Upload 12 files

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 checkpoint-8100/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

added_tokens.json

@@ -0,0 +1,837 @@
+{
+  "‼": 248733,
+  "⁉": 248850,
+  "⌛": 248707,
+  "⌨": 248450,
+  "⏏": 248572,
+  "⏫": 248890,
+  "⏭": 249212,
+  "⏮": 248920,
+  "⏯": 248576,
+  "⏲": 248922,
+  "⏸": 249193,
+  "⏹": 249213,
+  "Ⓜ": 248957,
+  "♉": 248638,
+  "♊": 248741,
+  "♌": 249062,
+  "♍": 248609,
+  "♎": 248611,
+  "♏": 249129,
+  "♑": 248994,
+  "♓": 248790,
+  "♾": 248725,
+  "♿": 248454,
+  "⚒": 248845,
+  "⚖": 248534,
+  "⚗": 248943,
+  "⚙": 248676,
+  "⚛": 248761,
+  "⚧": 248971,
+  "⚰": 248930,
+  "⚱": 249010,
+  "⛈": 249232,
+  "⛎": 248516,
+  "⛏": 248490,
+  "⛑": 248575,
+  "⛩": 248953,
+  "⛪": 249120,
+  "⛲": 249141,
+  "⛴": 248779,
+  "⛷": 248649,
+  "⛸": 248693,
+  "⛹": 249260,
+  "⛽": 249248,
+  "✝": 248664,
+  "➗": 248803,
+  "➰": 248665,
+  "➿": 248458,
+  "⬛": 249000,
+  "⬜": 248821,
+  "〽": 248791,
+  "㊗": 248521,
+  "㊙": 249060,
+  "🀄": 248929,
+  "🃏": 248468,
+  "🅿": 248755,
+  "🆎": 248619,
+  "🆒": 248798,
+  "🆕": 248655,
+  "🆖": 248837,
+  "🆗": 248917,
+  "🆙": 248978,
+  "🆚": 248972,
+  "🈁": 248495,
+  "🈂": 248844,
+  "🈚": 249053,
+  "🈯": 248568,
+  "🈲": 248512,
+  "🈳": 248785,
+  "🈴": 249003,
+  "🈵": 248752,
+  "🈶": 248502,
+  "🈷": 249153,
+  "🈸": 249079,
+  "🈹": 248690,
+  "🈺": 248472,
+  "🉐": 249070,
+  "🉑": 249224,
+  "🌁": 248694,
+  "🌂": 248584,
+  "🌃": 248739,
+  "🌇": 248528,
+  "🌉": 248964,
+  "🌌": 248587,
+  "🌑": 249069,
+  "🌒": 249187,
+  "🌓": 248598,
+  "🌔": 248514,
+  "🌕": 248452,
+  "🌖": 249242,
+  "🌗": 248478,
+  "🌘": 248782,
+  "🌚": 249190,
+  "🌛": 248984,
+  "🌜": 248635,
+  "🌠": 248623,
+  "🌡": 248904,
+  "🌤": 248812,
+  "🌥": 248881,
+  "🌦": 249172,
+  "🌧": 248666,
+  "🌨": 248730,
+  "🌩": 248795,
+  "🌪": 249022,
+  "🌫": 248734,
+  "🌬": 248530,
+  "🌭": 248868,
+  "🌮": 248976,
+  "🌯": 248716,
+  "🌰": 248872,
+  "🌵": 249081,
+  "🍈": 249173,
+  "🍐": 248933,
+  "🍖": 248742,
+  "🍗": 248433,
+  "🍘": 248647,
+  "🍙": 249057,
+  "🍛": 248869,
+  "🍜": 248979,
+  "🍝": 249087,
+  "🍞": 248697,
+  "🍟": 248748,
+  "🍠": 248735,
+  "🍡": 248591,
+  "🍢": 248833,
+  "🍣": 248744,
+  "🍤": 248777,
+  "🍥": 248722,
+  "🍧": 248692,
+  "🍩": 248935,
+  "🍭": 248923,
+  "🍮": 249067,
+  "🍱": 248715,
+  "🍵": 248873,
+  "🍶": 248814,
+  "🍸": 249251,
+  "🍼": 248650,
+  "🎃": 248736,
+  "🎆": 248678,
+  "🎇": 248589,
+  "🎌": 248805,
+  "🎍": 248892,
+  "🎎": 249231,
+  "🎏": 248670,
+  "🎐": 249014,
+  "🎑": 248954,
+  "🎒": 249072,
+  "🎚": 249103,
+  "🎛": 248704,
+  "🎟": 248802,
+  "🎠": 249005,
+  "🎡": 248932,
+  "🎣": 249137,
+  "🎦": 248858,
+  "🎪": 249086,
+  "🎫": 249258,
+  "🎰": 248683,
+  "🎱": 248961,
+  "🎲": 248455,
+  "🎳": 249073,
+  "🎴": 248769,
+  "🎷": 248451,
+  "🎹": 248503,
+  "🎺": 248586,
+  "🎻": 249148,
+  "🎽": 248577,
+  "🎾": 249108,
+  "🎿": 248915,
+  "🏁": 249031,
+  "🏂": 248792,
+  "🏅": 249063,
+  "🏇": 248612,
+  "🏈": 248891,
+  "🏉": 248434,
+  "🏊": 249052,
+  "🏍": 248525,
+  "🏎": 248992,
+  "🏏": 248515,
+  "🏐": 248660,
+  "🏑": 249144,
+  "🏒": 248866,
+  "🏓": 248459,
+  "🏔": 248588,
+  "🏕": 248916,
+  "🏗": 249098,
+  "🏘": 249199,
+  "🏙": 248567,
+  "🏚": 248888,
+  "🏛": 248914,
+  "🏜": 248685,
+  "🏞": 249200,
+  "🏟": 248815,
+  "🏣": 248828,
+  "🏤": 248773,
+  "🏥": 248835,
+  "🏦": 248553,
+  "🏧": 248642,
+  "🏨": 248561,
+  "🏩": 249182,
+  "🏪": 248999,
+  "🏬": 248996,
+  "🏭": 248934,
+  "🏮": 248644,
+  "🏯": 249175,
+  "🏰": 248774,
+  "🏴": 248501,
+  "🏷": 248427,
+  "🏸": 248804,
+  "🏹": 249130,
+  "🏺": 249250,
+  "🐀": 249257,
+  "🐁": 249252,
+  "🐂": 248621,
+  "🐃": 248695,
+  "🐄": 249194,
+  "🐅": 249117,
+  "🐆": 249026,
+  "🐇": 249015,
+  "🐈": 248527,
+  "🐉": 248462,
+  "🐊": 248574,
+  "🐋": 249208,
+  "🐌": 248880,
+  "🐎": 248479,
+  "🐏": 248865,
+  "🐐": 249198,
+  "🐑": 249122,
+  "🐓": 248680,
+  "🐔": 249068,
+  "🐕": 248756,
+  "🐖": 248603,
+  "🐗": 248750,
+  "🐙": 248701,
+  "🐚": 248886,
+  "🐛": 248540,
+  "🐜": 249249,
+  "🐟": 248912,
+  "🐠": 249253,
+  "🐡": 248480,
+  "🐢": 249210,
+  "🐧": 249254,
+  "🐨": 249019,
+  "🐩": 248580,
+  "🐪": 248775,
+  "🐫": 248505,
+  "🐬": 249002,
+  "🐭": 248615,
+  "🐮": 249074,
+  "🐵": 248533,
+  "🐹": 248441,
+  "🐺": 249055,
+  "🐽": 248602,
+  "🐿": 248823,
+  "👂": 249140,
+  "👓": 249170,
+  "👘": 248542,
+  "👛": 249217,
+  "👝": 248709,
+  "👞": 248728,
+  "👡": 248549,
+  "👢": 248910,
+  "👲": 249001,
+  "👴": 248820,
+  "👵": 248960,
+  "👷": 249146,
+  "👹": 249099,
+  "👺": 249106,
+  "👽": 248883,
+  "👾": 249059,
+  "💂": 248778,
+  "💈": 248554,
+  "💒": 248594,
+  "💤": 248466,
+  "💨": 248913,
+  "💩": 248508,
+  "💱": 249100,
+  "💴": 248931,
+  "💷": 248711,
+  "💹": 249158,
+  "💺": 248831,
+  "💽": 249021,
+  "💾": 249215,
+  "💿": 248604,
+  "📀": 248518,
+  "📁": 248444,
+  "📂": 248496,
+  "📄": 248801,
+  "📆": 248487,
+  "📇": 249007,
+  "📈": 248438,
+  "📉": 248617,
+  "📊": 248473,
+  "📏": 248556,
+  "📐": 248679,
+  "📑": 249256,
+  "📒": 248710,
+  "📓": 248632,
+  "📔": 248909,
+  "📗": 248807,
+  "📘": 248708,
+  "📙": 249160,
+  "📛": 248484,
+  "📟": 248727,
+  "📠": 249065,
+  "📡": 248532,
+  "📤": 248718,
+  "📨": 249196,
+  "📪": 248565,
+  "📫": 248469,
+  "📭": 248998,
+  "📯": 249220,
+  "📳": 248471,
+  "📴": 249004,
+  "📵": 249013,
+  "📶": 248969,
+  "📻": 249084,
+  "📼": 249230,
+  "📿": 249181,
+  "🔀": 248946,
+  "🔂": 248863,
+  "🔃": 249050,
+  "🔄": 248995,
+  "🔆": 249046,
+  "🔇": 249009,
+  "🔈": 248559,
+  "🔉": 248590,
+  "🔋": 248907,
+  "🔌": 248842,
+  "🔏": 248622,
+  "🔐": 248546,
+  "🔑": 248937,
+  "🔒": 248498,
+  "🔓": 249058,
+  "🔕": 248780,
+  "🔙": 249206,
+  "🔚": 248903,
+  "🔛": 248671,
+  "🔟": 249189,
+  "🔠": 249075,
+  "🔡": 248600,
+  "🔢": 249142,
+  "🔣": 248579,
+  "🔤": 248997,
+  "🔦": 248813,
+  "🔧": 248618,
+  "🔨": 248849,
+  "🔩": 248781,
+  "🔪": 248906,
+  "🔭": 249261,
+  "🔯": 248674,
+  "🔲": 248988,
+  "🔳": 248985,
+  "🔼": 248754,
+  "🔽": 248581,
+  "🕋": 249027,
+  "🕌": 249241,
+  "🕍": 248719,
+  "🕎": 248626,
+  "🕐": 248855,
+  "🕑": 248818,
+  "🕒": 248485,
+  "🕓": 248465,
+  "🕕": 248494,
+  "🕖": 248847,
+  "🕗": 249221,
+  "🕘": 248552,
+  "🕙": 248539,
+  "🕚": 248607,
+  "🕛": 248889,
+  "🕜": 248513,
+  "🕝": 248763,
+  "🕞": 248601,
+  "🕟": 248884,
+  "🕠": 249109,
+  "🕡": 248982,
+  "🕢": 248896,
+  "🕣": 248854,
+  "🕤": 248682,
+  "🕥": 248627,
+  "🕦": 248938,
+  "🕧": 248482,
+  "🕰": 248810,
+  "🕳": 248700,
+  "🕴": 248483,
+  "🕶": 248688,
+  "🕷": 248437,
+  "🕸": 248749,
+  "🕹": 248871,
+  "🖇": 248563,
+  "🖊": 248760,
+  "🖌": 249209,
+  "🖍": 248560,
+  "🖕": 249105,
+  "🖖": 248864,
+  "🖨": 248947,
+  "🖱": 248784,
+  "🖲": 248766,
+  "🖼": 248506,
+  "🗂": 248631,
+  "🗃": 248497,
+  "🗄": 248980,
+  "🗑": 249088,
+  "🗒": 249048,
+  "🗜": 248764,
+  "🗝": 248629,
+  "🗞": 248867,
+  "🗡": 249134,
+  "🗨": 249150,
+  "🗯": 248585,
+  "🗳": 248461,
+  "🗺": 248493,
+  "🗻": 249133,
+  "🗾": 248464,
+  "🗿": 249135,
+  "😧": 249227,
+  "😸": 248908,
+  "😺": 248851,
+  "😼": 248614,
+  "😽": 249121,
+  "😾": 248675,
+  "😿": 248796,
+  "🙉": 248633,
+  "🙍": 249066,
+  "🙎": 248808,
+  "🚁": 248677,
+  "🚂": 248504,
+  "🚃": 249184,
+  "🚄": 248648,
+  "🚅": 249102,
+  "🚆": 248974,
+  "🚈": 248901,
+  "🚉": 248856,
+  "🚊": 249178,
+  "🚋": 248878,
+  "🚍": 249016,
+  "🚎": 249163,
+  "🚏": 249119,
+  "🚐": 248731,
+  "🚑": 249188,
+  "🚒": 249225,
+  "🚓": 248817,
+  "🚔": 249259,
+  "🚕": 248608,
+  "🚖": 248667,
+  "🚜": 249076,
+  "🚝": 248882,
+  "🚞": 248596,
+  "🚟": 249154,
+  "🚠": 248640,
+  "🚡": 249020,
+  "🚢": 248547,
