tomercagan's picture
Training in progress, epoch 1
387b052 verified
raw
history blame
10.8 kB
{
"architectures": [
"ModernBertForSequenceClassification"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 50281,
"classifier_activation": "gelu",
"classifier_bias": false,
"classifier_dropout": 0.0,
"classifier_pooling": "mean",
"cls_token_id": 50281,
"decoder_bias": true,
"deterministic_flash_attn": false,
"dtype": "float32",
"embedding_dropout": 0.0,
"eos_token_id": 50282,
"global_attn_every_n_layers": 3,
"global_rope_theta": 160000.0,
"gradient_checkpointing": false,
"hidden_activation": "gelu",
"hidden_size": 768,
"id2label": {
"0": "Scheme",
"1": "R",
"2": "Protocol Buffer Text Format",
"3": "Java",
"4": "J",
"5": "Erlang",
"6": "Smali",
"7": "Vim Snippet",
"8": "ApacheConf",
"9": "Edoid",
"10": "Csound Document",
"11": "Gherkin",
"12": "Unknown",
"13": "GLSL",
"14": "Bluespec",
"15": "VCL",
"16": "DIGITAL Command Language",
"17": "Redcode",
"18": "POV-Ray SDL",
"19": "Objective-C",
"20": "HCL",
"21": "Inform 7",
"22": "Groovy",
"23": "TypeScript",
"24": "C#",
"25": "XML Property List",
"26": "GAP",
"27": "Csound",
"28": "Python",
"29": "Assembly",
"30": "FreeBasic",
"31": "Fish",
"32": "MATLAB",
"33": "Thrift",
"34": "G-code",
"35": "HTML+Razor",
"36": "Hoon",
"37": "Vim Script",
"38": "Scilab",
"39": "PHP",
"40": "CMake",
"41": "Ragel in Ruby Host",
"42": "Prolog",
"43": "Fortran Free Form",
"44": "PowerShell",
"45": "FreeMarker",
"46": "Metal",
"47": "PostScript",
"48": "EJS",
"49": "Vue",
"50": "VBScript",
"51": "Mathematica",
"52": "Rust",
"53": "Fluent",
"54": "Starlark",
"55": "LookML",
"56": "Crystal",
"57": "Modelica",
"58": "TSV",
"59": "PicoLisp",
"60": "Blade",
"61": "Isabelle",
"62": "Objective-C++",
"63": "Eagle",
"64": "Ruby",
"65": "Inno Setup",
"66": "Gradle",
"67": "Visual Basic",
"68": "KiCad Layout",
"69": "AsciiDoc",
"70": "Gnuplot",
"71": "CODEOWNERS",
"72": "Git Config",
"73": "XS",
"74": "OpenStep Property List",
"75": "OCaml",
"76": "Go",
"77": "Lean",
"78": "GDScript",
"79": "Ignore List",
"80": "RDoc",
"81": "Swift",
"82": "Visual Basic .NET",
"83": "GAS",
"84": "Unity3D Asset",
"85": "Gerber Image",
"86": "PureScript",
"87": "SCSS",
"88": "Scala",
"89": "C",
"90": "Verilog",
"91": "Lua",
"92": "OpenEdge ABL",
"93": "TSX",
"94": "Nim",
"95": "GDB",
"96": "Solidity",
"97": "ObjDump",
"98": "Ada",
"99": "ImageJ Macro",
"100": "REALbasic",
"101": "VHDL",
"102": "Lex",
"103": "YANG",
"104": "Hack",
"105": "Rascal",
"106": "AGS Script",
"107": "Sass",
"108": "JAR Manifest",
"109": "Kotlin",
"110": "Dockerfile",
"111": "SQL",
"112": "ColdFusion",
"113": "Chapel",
"114": "Open Policy Agent",
"115": "PLpgSQL",
"116": "Twig",
"117": "CoffeeScript",
"118": "Haskell",
"119": "JavaScript",
"120": "ECL",
"121": "Adobe Font Metrics",
"122": "OpenType Feature File",
"123": "Logtalk",
"124": "Ioke",
"125": "Haxe",
"126": "Raw token data",
"127": "Apex",
"128": "D",
"129": "Perl",
"130": "Jupyter Notebook",
"131": "LLVM",
"132": "SQF",
