":8,"avg_attn_entropy":5.9817439265}
{"layer":5,"head":9,"avg_attn_entropy":5.7653453698}
{"layer":5,"head":10,"avg_attn_entropy":5.7180798602}
{"layer":5,"head":11,"avg_attn_entropy":6.0889825735}
{"layer":6,"head":0,"avg_attn_entropy":5.8790088012}
{"layer":6,"head":1,"avg_attn_entropy":5.7733275442}
{"layer":6,"head":2,"avg_attn_entropy":5.8709836962}
{"layer":6,"head":3,"avg_attn_entropy":5.8874621382}
{"layer":6,"head":4,"avg_attn_entropy":5.8479917793}
{"layer":6,"head":5,"avg_attn_entropy":5.9794743543}
{"layer":6,"head":6,"avg_attn_entropy":5.9015631943}
{"layer":6,"head":7,"avg_attn_entropy":5.8486342669}
{"layer":6,"head":8,"avg_attn_entropy":5.7872277117}
{"layer":6,"head":9,"avg_attn_entropy":5.7529757719}
{"layer":6,"head":10,"avg_attn_entropy":5.857016542}
{"layer":6,"head":11,"avg_attn_entropy":5.9148236213}
{"layer":7,"head":0,"avg_attn_entropy":5.7042275209}
{"layer":7,"head":1,"avg_attn_entropy":5.959549922}
{"layer":7,"head":2,"avg_attn_entropy":5.7165108323}
{"layer":7,"head":3,"avg_attn_entropy":5.8884048076}
{"layer":7,"head":4,"avg_attn_entropy":5.7545999279}
{"layer":7,"head":5,"avg_attn_entropy":5.8268059106}
{"layer":7,"head":6,"avg_attn_entropy":5.8753777838}
{"layer":7,"head":7,"avg_attn_entropy":5.7040046787}
{"layer":7,"head":8,"avg_attn_entropy":5.8114910622}
{"layer":7,"head":9,"avg_attn_entropy":5.8918173699}
{"layer":7,"head":10,"avg_attn_entropy":5.6642464123}
{"layer":7,"head":11,"avg_attn_entropy":5.9899081182}
{"layer":8,"head":0,"avg_attn_entropy":5.8113057551}
{"layer":8,"head":1,"avg_attn_entropy":5.9751627235}
{"layer":8,"head":2,"avg_attn_entropy":5.870235693}
{"layer":8,"head":3,"avg_attn_entropy":6.00851261}
{"layer":8,"head":4,"avg_attn_entropy":5.874881382}
{"layer":8,"head":5,"avg_attn_entropy":5.9173713131}
{"layer":8,"head":6,"avg_attn_entropy":5.7487669768}
{"layer":8,"head":7,"avg_attn_entropy":5.9151715322}
{"layer":8,"head":8,"avg_attn_entropy":5.9416119189}
{"layer":8,"head":9,"avg_attn_entropy":6.0304414339}
{"layer":8,"head":10,"avg_attn_entropy":5.9463654428}
{"layer":8,"head":11,"avg_attn_entropy":5.7552012491}
{"layer":9,"head":0,"avg_attn_entropy":5.8410833721}
{"layer":9,"head":1,"avg_attn_entropy":5.8150470786}
{"layer":9,"head":2,"avg_attn_entropy":5.9479796762}
{"layer":9,"head":3,"avg_attn_entropy":5.9216109166}
{"layer":9,"head":4,"avg_attn_entropy":5.8009545419}
{"layer":9,"head":5,"avg_attn_entropy":5.8589636931}
{"layer":9,"head":6,"avg_attn_entropy":5.8163571286}
{"layer":9,"head":7,"avg_attn_entropy":5.8836691439}
{"layer":9,"head":8,"avg_attn_entropy":5.9429277472}
{"layer":9,"head":9,"avg_attn_entropy":5.8739997783}
{"layer":9,"head":10,"avg_attn_entropy":5.8936929183}
{"layer":9,"head":11,"avg_attn_entropy":5.94838977}
{"layer":10,"head":0,"avg_attn_entropy":5.8806723328}
{"layer":10,"head":1,"avg_attn_entropy":5.8968573709}
{"layer":10,"head":2,"avg_attn_entropy":5.6110674973}
{"layer":10,"head":3,"avg_attn_entropy":5.8375773864}
{"layer":10,"head":4,"avg_attn_entropy":5.885481719}
{"layer":10,"head":5,"avg_attn_entropy":5.8896907644}
{"layer":10,"head":6,"avg_attn_entropy":5.9586319361}
{"layer":10,"head":7,"avg_attn_entropy":5.9226288447}
{"layer":10,"head":8,"avg_attn_entropy":5.8263332944}
{"layer":10,"head":9,"avg_attn_entropy":5.766307528}
{"layer":10,"head":10,"avg_attn_entropy":5.8988998795}
{"layer":10,"head":11,"avg_attn_entropy":5.7453649101}
{"layer":11,"head":0,"avg_attn_entropy":5.8651043029}
{"layer":11,"head":1,"avg_attn_entropy":5.9354231782}
{"layer":11,"head":2,"avg_attn_entropy":5.826529108}
{"layer":11,"head":3,"avg_attn_entropy":5.9480430651}
{"layer":11,"head":4,"avg_attn_entropy":5.970781518}
{"layer":11,"head":5,"avg_attn_entropy":5.8623771658}
{"layer":11,"head":6,"avg_attn_entropy":5.8123741412}
{"layer":11,"head":7,"avg_attn_entropy":5.8292103839}
{"layer":11,"head":8,"avg_attn_entropy":5.9031762438}
{"layer":11,"head":9,"avg_attn_entropy":5.8772822061}
{"layer":11,"head":10,"avg_attn_entropy":5.928469512}
{"layer":11,"head":11,"avg_attn_entropy":5.9718926072}