+  "🚣": 249186,
+  "🚤": 249204,
+  "🚥": 249164,
+  "🚦": 248717,
+  "🚧": 249143,
+  "🚪": 248861,
+  "🚬": 248918,
+  "🚭": 248836,
+  "🚮": 248583,
+  "🚯": 248659,
+  "🚰": 249202,
+  "🚱": 248786,
+  "🚳": 248758,
+  "🚷": 249149,
+  "🚸": 248550,
+  "🚹": 249024,
+  "🚺": 249044,
+  "🚻": 248636,
+  "🚼": 249214,
+  "🚽": 248726,
+  "🚾": 249205,
+  "🚿": 248941,
+  "🛀": 249097,
+  "🛁": 248430,
+  "🛂": 248770,
+  "🛃": 248732,
+  "🛄": 248879,
+  "🛅": 249041,
+  "🛋": 249203,
+  "🛎": 249113,
+  "🛏": 249078,
+  "🛐": 248830,
+  "🛕": 248859,
+  "🛖": 248435,
+  "🛗": 248551,
+  "🛜": 248519,
+  "🛝": 248510,
+  "🛞": 248897,
+  "🛟": 248981,
+  "🛠": 248966,
+  "🛡": 248811,
+  "🛢": 249240,
+  "🛣": 249040,
+  "🛤": 248973,
+  "🛥": 248436,
+  "🛩": 249156,
+  "🛬": 248951,
+  "🛰": 248656,
+  "🛳": 249123,
+  "🛴": 249111,
+  "🛵": 248857,
+  "🛶": 248940,
+  "🛷": 248824,
+  "🛸": 248562,
+  "🛹": 248669,
+  "🛺": 248569,
+  "🛻": 249017,
+  "🛼": 248991,
+  "🟠": 249195,
+  "🟡": 248460,
+  "🟢": 248489,
+  "🟣": 248737,
+  "🟤": 248936,
+  "🟥": 249239,
+  "🟦": 248491,
+  "🟧": 249166,
+  "🟨": 248765,
+  "🟩": 248439,
+  "🟪": 248787,
+  "🟫": 249255,
+  "🟰": 248874,
+  "🤌": 248620,
+  "🤍": 248967,
+  "🤎": 249116,
+  "🤏": 248699,
+  "🤐": 248806,
+  "🤒": 248610,
+  "🤕": 248475,
+  "🤚": 249011,
+  "🤛": 249107,
+  "🤜": 249152,
+  "🤠": 248862,
+  "🤢": 249023,
+  "🤥": 248628,
+  "🤧": 249064,
+  "🤫": 248706,
+  "🤬": 249219,
+  "🤮": 248819,
+  "🤯": 249096,
+  "🤰": 249207,
+  "🤱": 248925,
+  "🤳": 248541,
+  "🤴": 248573,
+  "🤶": 249094,
+  "🤸": 248809,
+  "🤹": 248570,
+  "🤺": 248443,
+  "🤼": 249115,
+  "🤽": 248794,
+  "🤾": 248721,
+  "🤿": 249191,
+  "🥁": 249118,
+  "🥃": 248788,
+  "🥄": 248944,
+  "🥅": 249101,
+  "🥈": 248899,
+  "🥉": 248762,
+  "🥊": 248768,
+  "🥋": 248772,
+  "🥌": 249177,
+  "🥍": 249051,
+  "🥎": 248645,
+  "🥏": 248557,
+  "🥐": 249131,
+  "🥑": 248511,
+  "🥒": 249229,
+  "🥓": 248751,
+  "🥔": 248905,
+  "🥕": 248838,
+  "🥖": 248470,
+  "🥗": 248745,
+  "🥘": 248445,
+  "🥙": 248463,
+  "🥚": 248520,
+  "🥛": 248582,
+  "🥜": 249083,
+  "🥝": 248848,
+  "🥞": 248453,
+  "🥟": 249033,
+  "🥠": 248942,
+  "🥡": 248720,
+  "🥢": 249168,
+  "🥤": 248949,
+  "🥥": 248956,
+  "🥦": 248432,
+  "🥧": 248595,
+  "🥨": 248799,
+  "🥩": 248928,
+  "🥪": 248963,
+  "🥫": 248509,
+  "🥬": 249237,
+  "🥭": 248634,
+  "🥮": 248446,
+  "🥯": 248757,
+  "🥱": 249244,
+  "🥲": 248429,
+  "🥳": 249128,
+  "🥴": 248860,
+  "🥵": 249029,
+  "🥶": 249147,
+  "🥷": 248702,
+  "🥸": 248486,
+  "🥹": 248431,
+  "🥻": 249039,
+  "🥼": 248832,
+  "🥽": 248875,
+  "🥾": 249056,
+  "🥿": 249138,
+  "🦀": 249201,
+  "🦂": 249034,
+  "🦃": 248834,
+  "🦆": 249042,
+  "🦇": 248543,
+  "🦈": 249139,
+  "🦉": 249008,
+  "🦌": 248625,
+  "🦍": 248713,
+  "🦎": 248911,
+  "🦏": 248555,
+  "🦐": 248950,
+  "🦑": 248746,
+  "🦒": 248843,
+  "🦓": 248789,
+  "🦔": 248641,
+  "🦕": 248654,
+  "🦖": 248686,
+  "🦗": 249235,
+  "🦘": 248825,
+  "🦙": 248523,
+  "🦚": 248797,
+  "🦛": 249080,
+  "🦜": 248816,
+  "🦝": 248661,
+  "🦞": 249045,
+  "🦟": 248753,
+  "🦠": 249085,
+  "🦡": 249246,
+  "🦢": 248965,
+  "🦣": 248578,
+  "🦤": 248662,
+  "🦥": 248939,
+  "🦦": 249161,
+  "🦧": 249030,
+  "🦨": 248776,
+  "🦩": 248624,
+  "🦪": 248738,
+  "🦫": 248566,
+  "🦬": 249245,
+  "🦭": 249159,
+  "🦮": 248517,
+  "🦯": 249089,
+  "🦰": 249132,
+  "🦱": 248613,
+  "🦲": 249047,
+  "🦳": 248927,
+  "🦴": 248691,
+  "🦵": 248885,
+  "🦶": 248876,
+  "🦷": 248771,
+  "🦸": 248538,
+  "🦹": 248958,
+  "🦺": 248948,
+  "🦻": 248729,
+  "🦼": 248639,
+  "🦽": 248959,
+  "🦾": 248507,
+  "🦿": 249222,
+  "🧀": 248712,
+  "🧁": 248840,
+  "🧂": 249018,
+  "🧃": 249125,
+  "🧄": 248723,
+  "🧅": 248895,
+  "🧆": 248571,
+  "🧇": 248767,
+  "🧈": 249211,
+  "🧉": 248684,
+  "🧊": 249092,
+  "🧋": 248696,
+  "🧌": 249061,
+  "🧍": 249218,
+  "🧎": 248537,
+  "🧏": 248544,
+  "🧐": 249095,
+  "🧑": 248672,
+  "🧒": 248536,
+  "🧓": 249145,
+  "🧔": 249110,
+  "🧕": 249176,
+  "🧖": 249025,
+  "🧗": 248853,
+  "🧘": 248663,
+  "🧙": 248605,
+  "🧛": 248887,
+  "🧝": 248524,
+  "🧞": 248558,
+  "🧟": 249197,
+  "🧠": 248893,
+  "🧢": 249035,
+  "🧣": 249155,
+  "🧤": 248740,
+  "🧥": 248657,
+  "🧦": 248900,
+  "🧧": 248428,
+  "🧨": 248793,
+  "🧩": 249234,
+  "🧪": 249192,
+  "🧫": 248975,
+  "🧬": 248681,
+  "🧭": 249180,
+  "🧮": 249049,
+  "🧯": 249167,
+  "🧰": 248593,
+  "🧱": 249112,
+  "🧲": 249082,
+  "🧳": 249179,
+  "🧴": 249104,
+  "🧵": 249169,
+  "🧶": 248651,
+  "🧷": 248977,
+  "🧸": 248488,
+  "🧹": 249233,
+  "🧺": 248921,
+  "🧻": 249236,
+  "🧼": 248993,
+  "🧽": 248545,
+  "🧾": 248827,
+  "🧿": 249114,
+  "🩰": 248852,
+  "🩱": 248529,
+  "🩲": 249171,
+  "🩳": 248970,
+  "🩴": 248705,
+  "🩵": 248989,
+  "🩶": 249243,
+  "🩷": 248945,
+  "🩸": 248839,
+  "🩹": 248477,
+  "🩺": 248983,
+  "🩻": 248474,
+  "🩼": 249127,
+  "🪀": 249216,
+  "🪁": 249183,
+  "🪂": 248448,
+  "🪃": 248689,
+  "🪄": 248653,
+  "🪅": 248440,
+  "🪆": 248616,
+  "🪇": 249238,
+  "🪈": 249038,
+  "🪐": 249032,
+  "🪑": 249185,
+  "🪒": 248467,
+  "🪓": 249028,
+  "🪔": 248919,
+  "🪕": 248759,
+  "🪖": 248447,
+  "🪗": 249006,
+  "🪘": 248643,
+  "🪙": 248476,
+  "🪚": 249165,
+  "🪛": 249077,
+  "🪜": 248597,
+  "🪝": 248714,
+  "🪞": 248687,
+  "🪟": 249226,
+  "🪠": 248499,
+  "🪡": 248841,
+  "🪢": 249124,
+  "🪣": 248955,
+  "🪤": 248668,
+  "🪥": 248599,
+  "🪦": 248902,
+  "🪧": 248535,
+  "🪨": 248673,
+  "🪩": 248962,
+  "🪪": 249223,
+  "🪫": 248990,
+  "🪬": 248564,
+  "🪭": 248926,
+  "🪮": 248630,
+  "🪯": 248822,
+  "🪰": 248829,
+  "🪱": 249162,
+  "🪲": 248522,
+  "🪳": 249037,
+  "🪴": 248457,
+  "🪵": 248877,
+  "🪶": 249151,
+  "🪷": 249036,
+  "🪸": 248492,
+  "🪹": 248592,
+  "🪺": 248646,
+  "🪻": 248898,
+  "🪼": 248952,
+  "🪽": 249136,
+  "🪿": 249071,
+  "🫀": 248442,
+  "🫁": 248449,
+  "🫂": 248606,
+  "🫃": 248637,
+  "🫄": 248658,
+  "🫅": 248846,
+  "🫎": 248894,
+  "🫏": 248826,
+  "🫐": 248500,
+  "🫑": 248456,
+  "🫒": 248703,
+  "🫓": 248743,
+  "🫔": 249126,
+  "🫕": 249174,
+  "🫖": 248724,
+  "🫗": 248652,
+  "🫘": 249093,
+  "🫙": 249054,
+  "🫚": 248800,
+  "🫛": 248698,
+  "🫠": 249247,
+  "🫡": 248968,
+  "🫢": 248783,
+  "🫣": 248481,
+  "🫤": 248986,
+  "🫥": 249043,
+  "🫦": 248526,
+  "🫧": 248548,
+  "🫨": 248747,
+  "🫰": 249012,
+  "🫱": 249090,
+  "🫲": 248924,
+  "🫳": 249228,
+  "🫴": 248987,
+  "🫵": 248531,
+  "🫶": 249157,
+  "🫷": 249091,
+  "🫸": 248870
+}

config.json

@@ -0,0 +1,40 @@
+{
+  "_name_or_path": "clicknext/phayathaibert",
+  "architectures": [
+    "CamembertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2",
+    "3": "LABEL_3"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2,
+    "LABEL_3": 3
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "camembert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.40.1",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 249262
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aaf08ea07d6399123e24ccea22f9a6192783c65aff174f8e2135aedd03db6c52
+size 1109931736

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46272d55849f78c8946e719105ad77f106ab150686b803da3ac005850cdd8676
+size 2219983098

rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0756100fa218e165a9a20bec4d70d745156140c6e5abfd5647935cea6c5bcb5
+size 14244

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d432300d416e7d574db386599aa107407f0bd61bdd23f62c7d90a6b7c4ff1d1
+size 1064

sentencepiece.bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e295c936bc0d8b6669ae769a2f8a0363e6d3abcfd8d0869134aa1e903a447d26
+size 5261686

special_tokens_map.json

@@ -0,0 +1,56 @@
+{
+  "additional_special_tokens": [