"133": "Less",
"134": "Makefile",
"135": "Io",
"136": "JQ",
"137": "Stylus",
"138": "C++",
"139": "PlantUML",
"140": "Kit",
"141": "Shell",
"142": "Julia",
"143": "Kvlang",
"144": "Common Lisp",
"145": "Wavefront Object",
"146": "Squirrel",
"147": "Gettext Catalog",
"148": "Dart",
"149": "Unix Assembly",
"150": "Pascal",
"151": "Turtle",
"152": "Go Module",
"153": "Graphviz (DOT)",
"154": "Scheme",
"155": "R",
"156": "Protocol Buffer Text Format",
"157": "Java",
"158": "Erlang",
"159": "J",
"160": "Vim Snippet",
"161": "ApacheConf",
"162": "Edoid",
"163": "Gherkin",
"164": "GLSL",
"165": "Bluespec",
"166": "DIGITAL Command Language",
"167": "POV-Ray SDL",
"168": "Objective-C",
"169": "HCL",
"170": "Groovy",
"171": "TypeScript",
"172": "C#",
"173": "XML Property List",
"174": "Python",
"175": "Assembly",
"176": "FreeBasic",
"177": "Fish",
"178": "MATLAB",
"179": "Thrift",
"180": "G-code",
"181": "HTML+Razor",
"182": "Hoon",
"183": "Vim Script",
"184": "Scilab",
"185": "PHP",
"186": "CMake",
"187": "Ragel in Ruby Host",
"188": "Fortran Free Form",
"189": "PowerShell",
"190": "FreeMarker",
"191": "Metal",
"192": "PostScript",
"193": "EJS",
"194": "Vue",
"195": "VBScript",
"196": "Rust",
"197": "Fluent",
"198": "Starlark",
"199": "LookML",
"200": "Crystal",
"201": "Modelica",
"202": "TSV",
"203": "Blade",
"204": "Isabelle",
"205": "Objective-C++",
"206": "Eagle",
"207": "Ruby",
"208": "Inno Setup",
"209": "Gradle",
"210": "AsciiDoc",
"211": "Visual Basic",
"212": "KiCad Layout",
"213": "Gnuplot",
"214": "OCaml",
"215": "Go",
"216": "Lean",
"217": "GDScript",
"218": "Ignore List",
"219": "RDoc",
"220": "Swift",
"221": "Visual Basic .NET",
"222": "GAS",
"223": "Unity3D Asset",
"224": "Gerber Image",
"225": "PureScript",
"226": "SCSS",
"227": "Scala",
"228": "C",
"229": "Verilog",
"230": "Lua",
"231": "OpenEdge ABL",
"232": "TSX",
"233": "Nim",
"234": "GDB",
"235": "Solidity",
"236": "ObjDump",
"237": "Ada",
"238": "ImageJ Macro",
"239": "VHDL",
"240": "Lex",
"241": "YANG",
"242": "Hack",
"243": "Rascal",
"244": "Sass",
"245": "Kotlin",
"246": "Dockerfile",
"247": "SQL",
"248": "ColdFusion",
"249": "Open Policy Agent",
"250": "Chapel",
"251": "PLpgSQL",
"252": "Twig",
"253": "CoffeeScript",
"254": "Haskell",
"255": "JavaScript",
"256": "ECL",
"257": "OpenType Feature File",
"258": "Logtalk",
"259": "Haxe",
"260": "Raw token data",
"261": "Apex",
"262": "D",
"263": "Perl",
"264": "Jupyter Notebook",
"265": "LLVM",
"266": "SQF",
"267": "Less",
"268": "Makefile",
"269": "Io",
"270": "Pickle",
"271": "Stylus",
"272": "C++",
"273": "PlantUML",
"274": "Kit",
"275": "Julia",
"276": "Shell",
"277": "Wavefront Object",
"278": "Common Lisp",
"279": "Squirrel",
"280": "Gettext Catalog",
"281": "Dart",
"282": "Turtle",
"283": "Go Module",
"284": "Graphviz (DOT)"
},
"initializer_cutoff_factor": 2.0,
"initializer_range": 0.02,
"intermediate_size": 1152,
"label2id": {
"AGS Script": 106,
"Ada": 237,
"Adobe Font Metrics": 121,
"ApacheConf": 161,
"Apex": 261,
"AsciiDoc": 210,
