my_awesome_model / config.json
yangwooko's picture
Training in progress, epoch 1
9a29eb1
{
"_name_or_path": "distilbert-base-uncased",
"activation": "gelu",
"architectures": [
"DistilBertForSequenceClassification"
],
"attention_dropout": 0.1,
"dim": 768,
"dropout": 0.1,
"hidden_dim": 3072,
"id2label": {
"0": "32bit",
"1": "admin",
"2": "algos",
"3": "api - consistency",
"4": "api design",
"5": "apply",
"6": "arm",
"7": "arraymanager",
"8": "benchmark",
"9": "blocker",
"10": "bug",
"11": "build",
"12": "categorical",
"13": "ci",
"14": "clean",
"15": "closing candidate",
"16": "code style",
"17": "community",
"18": "compat",
"19": "complex",
"20": "constructors",
"21": "copy / view semantics",
"22": "cut",
"23": "dataframe",
"24": "dependencies",
"25": "deprecate",
"26": "docs",
"27": "dtype conversions",
"28": "duplicate report",
"29": "enhancement",
"30": "error reporting",
"31": "expressions",
"32": "extensionarray",
"33": "frequency",
"34": "good first issue",
"35": "groupby",
"36": "ideas",
"37": "index",
"38": "indexing",
"39": "internals",
"40": "interval",
"41": "io csv",
"42": "io data",
"43": "io excel",
"44": "io google",
"45": "io hdf5",
"46": "io html",
"47": "io json",
"48": "io latex",
"49": "io network",
"50": "io parquet",
"51": "io pickle",
"52": "io sas",
"53": "io sql",
"54": "io stata",
"55": "io xml",
"56": "master tracker",
"57": "metadata",
"58": "missing-data",
"59": "mothballed",
"60": "multiindex",
"61": "multithreading",
"62": "na - maskedarrays",
"63": "needs discussion",
"64": "needs info",
"65": "needs review",
"66": "needs tests",
"67": "numba",
"68": "numeric operations",
"69": "output-formatting",
"70": "performance",
"71": "period",
"72": "python 3.10",
"73": "python 3.8",
"74": "python 3.9",
"75": "quantile",
"76": "reduction operations",
"77": "refactor",
"78": "regression",
"79": "release",
"80": "replace",
"81": "resample",
"82": "reshaping",
"83": "roadmap",
"84": "segfault",
"85": "series",
"86": "sparse",
"87": "stale",
"88": "strings",
"89": "styler",
"90": "testing",
"91": "timedelta",
"92": "timeseries",
"93": "timezones",
"94": "typing",
"95": "unicode",
"96": "unreliable test",
"97": "upstream issue",
"98": "usage question",
"99": "visualization",
"100": "warnings",
"101": "web",
"102": "window",
"103": "windows"
},
"initializer_range": 0.02,
"label2id": {
"32bit": 0,
"admin": 1,
"algos": 2,
"api - consistency": 3,
"api design": 4,
"apply": 5,
"arm": 6,
"arraymanager": 7,
"benchmark": 8,
"blocker": 9,
"bug": 10,
"build": 11,
"categorical": 12,
"ci": 13,
"clean": 14,
"closing candidate": 15,
"code style": 16,
"community": 17,
"compat": 18,
"complex": 19,
"constructors": 20,
"copy / view semantics": 21,
"cut": 22,
"dataframe": 23,
"dependencies": 24,
"deprecate": 25,
"docs": 26,
"dtype conversions": 27,
"duplicate report": 28,
"enhancement": 29,
"error reporting": 30,
"expressions": 31,
"extensionarray": 32,
"frequency": 33,
"good first issue": 34,
"groupby": 35,
"ideas": 36,
"index": 37,
"indexing": 38,
"internals": 39,
"interval": 40,
"io csv": 41,
"io data": 42,
"io excel": 43,
"io google": 44,
"io hdf5": 45,
"io html": 46,
"io json": 47,
"io latex": 48,
"io network": 49,
"io parquet": 50,
"io pickle": 51,
"io sas": 52,
"io sql": 53,
"io stata": 54,
"io xml": 55,
"master tracker": 56,
"metadata": 57,
"missing-data": 58,
"mothballed": 59,
"multiindex": 60,
"multithreading": 61,
"na - maskedarrays": 62,
"needs discussion": 63,
"needs info": 64,
"needs review": 65,
"needs tests": 66,
"numba": 67,
"numeric operations": 68,
"output-formatting": 69,
"performance": 70,
"period": 71,
"python 3.10": 72,
"python 3.8": 73,
"python 3.9": 74,
"quantile": 75,
"reduction operations": 76,
"refactor": 77,
"regression": 78,
"release": 79,
"replace": 80,
"resample": 81,
"reshaping": 82,
"roadmap": 83,
"segfault": 84,
"series": 85,
"sparse": 86,
"stale": 87,
"strings": 88,
"styler": 89,
"testing": 90,
"timedelta": 91,
"timeseries": 92,
"timezones": 93,
"typing": 94,
"unicode": 95,
"unreliable test": 96,
"upstream issue": 97,
"usage question": 98,
"visualization": 99,
"warnings": 100,
"web": 101,
"window": 102,
"windows": 103
},
"max_position_embeddings": 512,
"model_type": "distilbert",
"n_heads": 12,
"n_layers": 6,
"pad_token_id": 0,
"problem_type": "single_label_classification",
"qa_dropout": 0.1,
"seq_classif_dropout": 0.2,
"sinusoidal_pos_embds": false,
"tie_weights_": true,
"torch_dtype": "float32",
"transformers_version": "4.26.1",
"vocab_size": 30522
}