+    "<s>NOTUSED",
+    "</s>NOTUSED",
+    "<_>"
+  ],
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bd9e947fbe9c970a202ea2a4dd511892b6b239078cf5919690e58d35a43e3f2
+size 17349635

trainer_state.json

@@ -0,0 +1,1429 @@
+{
+  "best_metric": 0.762063227953411,
+  "best_model_checkpoint": "finetuned_models/wisesight_sentiment/checkpoint-2400",
+  "epoch": 5.991124260355029,
+  "eval_steps": 100,
+  "global_step": 8100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.07396449704142012,
+      "eval_class_f1": {
+        "neg": 0.018604651162790697,
+        "neu": 0.6997558991049634,
+        "pos": 0.0,
+        "q": 0.0
+      },
+      "eval_loss": 1.0810712575912476,
+      "eval_macro_average_f1": 0.17959013756693853,
+      "eval_micro_average_f1": 0.5391014975041597,
+      "eval_runtime": 6.8397,
+      "eval_samples_per_second": 351.476,
+      "eval_steps_per_second": 22.077,
+      "step": 100
+    },
+    {
+      "epoch": 0.14792899408284024,
+      "eval_class_f1": {
+        "neg": 0.6223055295220243,
+        "neu": 0.7497702909647779,
+        "pos": 0.0,
+        "q": 0.0
+      },
+      "eval_loss": 0.8820463418960571,
+      "eval_macro_average_f1": 0.34301895512170055,
+      "eval_micro_average_f1": 0.6472545757071547,
+      "eval_runtime": 6.8712,
+      "eval_samples_per_second": 349.865,
+      "eval_steps_per_second": 21.976,
+      "step": 200
+    },
+    {
+      "epoch": 0.22189349112426035,
+      "eval_class_f1": {
+        "neg": 0.7588757396449703,
+        "neu": 0.7731384829505916,
+        "pos": 0.27037037037037037,
+        "q": 0.0
+      },
+      "eval_loss": 0.7263810038566589,
+      "eval_macro_average_f1": 0.4505961482414831,
+      "eval_micro_average_f1": 0.7059068219633944,
+      "eval_runtime": 7.0136,
+      "eval_samples_per_second": 342.761,
+      "eval_steps_per_second": 21.529,
+      "step": 300
+    },
+    {
+      "epoch": 0.2958579881656805,
+      "eval_class_f1": {
+        "neg": 0.7806637806637807,
+        "neu": 0.760541586073501,
+        "pos": 0.501891551071879,
+        "q": 0.0909090909090909
+      },
+      "eval_loss": 0.6896220445632935,
+      "eval_macro_average_f1": 0.5335015021795629,
+      "eval_micro_average_f1": 0.7175540765391015,
+      "eval_runtime": 7.0909,
+      "eval_samples_per_second": 339.026,
+      "eval_steps_per_second": 21.295,
+      "step": 400
+    },
+    {
+      "epoch": 0.3698224852071006,
+      "grad_norm": 5.410265922546387,
+      "learning_rate": 1.828817733990148e-05,
+      "loss": 0.8994,
+      "step": 500
+    },
+    {
+      "epoch": 0.3698224852071006,
+      "eval_class_f1": {
+        "neg": 0.7767988252569751,
+        "neu": 0.7759882869692534,
+        "pos": 0.41987179487179493,
+        "q": 0.3777777777777778
+      },
+      "eval_loss": 0.6673027276992798,
+      "eval_macro_average_f1": 0.5876091712189503,
+      "eval_micro_average_f1": 0.7225457570715474,
+      "eval_runtime": 7.071,
+      "eval_samples_per_second": 339.978,
+      "eval_steps_per_second": 21.355,
+      "step": 500
+    },
+    {
+      "epoch": 0.4437869822485207,
+      "eval_class_f1": {
+        "neg": 0.7704042715484364,
+        "neu": 0.7415287628053587,
+        "pos": 0.5426356589147286,
+        "q": 0.14285714285714285
+      },
+      "eval_loss": 0.6752218008041382,
+      "eval_macro_average_f1": 0.5493564590314166,
+      "eval_micro_average_f1": 0.7050748752079867,
+      "eval_runtime": 7.2386,
+      "eval_samples_per_second": 332.107,
+      "eval_steps_per_second": 20.86,
+      "step": 600
+    },
+    {
+      "epoch": 0.5177514792899408,
+      "eval_class_f1": {
+        "neg": 0.7780979827089337,
+        "neu": 0.7535296490520371,
+        "pos": 0.5532435740514076,
+        "q": 0.3870967741935484
+      },
+      "eval_loss": 0.667128324508667,
+      "eval_macro_average_f1": 0.6179919950014817,
+      "eval_micro_average_f1": 0.7171381031613977,
+      "eval_runtime": 7.3636,
+      "eval_samples_per_second": 326.472,
+      "eval_steps_per_second": 20.506,
+      "step": 700
+    },
+    {
+      "epoch": 0.591715976331361,
+      "eval_class_f1": {
+        "neg": 0.7555923777961889,
+        "neu": 0.7529501332318232,
+        "pos": 0.5357575757575757,
+        "q": 0.4161073825503356
+      },
+      "eval_loss": 0.662220299243927,
+      "eval_macro_average_f1": 0.6151018673339809,
+      "eval_micro_average_f1": 0.7059068219633944,
+      "eval_runtime": 7.3848,
+      "eval_samples_per_second": 325.535,
+      "eval_steps_per_second": 20.448,
+      "step": 800
+    },
+    {
+      "epoch": 0.665680473372781,
+      "eval_class_f1": {
+        "neg": 0.7756714060031595,
+        "neu": 0.7722698471859858,
+        "pos": 0.5253807106598984,
+        "q": 0.2535211267605634
+      },
+      "eval_loss": 0.6270455121994019,
+      "eval_macro_average_f1": 0.5817107726524018,
+      "eval_micro_average_f1": 0.7250415973377704,
+      "eval_runtime": 7.3625,
+      "eval_samples_per_second": 326.52,
+      "eval_steps_per_second": 20.509,
+      "step": 900
+    },
+    {
+      "epoch": 0.7396449704142012,
+      "grad_norm": 7.580224514007568,
+      "learning_rate": 2.9252054794520548e-05,
+      "loss": 0.6495,
+      "step": 1000
+    },
+    {
+      "epoch": 0.7396449704142012,
+      "eval_class_f1": {
+        "neg": 0.7862993298585256,
+        "neu": 0.7885968159940763,
+        "pos": 0.5219858156028369,
+        "q": 0.30508474576271183
+      },
+      "eval_loss": 0.6415818929672241,
+      "eval_macro_average_f1": 0.6004916768045376,
+      "eval_micro_average_f1": 0.742928452579035,
+      "eval_runtime": 7.3324,
+      "eval_samples_per_second": 327.861,
+      "eval_steps_per_second": 20.594,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8136094674556213,
+      "eval_class_f1": {
+        "neg": 0.7824267782426777,
+        "neu": 0.7750972762645915,
+        "pos": 0.550531914893617,
+        "q": 0.1923076923076923
+      },
+      "eval_loss": 0.6599770188331604,
+      "eval_macro_average_f1": 0.5750909154271446,
+      "eval_micro_average_f1": 0.7358569051580699,
+      "eval_runtime": 7.4168,
+      "eval_samples_per_second": 324.129,
+      "eval_steps_per_second": 20.359,
+      "step": 1100
+    },
+    {
+      "epoch": 0.8875739644970414,
+      "eval_class_f1": {
+        "neg": 0.7531806615776081,
+        "neu": 0.7700414000752727,
+        "pos": 0.5685164212910533,
+        "q": 0.35955056179775274
+      },
+      "eval_loss": 0.6348879933357239,
+      "eval_macro_average_f1": 0.6128222611854217,
+      "eval_micro_average_f1": 0.721297836938436,
+      "eval_runtime": 7.4408,
+      "eval_samples_per_second": 323.084,
+      "eval_steps_per_second": 20.294,
+      "step": 1200
+    },
+    {
+      "epoch": 0.9615384615384616,
+      "eval_class_f1": {
+        "neg": 0.7920646583394563,
+        "neu": 0.783076923076923,
+        "pos": 0.535014005602241,
+        "q": 0.46616541353383456
+      },
+      "eval_loss": 0.6110679507255554,
+      "eval_macro_average_f1": 0.6440802501381137,
+      "eval_micro_average_f1": 0.7400166389351082,
+      "eval_runtime": 7.4775,
+      "eval_samples_per_second": 321.498,
+      "eval_steps_per_second": 20.194,
+      "step": 1300
+    },
+    {
+      "epoch": 1.0355029585798816,
+      "eval_class_f1": {