"Assembly": 175,
"Blade": 203,
"Bluespec": 165,
"C": 228,
"C#": 172,
"C++": 272,
"CMake": 186,
"CODEOWNERS": 71,
"Chapel": 250,
"CoffeeScript": 253,
"ColdFusion": 248,
"Common Lisp": 278,
"Crystal": 200,
"Csound": 27,
"Csound Document": 10,
"D": 262,
"DIGITAL Command Language": 166,
"Dart": 281,
"Dockerfile": 246,
"ECL": 256,
"EJS": 193,
"Eagle": 206,
"Edoid": 162,
"Erlang": 158,
"Fish": 177,
"Fluent": 197,
"Fortran Free Form": 188,
"FreeBasic": 176,
"FreeMarker": 190,
"G-code": 180,
"GAP": 26,
"GAS": 222,
"GDB": 234,
"GDScript": 217,
"GLSL": 164,
"Gerber Image": 224,
"Gettext Catalog": 280,
"Gherkin": 163,
"Git Config": 72,
"Gnuplot": 213,
"Go": 215,
"Go Module": 283,
"Gradle": 209,
"Graphviz (DOT)": 284,
"Groovy": 170,
"HCL": 169,
"HTML+Razor": 181,
"Hack": 242,
"Haskell": 254,
"Haxe": 259,
"Hoon": 182,
"Ignore List": 218,
"ImageJ Macro": 238,
"Inform 7": 21,
"Inno Setup": 208,
"Io": 269,
"Ioke": 124,
"Isabelle": 204,
"J": 159,
"JAR Manifest": 108,
"JQ": 136,
"Java": 157,
"JavaScript": 255,
"Julia": 275,
"Jupyter Notebook": 264,
"KiCad Layout": 212,
"Kit": 274,
"Kotlin": 245,
"Kvlang": 143,
"LLVM": 265,
"Lean": 216,
"Less": 267,
"Lex": 240,
"Logtalk": 258,
"LookML": 199,
"Lua": 230,
"MATLAB": 178,
"Makefile": 268,
"Mathematica": 51,
"Metal": 191,
"Modelica": 201,
"Nim": 233,
"OCaml": 214,
"ObjDump": 236,
"Objective-C": 168,
"Objective-C++": 205,
"Open Policy Agent": 249,
"OpenEdge ABL": 231,
"OpenStep Property List": 74,
"OpenType Feature File": 257,
"PHP": 185,
"PLpgSQL": 251,
"POV-Ray SDL": 167,
"Pascal": 150,
"Perl": 263,
"Pickle": 270,
"PicoLisp": 59,
"PlantUML": 273,
"PostScript": 192,
"PowerShell": 189,
"Prolog": 42,
"Protocol Buffer Text Format": 156,
"PureScript": 225,
"Python": 174,
"R": 155,
"RDoc": 219,
"REALbasic": 100,
"Ragel in Ruby Host": 187,
"Rascal": 243,
"Raw token data": 260,
"Redcode": 17,
"Ruby": 207,
"Rust": 196,
"SCSS": 226,
"SQF": 266,
"SQL": 247,
"Sass": 244,
"Scala": 227,
"Scheme": 154,
"Scilab": 184,
"Shell": 276,
"Smali": 6,
"Solidity": 235,
"Squirrel": 279,
"Starlark": 198,
"Stylus": 271,
"Swift": 220,
"TSV": 202,
"TSX": 232,
"Thrift": 179,
"Turtle": 282,
"Twig": 252,
"TypeScript": 171,
"Unity3D Asset": 223,
"Unix Assembly": 149,
"Unknown": 12,
"VBScript": 195,
"VCL": 15,
"VHDL": 239,
"Verilog": 229,
"Vim Script": 183,
"Vim Snippet": 160,
"Visual Basic": 211,
"Visual Basic .NET": 221,
"Vue": 194,
"Wavefront Object": 277,
"XML Property List": 173,
"XS": 73,
"YANG": 241
},
"layer_norm_eps": 1e-05,
"local_attention": 128,
"local_rope_theta": 10000.0,
"max_position_embeddings": 8192,
"mlp_bias": false,
"mlp_dropout": 0.0,
"model_type": "modernbert",
"norm_bias": false,
"norm_eps": 1e-05,
"num_attention_heads": 12,
"num_hidden_layers": 22,
"pad_token_id": 50283,
"position_embedding_type": "absolute",
"problem_type": "single_label_classification",
"repad_logits_with_grad": false,
"sep_token_id": 50282,
"sparse_pred_ignore_index": -100,
"sparse_prediction": false,
"transformers_version": "4.57.1",
"vocab_size": 50368
}