+        "neg": 0.8015094339622642,
+        "neu": 0.8065934065934067,
+        "pos": 0.5252225519287834,
+        "q": 0.379746835443038
+      },
+      "eval_loss": 0.6416576504707336,
+      "eval_macro_average_f1": 0.628268056981873,
+      "eval_micro_average_f1": 0.7587354409317804,
+      "eval_runtime": 7.358,
+      "eval_samples_per_second": 326.721,
+      "eval_steps_per_second": 20.522,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1094674556213018,
+      "grad_norm": 3.8226146697998047,
+      "learning_rate": 2.72013698630137e-05,
+      "loss": 0.6084,
+      "step": 1500
+    },
+    {
+      "epoch": 1.1094674556213018,
+      "eval_class_f1": {
+        "neg": 0.7846277021617293,
+        "neu": 0.8024917552216929,
+        "pos": 0.5813953488372093,
+        "q": 0.2857142857142857
+      },
+      "eval_loss": 0.6498740315437317,
+      "eval_macro_average_f1": 0.6135572729837293,
+      "eval_micro_average_f1": 0.7562396006655574,
+      "eval_runtime": 7.2975,
+      "eval_samples_per_second": 329.428,
+      "eval_steps_per_second": 20.692,
+      "step": 1500
+    },
+    {
+      "epoch": 1.183431952662722,
+      "eval_class_f1": {
+        "neg": 0.7917525773195877,
+        "neu": 0.7863247863247863,
+        "pos": 0.5444126074498568,
+        "q": 0.345679012345679
+      },
+      "eval_loss": 0.685055673122406,
+      "eval_macro_average_f1": 0.6170422458599774,
+      "eval_micro_average_f1": 0.7454242928452579,
+      "eval_runtime": 7.1996,
+      "eval_samples_per_second": 333.908,
+      "eval_steps_per_second": 20.973,
+      "step": 1600
+    },
+    {
+      "epoch": 1.2573964497041419,
+      "eval_class_f1": {
+        "neg": 0.7966231772831925,
+        "neu": 0.7783018867924529,
+        "pos": 0.5695216907675196,
+        "q": 0.25806451612903225
+      },
+      "eval_loss": 0.6685267090797424,
+      "eval_macro_average_f1": 0.6006278177430493,
+      "eval_micro_average_f1": 0.7375207986688852,
+      "eval_runtime": 7.2849,
+      "eval_samples_per_second": 329.999,
+      "eval_steps_per_second": 20.728,
+      "step": 1700
+    },
+    {
+      "epoch": 1.331360946745562,
+      "eval_class_f1": {
+        "neg": 0.8018362662586075,
+        "neu": 0.7977570093457944,
+        "pos": 0.5578947368421052,
+        "q": 0.36363636363636365
+      },
+      "eval_loss": 0.6347253918647766,
+      "eval_macro_average_f1": 0.6302810940207177,
+      "eval_micro_average_f1": 0.7549916805324459,
+      "eval_runtime": 7.1263,
+      "eval_samples_per_second": 337.343,
+      "eval_steps_per_second": 21.189,
+      "step": 1800
+    },
+    {
+      "epoch": 1.4053254437869822,
+      "eval_class_f1": {
+        "neg": 0.7917329093799682,
+        "neu": 0.7925512104283055,
+        "pos": 0.5742821473158551,
+        "q": 0.28125
+      },
+      "eval_loss": 0.6284430027008057,
+      "eval_macro_average_f1": 0.6099540667810323,
+      "eval_micro_average_f1": 0.7491680532445923,
+      "eval_runtime": 7.2447,
+      "eval_samples_per_second": 331.831,
+      "eval_steps_per_second": 20.843,
+      "step": 1900
+    },
+    {
+      "epoch": 1.4792899408284024,
+      "grad_norm": 4.046507835388184,
+      "learning_rate": 2.5146575342465757e-05,
+      "loss": 0.5135,
+      "step": 2000
+    },
+    {
+      "epoch": 1.4792899408284024,
+      "eval_class_f1": {
+        "neg": 0.7920792079207921,
+        "neu": 0.7867370007535796,
+        "pos": 0.5517241379310345,
+        "q": 0.27586206896551724
+      },
+      "eval_loss": 0.6431812644004822,
+      "eval_macro_average_f1": 0.601600603892731,
+      "eval_micro_average_f1": 0.7437603993344426,
+      "eval_runtime": 7.2356,
+      "eval_samples_per_second": 332.247,
+      "eval_steps_per_second": 20.869,
+      "step": 2000
+    },
+    {
+      "epoch": 1.5532544378698225,
+      "eval_class_f1": {
+        "neg": 0.7887550200803214,
+        "neu": 0.7950581395348836,
+        "pos": 0.555407209612817,
+        "q": 0.29032258064516125
+      },
+      "eval_loss": 0.6327183842658997,
+      "eval_macro_average_f1": 0.6073857374682958,
+      "eval_micro_average_f1": 0.7495840266222962,
+      "eval_runtime": 7.3333,
+      "eval_samples_per_second": 327.82,
+      "eval_steps_per_second": 20.591,
+      "step": 2100
+    },
+    {
+      "epoch": 1.6272189349112427,
+      "eval_class_f1": {
+        "neg": 0.7658119658119659,
+        "neu": 0.8002847988608045,
+        "pos": 0.56951871657754,
+        "q": 0.4197530864197531
+      },
+      "eval_loss": 0.6534045338630676,
+      "eval_macro_average_f1": 0.6388421419175159,
+      "eval_micro_average_f1": 0.7495840266222962,
+      "eval_runtime": 7.2578,
+      "eval_samples_per_second": 331.231,
+      "eval_steps_per_second": 20.805,
+      "step": 2200
+    },
+    {
+      "epoch": 1.7011834319526629,
+      "eval_class_f1": {
+        "neg": 0.7832369942196531,
+        "neu": 0.7650099403578529,
+        "pos": 0.5829268292682928,
+        "q": 0.35955056179775274
+      },
+      "eval_loss": 0.6581071019172668,
+      "eval_macro_average_f1": 0.6226810814108878,
+      "eval_micro_average_f1": 0.7316971713810316,
+      "eval_runtime": 7.1824,
+      "eval_samples_per_second": 334.708,
+      "eval_steps_per_second": 21.024,
+      "step": 2300
+    },
+    {
+      "epoch": 1.7751479289940828,
+      "eval_class_f1": {
+        "neg": 0.7951807228915662,
+        "neu": 0.8109843081312411,
+        "pos": 0.5441595441595442,
+        "q": 0.3157894736842105
+      },
+      "eval_loss": 0.6206311583518982,
+      "eval_macro_average_f1": 0.6165285122166405,
+      "eval_micro_average_f1": 0.762063227953411,
+      "eval_runtime": 7.2501,
+      "eval_samples_per_second": 331.583,
+      "eval_steps_per_second": 20.827,
+      "step": 2400
+    },
+    {
+      "epoch": 1.849112426035503,
+      "grad_norm": 6.195135593414307,
+      "learning_rate": 2.3091780821917807e-05,
+      "loss": 0.4995,
+      "step": 2500
+    },
+    {
+      "epoch": 1.849112426035503,
+      "eval_class_f1": {
+        "neg": 0.7932148626817447,
+        "neu": 0.8,
+        "pos": 0.5830164765525983,
+        "q": 0.196078431372549
+      },
+      "eval_loss": 0.6029447913169861,
+      "eval_macro_average_f1": 0.5930774426517229,
+      "eval_micro_average_f1": 0.7562396006655574,
+      "eval_runtime": 7.1935,
+      "eval_samples_per_second": 334.192,
+      "eval_steps_per_second": 20.991,
+      "step": 2500
+    },
+    {
+      "epoch": 1.9230769230769231,
+      "eval_class_f1": {
+        "neg": 0.8059236165237724,
+        "neu": 0.7899159663865546,
+        "pos": 0.579415501905972,
+        "q": 0.36666666666666664
+      },
+      "eval_loss": 0.6066814064979553,
+      "eval_macro_average_f1": 0.6354804378707414,
+      "eval_micro_average_f1": 0.7491680532445923,
+      "eval_runtime": 7.2817,
+      "eval_samples_per_second": 330.143,
+      "eval_steps_per_second": 20.737,
+      "step": 2600
+    },
+    {
+      "epoch": 1.997041420118343,
+      "eval_class_f1": {
+        "neg": 0.8003157063930545,
+        "neu": 0.797884397431054,
+        "pos": 0.5773447015834348,
+        "q": 0.3835616438356164
+      },
+      "eval_loss": 0.630171537399292,
+      "eval_macro_average_f1": 0.63977661231079,
+      "eval_micro_average_f1": 0.7545757071547421,
+      "eval_runtime": 7.2376,
+      "eval_samples_per_second": 332.156,
+      "eval_steps_per_second": 20.863,
+      "step": 2700
+    },
+    {
+      "epoch": 2.0710059171597632,
+      "eval_class_f1": {
+        "neg": 0.7848509266720386,
+        "neu": 0.7945103857566765,
+        "pos": 0.5853051058530511,
+        "q": 0.35294117647058826
+      },
+      "eval_loss": 0.7064331769943237,
+      "eval_macro_average_f1": 0.6294018986880886,
+      "eval_micro_average_f1": 0.7508319467554077,
+      "eval_runtime": 7.2934,
+      "eval_samples_per_second": 329.612,
+      "eval_steps_per_second": 20.704,
+      "step": 2800
+    },
+    {
+      "epoch": 2.1449704142011834,
+      "eval_class_f1": {
+        "neg": 0.797752808988764,
+        "neu": 0.8026412325752018,
+        "pos": 0.5824742268041238,
+        "q": 0.26666666666666666
+      },
+      "eval_loss": 0.7201129794120789,
+      "eval_macro_average_f1": 0.612383733758689,
+      "eval_micro_average_f1": 0.7591514143094842,
+      "eval_runtime": 7.1883,
+      "eval_samples_per_second": 334.433,
+      "eval_steps_per_second": 21.006,
+      "step": 2900
+    },
+    {
+      "epoch": 2.2189349112426036,
+      "grad_norm": 6.065237045288086,
+      "learning_rate": 2.1036986301369864e-05,
+      "loss": 0.4003,
+      "step": 3000
+    },
+    {
+      "epoch": 2.2189349112426036,
+      "eval_class_f1": {
+        "neg": 0.7861271676300579,
+        "neu": 0.800578034682081,
+        "pos": 0.5670391061452514,
+        "q": 0.3185840707964602
+      },
+      "eval_loss": 0.7178497910499573,
+      "eval_macro_average_f1": 0.6180820948134627,
+      "eval_micro_average_f1": 0.7508319467554077,
+      "eval_runtime": 7.2106,
+      "eval_samples_per_second": 333.398,
+      "eval_steps_per_second": 20.941,
+      "step": 3000
+    },
+    {
+      "epoch": 2.2928994082840237,
+      "eval_class_f1": {
+        "neg": 0.7955801104972374,
+        "neu": 0.781854043392505,
+        "pos": 0.5852585258525853,
+        "q": 0.28865979381443296
+      },
+      "eval_loss": 0.7727176547050476,
+      "eval_macro_average_f1": 0.6128381183891901,
+      "eval_micro_average_f1": 0.7383527454242929,
+      "eval_runtime": 7.2299,
+      "eval_samples_per_second": 332.51,
+      "eval_steps_per_second": 20.886,
+      "step": 3100
+    },
+    {
+      "epoch": 2.366863905325444,
+      "eval_class_f1": {
+        "neg": 0.7893462469733656,
+        "neu": 0.788983997022702,
+        "pos": 0.5606060606060607,
+        "q": 0.2888888888888889
+      },
+      "eval_loss": 0.7219040393829346,
+      "eval_macro_average_f1": 0.6069562983727543,
+      "eval_micro_average_f1": 0.7420965058236273,
+      "eval_runtime": 7.2669,
+      "eval_samples_per_second": 330.815,
+      "eval_steps_per_second": 20.779,
+      "step": 3200
+    },
+    {
+      "epoch": 2.440828402366864,
+      "eval_class_f1": {
+        "neg": 0.8073115003808072,
+        "neu": 0.7814829344841114,
+        "pos": 0.5855338691159586,
+        "q": 0.26666666666666666
+      },
+      "eval_loss": 0.7229210734367371,
+      "eval_macro_average_f1": 0.610248742661886,
+      "eval_micro_average_f1": 0.7450083194675541,
+      "eval_runtime": 7.1283,
+      "eval_samples_per_second": 337.248,
+      "eval_steps_per_second": 21.183,
+      "step": 3300
+    },
+    {
+      "epoch": 2.5147928994082838,
+      "eval_class_f1": {
+        "neg": 0.7984790874524715,
+        "neu": 0.7856049004594182,
+        "pos": 0.5773447015834348,
+        "q": 0.3
+      },
+      "eval_loss": 0.7037935853004456,
+      "eval_macro_average_f1": 0.615357172373831,
+      "eval_micro_average_f1": 0.747504159733777,
+      "eval_runtime": 7.2219,
+      "eval_samples_per_second": 332.876,
+      "eval_steps_per_second": 20.909,
+      "step": 3400
+    },
+    {
+      "epoch": 2.5887573964497044,
+      "grad_norm": 3.8475677967071533,
+      "learning_rate": 1.8982191780821918e-05,
+      "loss": 0.3579,
+      "step": 3500
+    },
+    {
+      "epoch": 2.5887573964497044,
+      "eval_class_f1": {
+        "neg": 0.7871815940838127,
+        "neu": 0.7871305649083427,
+        "pos": 0.5738916256157636,
+        "q": 0.37735849056603776
+      },
+      "eval_loss": 0.7569752931594849,
+      "eval_macro_average_f1": 0.6313905687934891,
+      "eval_micro_average_f1": 0.7420965058236273,
+      "eval_runtime": 7.3391,
+      "eval_samples_per_second": 327.56,
+      "eval_steps_per_second": 20.575,
+      "step": 3500
+    },
+    {
+      "epoch": 2.662721893491124,
+      "eval_class_f1": {
+        "neg": 0.8064269319051262,
+        "neu": 0.7905718701700155,
+        "pos": 0.5779927448609432,
+        "q": 0.3255813953488372
+      },
+      "eval_loss": 0.7201011180877686,
+      "eval_macro_average_f1": 0.6251432355712305,
+      "eval_micro_average_f1": 0.75,
+      "eval_runtime": 7.2188,
+      "eval_samples_per_second": 333.02,
+      "eval_steps_per_second": 20.918,
+      "step": 3600
+    },
+    {
+      "epoch": 2.7366863905325443,
+      "eval_class_f1": {
+        "neg": 0.7847896440129449,
+        "neu": 0.7701911822083495,
+        "pos": 0.5797413793103448,
+        "q": 0.3703703703703704
+      },
+      "eval_loss": 0.7302864789962769,
+      "eval_macro_average_f1": 0.6262731439755023,
+      "eval_micro_average_f1": 0.7304492512479202,
+      "eval_runtime": 7.2541,
+      "eval_samples_per_second": 331.401,
+      "eval_steps_per_second": 20.816,
+      "step": 3700
+    },
+    {
+      "epoch": 2.8106508875739644,
+      "eval_class_f1": {
+        "neg": 0.7971698113207547,
+        "neu": 0.8014842300556586,
+        "pos": 0.5839793281653747,
+        "q": 0.3283582089552239
+      },
+      "eval_loss": 0.7112248539924622,
+      "eval_macro_average_f1": 0.627747894624253,
+      "eval_micro_average_f1": 0.7587354409317804,
+      "eval_runtime": 7.2905,
+      "eval_samples_per_second": 329.743,
+      "eval_steps_per_second": 20.712,
+      "step": 3800
+    },
+    {
+      "epoch": 2.8846153846153846,
+      "eval_class_f1": {
+        "neg": 0.7999999999999999,
+        "neu": 0.7944066515495087,
+        "pos": 0.5961995249406176,
+        "q": 0.2545454545454545
+      },
+      "eval_loss": 0.7105884552001953,
+      "eval_macro_average_f1": 0.6112879077588952,
+      "eval_micro_average_f1": 0.7549916805324459,
+      "eval_runtime": 7.4167,
+      "eval_samples_per_second": 324.132,
+      "eval_steps_per_second": 20.359,
+      "step": 3900
+    },
+    {
+      "epoch": 2.9585798816568047,
+      "grad_norm": 8.97050666809082,
+      "learning_rate": 1.6927397260273975e-05,
+      "loss": 0.3409,
+      "step": 4000
+    },
+    {
+      "epoch": 2.9585798816568047,
+      "eval_class_f1": {
+        "neg": 0.803088803088803,
+        "neu": 0.7901328273244782,
+        "pos": 0.5671641791044775,
+        "q": 0.3513513513513513
+      },
+      "eval_loss": 0.7364293932914734,
+      "eval_macro_average_f1": 0.6279342902172774,
+      "eval_micro_average_f1": 0.7495840266222962,
+      "eval_runtime": 7.2488,
+      "eval_samples_per_second": 331.641,
+      "eval_steps_per_second": 20.831,
+      "step": 4000
+    },
+    {
+      "epoch": 3.032544378698225,
+      "eval_class_f1": {
+        "neg": 0.7924836601307189,
+        "neu": 0.7892777364110202,
+        "pos": 0.5696969696969698,
+        "q": 0.3287671232876712
+      },
+      "eval_loss": 0.8425710201263428,
+      "eval_macro_average_f1": 0.6200563723815951,
+      "eval_micro_average_f1": 0.7454242928452579,
+      "eval_runtime": 7.1671,
+      "eval_samples_per_second": 335.422,
+      "eval_steps_per_second": 21.068,
+      "step": 4100
+    },
+    {
+      "epoch": 3.106508875739645,
+      "eval_class_f1": {
+        "neg": 0.7883817427385893,
+        "neu": 0.7684537684537684,
+        "pos": 0.5720338983050848,
+        "q": 0.35294117647058826
+      },
+      "eval_loss": 0.9264113306999207,
+      "eval_macro_average_f1": 0.6204526464920077,
+      "eval_micro_average_f1": 0.7275374376039934,
+      "eval_runtime": 7.2924,
+      "eval_samples_per_second": 329.656,
+      "eval_steps_per_second": 20.706,
+      "step": 4200
+    },
+    {
+      "epoch": 3.1804733727810652,
+      "eval_class_f1": {
+        "neg": 0.8064269319051262,
+        "neu": 0.7787333854573885,
+        "pos": 0.5774647887323944,
+        "q": 0.32967032967032966
+      },
+      "eval_loss": 0.9222328662872314,
+      "eval_macro_average_f1": 0.6230738589413097,
+      "eval_micro_average_f1": 0.7420965058236273,
+      "eval_runtime": 7.2012,
+      "eval_samples_per_second": 333.833,
+      "eval_steps_per_second": 20.969,
+      "step": 4300
+    },
+    {
+      "epoch": 3.2544378698224854,
+      "eval_class_f1": {
+        "neg": 0.7999999999999999,
+        "neu": 0.7803557617942769,
+        "pos": 0.5765124555160142,
+        "q": 0.35955056179775274
+      },
+      "eval_loss": 0.9496058821678162,
+      "eval_macro_average_f1": 0.6291046947770109,
+      "eval_micro_average_f1": 0.7420965058236273,
+      "eval_runtime": 7.3334,
+      "eval_samples_per_second": 327.814,
+      "eval_steps_per_second": 20.591,
+      "step": 4400
+    },
+    {
+      "epoch": 3.328402366863905,
+      "grad_norm": 12.435276985168457,
+      "learning_rate": 1.4872602739726027e-05,
+      "loss": 0.2249,
+      "step": 4500
+    },
+    {
+      "epoch": 3.328402366863905,
+      "eval_class_f1": {
+        "neg": 0.8012718600953895,
+        "neu": 0.784238714613619,
+        "pos": 0.5663082437275986,
+        "q": 0.32323232323232326
+      },
+      "eval_loss": 0.9026820063591003,
+      "eval_macro_average_f1": 0.6187627854172325,
+      "eval_micro_average_f1": 0.7412645590682196,
+      "eval_runtime": 7.1404,
+      "eval_samples_per_second": 336.677,
+      "eval_steps_per_second": 21.147,
+      "step": 4500
+    },
+    {
+      "epoch": 3.4023668639053253,
+      "eval_class_f1": {
+        "neg": 0.8043647700701482,
+        "neu": 0.7884322678843227,
+        "pos": 0.5676328502415459,
+        "q": 0.2898550724637681
+      },
+      "eval_loss": 0.943065345287323,
+      "eval_macro_average_f1": 0.6125712401649462,
+      "eval_micro_average_f1": 0.747504159733777,
+      "eval_runtime": 7.2681,
+      "eval_samples_per_second": 330.759,
+      "eval_steps_per_second": 20.776,
+      "step": 4600
+    },
+    {
+      "epoch": 3.4763313609467454,
+      "eval_class_f1": {
+        "neg": 0.8018504240555128,
+        "neu": 0.7930382141505864,
+        "pos": 0.5692503176620076,
+        "q": 0.345679012345679
+      },
+      "eval_loss": 0.9825762510299683,
+      "eval_macro_average_f1": 0.6274544920534464,
+      "eval_micro_average_f1": 0.7512479201331115,
+      "eval_runtime": 7.2921,
+      "eval_samples_per_second": 329.672,
+      "eval_steps_per_second": 20.707,
+      "step": 4700
+    },
+    {
+      "epoch": 3.5502958579881656,
+      "eval_class_f1": {
+        "neg": 0.7946498819826908,
+        "neu": 0.7813455657492355,
+        "pos": 0.5795053003533569,
+        "q": 0.3055555555555555
+      },
+      "eval_loss": 0.9374552965164185,
+      "eval_macro_average_f1": 0.6152640759102097,
+      "eval_micro_average_f1": 0.7420965058236273,
+      "eval_runtime": 7.387,
+      "eval_samples_per_second": 325.436,
+      "eval_steps_per_second": 20.441,
+      "step": 4800
+    },
+    {
+      "epoch": 3.6242603550295858,
+      "eval_class_f1": {
+        "neg": 0.7984790874524715,
+        "neu": 0.7785547785547785,
+        "pos": 0.5748218527315915,
+        "q": 0.3116883116883117
+      },
+      "eval_loss": 0.9656402468681335,
+      "eval_macro_average_f1": 0.6158860076067884,
+      "eval_micro_average_f1": 0.7408485856905158,
+      "eval_runtime": 7.4821,
+      "eval_samples_per_second": 321.299,
+      "eval_steps_per_second": 20.181,
+      "step": 4900
+    },
+    {
+      "epoch": 3.698224852071006,
+      "grad_norm": 0.6623280644416809,
+      "learning_rate": 1.2817808219178083e-05,
+      "loss": 0.2207,
+      "step": 5000
+    },
+    {
+      "epoch": 3.698224852071006,
+      "eval_class_f1": {
+        "neg": 0.7987616099071206,
+        "neu": 0.7862857142857143,
+        "pos": 0.5647348951911221,
+        "q": 0.3
+      },
+      "eval_loss": 0.9422620534896851,
+      "eval_macro_average_f1": 0.6124455548459892,
+      "eval_micro_average_f1": 0.7441763727121464,
+      "eval_runtime": 7.2765,
+      "eval_samples_per_second": 330.379,
+      "eval_steps_per_second": 20.752,
+      "step": 5000
+    },
+    {
+      "epoch": 3.772189349112426,
+      "eval_class_f1": {
+        "neg": 0.8,
+        "neu": 0.7671342685370742,
+        "pos": 0.5714285714285714,
+        "q": 0.3414634146341463
+      },
+      "eval_loss": 0.9625053405761719,
+      "eval_macro_average_f1": 0.620006563649948,
+      "eval_micro_average_f1": 0.7296173044925125,
+      "eval_runtime": 7.3445,
+      "eval_samples_per_second": 327.318,
+      "eval_steps_per_second": 20.559,
+      "step": 5100
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "eval_class_f1": {
+        "neg": 0.8018942383583267,
+        "neu": 0.7968691762951919,
+        "pos": 0.5706874189364461,
+        "q": 0.25287356321839083
+      },
+      "eval_loss": 0.9822611212730408,
+      "eval_macro_average_f1": 0.6055810992020889,
+      "eval_micro_average_f1": 0.7520798668885191,
+      "eval_runtime": 7.4499,
+      "eval_samples_per_second": 322.688,
+      "eval_steps_per_second": 20.269,
+      "step": 5200
+    },
+    {
+      "epoch": 3.9201183431952664,
+      "eval_class_f1": {
+        "neg": 0.7891268533772653,
+        "neu": 0.7875375375375375,
+        "pos": 0.5810968494749125,
+        "q": 0.273972602739726
+      },
+      "eval_loss": 0.9442653656005859,
+      "eval_macro_average_f1": 0.6079334607823603,
+      "eval_micro_average_f1": 0.7433444259567388,
+      "eval_runtime": 7.3731,
+      "eval_samples_per_second": 326.052,
+      "eval_steps_per_second": 20.48,
+      "step": 5300
+    },
+    {
+      "epoch": 3.994082840236686,
+      "eval_class_f1": {
+        "neg": 0.8046511627906977,
+        "neu": 0.7945516458569808,
+        "pos": 0.5775,
+        "q": 0.32
+      },
+      "eval_loss": 0.9429491758346558,
+      "eval_macro_average_f1": 0.6241757021619195,
+      "eval_micro_average_f1": 0.7537437603993344,
+      "eval_runtime": 7.3966,
+      "eval_samples_per_second": 325.013,
+      "eval_steps_per_second": 20.415,
+      "step": 5400
+    },
+    {
+      "epoch": 4.068047337278107,
+      "grad_norm": 2.4124114513397217,
+      "learning_rate": 1.0763013698630138e-05,
+      "loss": 0.2077,
+      "step": 5500
+    },
+    {
+      "epoch": 4.068047337278107,
+      "eval_class_f1": {
+        "neg": 0.8063781321184511,
+        "neu": 0.7866927592954991,
+        "pos": 0.5862884160756501,
+        "q": 0.3333333333333333
+      },
+      "eval_loss": 1.1077452898025513,
+      "eval_macro_average_f1": 0.6281731602057334,
+      "eval_micro_average_f1": 0.7483361064891847,
+      "eval_runtime": 7.369,
+      "eval_samples_per_second": 326.23,
+      "eval_steps_per_second": 20.491,
+      "step": 5500
+    },
+    {
+      "epoch": 4.1420118343195265,
+      "eval_class_f1": {
+        "neg": 0.7993920972644377,
+        "neu": 0.7660256410256411,
+        "pos": 0.5726775956284154,
+        "q": 0.29629629629629634
+      },
+      "eval_loss": 1.1472598314285278,
+      "eval_macro_average_f1": 0.6085979075536977,
+      "eval_micro_average_f1": 0.7304492512479202,
+      "eval_runtime": 7.5033,
+      "eval_samples_per_second": 320.394,
+      "eval_steps_per_second": 20.125,
+      "step": 5600
+    },
+    {
+      "epoch": 4.215976331360947,
+      "eval_class_f1": {
+        "neg": 0.8024502297090352,
+        "neu": 0.7868978805394989,
+        "pos": 0.5731559854897219,
+        "q": 0.3
+      },
+      "eval_loss": 1.169406533241272,
+      "eval_macro_average_f1": 0.615626023934564,
+      "eval_micro_average_f1": 0.7462562396006656,
+      "eval_runtime": 7.4833,
+      "eval_samples_per_second": 321.249,
+      "eval_steps_per_second": 20.178,
+      "step": 5700
+    },
+    {
+      "epoch": 4.289940828402367,
+      "eval_class_f1": {
+        "neg": 0.803951367781155,
+        "neu": 0.7902550437761706,
+        "pos": 0.5685019206145967,
+        "q": 0.30952380952380953
+      },
+      "eval_loss": 1.1968339681625366,
+      "eval_macro_average_f1": 0.6180580354239329,
+      "eval_micro_average_f1": 0.7495840266222962,
+      "eval_runtime": 7.4382,
+      "eval_samples_per_second": 323.197,
+      "eval_steps_per_second": 20.301,
+      "step": 5800
+    },
+    {
+      "epoch": 4.363905325443787,
+      "eval_class_f1": {
+        "neg": 0.8024786986831913,
+        "neu": 0.7916030534351146,
+        "pos": 0.5773955773955775,
+        "q": 0.3373493975903615
+      },
+      "eval_loss": 1.1896393299102783,
+      "eval_macro_average_f1": 0.6272066817760612,
+      "eval_micro_average_f1": 0.7504159733777038,
+      "eval_runtime": 7.423,
+      "eval_samples_per_second": 323.857,
+      "eval_steps_per_second": 20.342,
+      "step": 5900
+    },
+    {
+      "epoch": 4.437869822485207,
+      "grad_norm": 0.9506312608718872,
+      "learning_rate": 8.708219178082192e-06,
+      "loss": 0.1324,
+      "step": 6000
+    },
+    {
+      "epoch": 4.437869822485207,
+      "eval_class_f1": {
+        "neg": 0.8024316109422492,
+        "neu": 0.7925840092699884,
+        "pos": 0.5878048780487805,
+        "q": 0.28915662650602414
+      },
+      "eval_loss": 1.2535008192062378,
+      "eval_macro_average_f1": 0.6179942811917606,
+      "eval_micro_average_f1": 0.7516638935108153,
+      "eval_runtime": 7.3808,
+      "eval_samples_per_second": 325.711,
+      "eval_steps_per_second": 20.459,
+      "step": 6000
+    },
+    {
+      "epoch": 4.511834319526627,
+      "eval_class_f1": {
+        "neg": 0.7901821060965954,
+        "neu": 0.7819374758780393,
+        "pos": 0.5821064552661382,
+        "q": 0.2535211267605634
+      },
+      "eval_loss": 1.2182434797286987,
+      "eval_macro_average_f1": 0.601936791000334,
+      "eval_micro_average_f1": 0.7396006655574043,
+      "eval_runtime": 7.3498,
+      "eval_samples_per_second": 327.085,
+      "eval_steps_per_second": 20.545,
+      "step": 6100
+    },
+    {
+      "epoch": 4.585798816568047,
+      "eval_class_f1": {
+        "neg": 0.7945425361155697,
+        "neu": 0.7956989247311828,
+        "pos": 0.5761006289308177,
+        "q": 0.22857142857142854
+      },
+      "eval_loss": 1.2836171388626099,
+      "eval_macro_average_f1": 0.5987283795872497,
+      "eval_micro_average_f1": 0.7508319467554077,
+      "eval_runtime": 7.4332,
+      "eval_samples_per_second": 323.412,
+      "eval_steps_per_second": 20.314,
+      "step": 6200
+    },
+    {
+      "epoch": 4.659763313609467,
+      "eval_class_f1": {
+        "neg": 0.8003025718608169,
+        "neu": 0.7733970529669454,
+        "pos": 0.5691609977324262,
+        "q": 0.3225806451612903
+      },
+      "eval_loss": 1.2842472791671753,
+      "eval_macro_average_f1": 0.6163603169303697,
+      "eval_micro_average_f1": 0.7346089850249584,
+      "eval_runtime": 7.4941,
+      "eval_samples_per_second": 320.786,
+      "eval_steps_per_second": 20.149,
+      "step": 6300
+    },
+    {
+      "epoch": 4.733727810650888,
+      "eval_class_f1": {
+        "neg": 0.799375487900078,
+        "neu": 0.7936865839909809,
+        "pos": 0.5614489003880984,
+        "q": 0.3225806451612903
+      },
+      "eval_loss": 1.3067219257354736,
+      "eval_macro_average_f1": 0.619272904360112,
+      "eval_micro_average_f1": 0.7487520798668885,
+      "eval_runtime": 7.3805,
+      "eval_samples_per_second": 325.723,
+      "eval_steps_per_second": 20.459,
+      "step": 6400
+    },
+    {
+      "epoch": 4.8076923076923075,
+      "grad_norm": 0.2732117772102356,
+      "learning_rate": 6.653424657534246e-06,
+      "loss": 0.1441,
+      "step": 6500
+    },
+    {
+      "epoch": 4.8076923076923075,
+      "eval_class_f1": {
+        "neg": 0.8027628549501151,
+        "neu": 0.7751572327044026,
+        "pos": 0.5694760820045559,
+        "q": 0.3132530120481927
+      },
+      "eval_loss": 1.2718240022659302,
+      "eval_macro_average_f1": 0.6151622954268166,
+      "eval_micro_average_f1": 0.7371048252911814,
+      "eval_runtime": 7.3822,
+      "eval_samples_per_second": 325.649,
+      "eval_steps_per_second": 20.455,
+      "step": 6500
+    },
+    {
+      "epoch": 4.881656804733728,
+      "eval_class_f1": {
+        "neg": 0.796875,
+        "neu": 0.7868601986249045,
+        "pos": 0.5735115431348725,
+        "q": 0.29885057471264365
+      },
+      "eval_loss": 1.261472225189209,
+      "eval_macro_average_f1": 0.6140243291181051,
+      "eval_micro_average_f1": 0.7441763727121464,
+      "eval_runtime": 7.5114,
+      "eval_samples_per_second": 320.048,
+      "eval_steps_per_second": 20.103,
+      "step": 6600
+    },
+    {
+      "epoch": 4.955621301775148,
+      "eval_class_f1": {
+        "neg": 0.7930763178599529,
+        "neu": 0.7766536964980545,
+        "pos": 0.5714285714285715,
+        "q": 0.35294117647058826
+      },
+      "eval_loss": 1.2753080129623413,
+      "eval_macro_average_f1": 0.6235249405642919,
+      "eval_micro_average_f1": 0.7358569051580699,
+      "eval_runtime": 7.3656,
+      "eval_samples_per_second": 326.381,
+      "eval_steps_per_second": 20.501,
+      "step": 6700
+    },
+    {
+      "epoch": 5.029585798816568,
+      "eval_class_f1": {
+        "neg": 0.7962962962962963,
+        "neu": 0.7754943776657619,
+        "pos": 0.5657276995305165,
+        "q": 0.345679012345679
+      },
+      "eval_loss": 1.3079357147216797,
+      "eval_macro_average_f1": 0.6207993464595634,
+      "eval_micro_average_f1": 0.7366888519134775,
+      "eval_runtime": 7.4026,
+      "eval_samples_per_second": 324.753,
+      "eval_steps_per_second": 20.398,
+      "step": 6800
+    },
+    {
+      "epoch": 5.103550295857988,
+      "eval_class_f1": {
+        "neg": 0.7972136222910216,
+        "neu": 0.7786790266512167,
+        "pos": 0.5721040189125295,
+        "q": 0.3703703703703704
+      },
+      "eval_loss": 1.3499900102615356,
+      "eval_macro_average_f1": 0.6295917595562845,
+      "eval_micro_average_f1": 0.740432612312812,
+      "eval_runtime": 7.3777,
+      "eval_samples_per_second": 325.846,
+      "eval_steps_per_second": 20.467,
+      "step": 6900
+    },
+    {
+      "epoch": 5.177514792899408,
+      "grad_norm": 11.024497985839844,
+      "learning_rate": 4.598630136986302e-06,
+      "loss": 0.1111,
+      "step": 7000
+    },
+    {
+      "epoch": 5.177514792899408,
+      "eval_class_f1": {
+        "neg": 0.7956147220046985,
+        "neu": 0.7807853602744949,
+        "pos": 0.5693606755126658,
+        "q": 0.3544303797468354
+      },
+      "eval_loss": 1.4051584005355835,
+      "eval_macro_average_f1": 0.6250477843846737,
+      "eval_micro_average_f1": 0.7412645590682196,
+      "eval_runtime": 7.2531,
+      "eval_samples_per_second": 331.444,
+      "eval_steps_per_second": 20.819,
+      "step": 7000
+    },
+    {
+      "epoch": 5.2514792899408285,
+      "eval_class_f1": {
+        "neg": 0.7925407925407926,
+        "neu": 0.7769230769230769,
+        "pos": 0.5737898465171192,
+        "q": 0.27027027027027023
+      },
+      "eval_loss": 1.4020917415618896,
+      "eval_macro_average_f1": 0.6033809965628147,
+      "eval_micro_average_f1": 0.7375207986688852,
+      "eval_runtime": 7.404,
+      "eval_samples_per_second": 324.69,
+      "eval_steps_per_second": 20.394,
+      "step": 7100
+    },
+    {
+      "epoch": 5.325443786982248,
+      "eval_class_f1": {
+        "neg": 0.7949326999208235,
+        "neu": 0.7753846153846154,
+        "pos": 0.5727482678983833,
+        "q": 0.27848101265822783
+      },
+      "eval_loss": 1.4238033294677734,
+      "eval_macro_average_f1": 0.6053866489655125,
+      "eval_micro_average_f1": 0.7358569051580699,
+      "eval_runtime": 7.4328,
+      "eval_samples_per_second": 323.433,
+      "eval_steps_per_second": 20.315,
+      "step": 7200
+    },
+    {
+      "epoch": 5.399408284023669,
+      "eval_class_f1": {
+        "neg": 0.7969348659003831,
+        "neu": 0.7798306389530408,
+        "pos": 0.5721212121212121,
+        "q": 0.3
+      },
+      "eval_loss": 1.4431192874908447,
+      "eval_macro_average_f1": 0.612221679243659,
+      "eval_micro_average_f1": 0.7408485856905158,
+      "eval_runtime": 7.3682,
+      "eval_samples_per_second": 326.266,
+      "eval_steps_per_second": 20.493,
+      "step": 7300
+    },
+    {
+      "epoch": 5.4733727810650885,
+      "eval_class_f1": {
+        "neg": 0.7940717628705148,
+        "neu": 0.783072817384674,
+        "pos": 0.5703883495145632,
+        "q": 0.3037974683544304
+      },
+      "eval_loss": 1.4316595792770386,
+      "eval_macro_average_f1": 0.6128325995310456,
+      "eval_micro_average_f1": 0.7416805324459235,
+      "eval_runtime": 7.3736,
+      "eval_samples_per_second": 326.03,
+      "eval_steps_per_second": 20.479,
+      "step": 7400
+    },
+    {
+      "epoch": 5.547337278106509,
+      "grad_norm": 0.4265735149383545,
+      "learning_rate": 2.543835616438356e-06,
+      "loss": 0.0933,
+      "step": 7500
+    },
+    {
+      "epoch": 5.547337278106509,
+      "eval_class_f1": {
+        "neg": 0.7978311386522074,
+        "neu": 0.7788089713843775,
+        "pos": 0.567409144196952,
+        "q": 0.30769230769230765
+      },
+      "eval_loss": 1.4399964809417725,
+      "eval_macro_average_f1": 0.6129353904814612,
+      "eval_micro_average_f1": 0.7387687188019967,
+      "eval_runtime": 7.2697,
+      "eval_samples_per_second": 330.689,
+      "eval_steps_per_second": 20.771,
+      "step": 7500
+    },
+    {
+      "epoch": 5.621301775147929,
+      "eval_class_f1": {
+        "neg": 0.7984375,
+        "neu": 0.7812379853902346,
+        "pos": 0.5714285714285714,
+        "q": 0.32500000000000007
+      },
+      "eval_loss": 1.4240373373031616,
+      "eval_macro_average_f1": 0.6190260142047015,
+      "eval_micro_average_f1": 0.7412645590682196,
+      "eval_runtime": 7.4341,
+      "eval_samples_per_second": 323.375,
+      "eval_steps_per_second": 20.312,
+      "step": 7600
+    },
+    {
+      "epoch": 5.695266272189349,
+      "eval_class_f1": {
+        "neg": 0.7987470634299139,
+        "neu": 0.7843286420692278,
+        "pos": 0.5703883495145632,
+        "q": 0.30769230769230765
+      },
+      "eval_loss": 1.4332064390182495,
+      "eval_macro_average_f1": 0.6152890906765031,
+      "eval_micro_average_f1": 0.7437603993344426,
+      "eval_runtime": 7.4434,
+      "eval_samples_per_second": 322.969,
+      "eval_steps_per_second": 20.286,
+      "step": 7700
+    },
+    {
+      "epoch": 5.769230769230769,
+      "eval_class_f1": {
+        "neg": 0.7981220657276996,
+        "neu": 0.781874039938556,
+        "pos": 0.5731132075471698,
+        "q": 0.30769230769230765
+      },
+      "eval_loss": 1.4344979524612427,
+      "eval_macro_average_f1": 0.6152004052264332,
+      "eval_micro_average_f1": 0.7416805324459235,
+      "eval_runtime": 7.3808,
+      "eval_samples_per_second": 325.708,
+      "eval_steps_per_second": 20.458,
+      "step": 7800
+    },
+    {
+      "epoch": 5.84319526627219,
+      "eval_class_f1": {
+        "neg": 0.7990654205607477,
+        "neu": 0.7815384615384616,
+        "pos": 0.5724465558194775,
+        "q": 0.3414634146341463
+      },
+      "eval_loss": 1.4412455558776855,
+      "eval_macro_average_f1": 0.6236284631382082,
+      "eval_micro_average_f1": 0.7420965058236273,
+      "eval_runtime": 7.3915,
+      "eval_samples_per_second": 325.237,
+      "eval_steps_per_second": 20.429,
+      "step": 7900
+    },
+    {
+      "epoch": 5.9171597633136095,
+      "grad_norm": 16.41318702697754,
+      "learning_rate": 4.89041095890411e-07,
+      "loss": 0.1006,
+      "step": 8000
+    },
+    {
+      "epoch": 5.9171597633136095,
+      "eval_class_f1": {
+        "neg": 0.7987519500780033,
+        "neu": 0.7813098429720413,
+        "pos": 0.5700598802395208,
+        "q": 0.32500000000000007
+      },
+      "eval_loss": 1.4469937086105347,
+      "eval_macro_average_f1": 0.6187804183223914,
+      "eval_micro_average_f1": 0.7416805324459235,
+      "eval_runtime": 7.3689,
+      "eval_samples_per_second": 326.236,
+      "eval_steps_per_second": 20.492,
+      "step": 8000
+    },
+    {
+      "epoch": 5.991124260355029,
+      "eval_class_f1": {
+        "neg": 0.7990654205607477,
+        "neu": 0.781441717791411,
+        "pos": 0.569377990430622,
+        "q": 0.32500000000000007
+      },
+      "eval_loss": 1.4454258680343628,
+      "eval_macro_average_f1": 0.6187212821956952,
+      "eval_micro_average_f1": 0.7416805324459235,
+      "eval_runtime": 7.448,
+      "eval_samples_per_second": 322.77,
+      "eval_steps_per_second": 20.274,
+      "step": 8100
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 8112,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 100,
+  "total_flos": 1.0485727069042368e+16,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d136bd3a72331c1ee1902b7c4073c9b6bb6f902bd3ce56bb0ff2ee4eab86fe5c
+size 5048