artemis13fowl commited on
Commit
1549a10
·
verified ·
1 Parent(s): a921726

Training in progress, epoch 1

Browse files
Files changed (3) hide show
  1. config.json +440 -440
  2. model.safetensors +1 -1
  3. training_args.bin +1 -1
config.json CHANGED
@@ -21,451 +21,451 @@
21
  "hidden_activation": "gelu",
22
  "hidden_size": 768,
23
  "id2label": {
24
- "0": "Lua",
25
- "1": "KiCad Layout",
26
- "10": "Rust",
27
- "100": "Unix Assembly",
28
- "101": "OpenType Feature File",
29
- "102": "Prolog",
30
- "103": "Protocol Buffer Text Format",
31
- "104": "HTML+Razor",
32
- "105": "Fortran Free Form",
33
- "106": "Logtalk",
34
- "107": "Kit",
35
- "108": "Graphviz (DOT)",
36
- "109": "Erlang",
37
- "11": "Scheme",
38
- "110": "C",
39
- "111": "LLVM",
40
- "112": "Visual Basic .NET",
41
- "113": "PHP",
42
- "114": "J",
43
- "115": "Ragel in Ruby Host",
44
- "116": "CoffeeScript",
45
- "117": "PlantUML",
46
- "118": "Vim Script",
47
- "119": "Go",
48
- "12": "FreeBasic",
49
- "120": "Vim Snippet",
50
- "121": "R",
51
- "122": "Dockerfile",
52
- "123": "Lex",
53
- "124": "G-code",
54
- "125": "Scilab",
55
- "126": "Csound",
56
- "127": "Redcode",
57
- "128": "VCL",
58
- "129": "Perl",
59
- "13": "Starlark",
60
- "130": "Java",
61
- "131": "Csound Document",
62
- "132": "Julia",
63
- "133": "Turtle",
64
- "134": "Gherkin",
65
- "135": "Smali",
66
- "136": "C++",
67
- "137": "Python",
68
- "138": "CODEOWNERS",
69
- "139": "Io",
70
- "14": "D",
71
- "140": "GDB",
72
- "141": "Makefile",
73
- "142": "Common Lisp",
74
- "143": "Apex",
75
- "144": "PostScript",
76
- "145": "Edoid",
77
- "146": "Unity3D Asset",
78
- "147": "OpenEdge ABL",
79
- "148": "Open Policy Agent",
80
- "149": "Sass",
81
- "15": "Raw token data",
82
- "150": "Rascal",
83
- "151": "Lean",
84
- "152": "Assembly",
85
- "153": "Inform 7",
86
- "154": "Lua",
87
- "155": "Chapel",
88
- "156": "KiCad Layout",
89
- "157": "Nim",
90
- "158": "Ignore List",
91
- "159": "Objective-C++",
92
- "16": "Modelica",
93
- "160": "Squirrel",
94
- "161": "Ruby",
95
- "162": "Objective-C",
96
- "163": "Rust",
97
- "164": "FreeBasic",
98
- "165": "Scheme",
99
- "166": "Starlark",
100
- "167": "D",
101
- "168": "Raw token data",
102
- "169": "Modelica",
103
- "17": "PicoLisp",
104
- "170": "Pickle",
105
- "171": "SCSS",
106
- "172": "Shell",
107
- "173": "PowerShell",
108
- "174": "TSX",
109
- "175": "C#",
110
- "176": "Gnuplot",
111
- "177": "Haxe",
112
- "178": "Go Module",
113
- "179": "Vue",
114
- "18": "SCSS",
115
- "180": "ObjDump",
116
- "181": "SQL",
117
- "182": "AsciiDoc",
118
- "183": "Fish",
119
- "184": "Wavefront Object",
120
- "185": "DIGITAL Command Language",
121
- "186": "Thrift",
122
- "187": "ApacheConf",
123
- "188": "Gerber Image",
124
- "189": "YANG",
125
- "19": "Shell",
126
- "190": "Groovy",
127
- "191": "MATLAB",
128
- "192": "Bluespec",
129
- "193": "POV-Ray SDL",
130
- "194": "Solidity",
131
- "195": "Dart",
132
- "196": "VBScript",
133
- "197": "Fluent",
134
- "198": "SQF",
135
- "199": "Gettext Catalog",
136
- "2": "Nim",
137
- "20": "PowerShell",
138
- "200": "Haskell",
139
- "201": "ImageJ Macro",
140
- "202": "PureScript",
141
- "203": "GDScript",
142
- "204": "Inno Setup",
143
- "205": "ColdFusion",
144
- "206": "TSV",
145
- "207": "Visual Basic",
146
- "208": "CMake",
147
- "209": "Eagle",
148
- "21": "TSX",
149
- "210": "Blade",
150
- "211": "OCaml",
151
- "212": "LookML",
152
- "213": "VHDL",
153
- "214": "HCL",
154
- "215": "GLSL",
155
- "216": "RDoc",
156
- "217": "Metal",
157
- "218": "Isabelle",
158
- "219": "TypeScript",
159
- "22": "C#",
160
- "220": "Swift",
161
- "221": "Stylus",
162
- "222": "Ada",
163
- "223": "Crystal",
164
- "224": "Less",
165
- "225": "Verilog",
166
- "226": "XML Property List",
167
- "227": "Hoon",
168
- "228": "Hack",
169
- "229": "PLpgSQL",
170
- "23": "Gnuplot",
171
- "230": "Scala",
172
- "231": "FreeMarker",
173
- "232": "Kotlin",
174
- "233": "GAS",
175
- "234": "Twig",
176
- "235": "EJS",
177
- "236": "JavaScript",
178
- "237": "Gradle",
179
- "238": "ECL",
180
- "239": "Jupyter Notebook",
181
- "24": "Haxe",
182
- "240": "OpenType Feature File",
183
- "241": "Protocol Buffer Text Format",
184
- "242": "HTML+Razor",
185
- "243": "Fortran Free Form",
186
- "244": "Logtalk",
187
- "245": "Kit",
188
- "246": "Graphviz (DOT)",
189
- "247": "Erlang",
190
- "248": "C",
191
- "249": "LLVM",
192
- "25": "Go Module",
193
- "250": "Visual Basic .NET",
194
- "251": "PHP",
195
- "252": "J",
196
- "253": "Ragel in Ruby Host",
197
- "254": "CoffeeScript",
198
- "255": "PlantUML",
199
- "256": "Vim Script",
200
- "257": "Go",
201
- "258": "Vim Snippet",
202
- "259": "R",
203
- "26": "Vue",
204
- "260": "Dockerfile",
205
- "261": "Lex",
206
- "262": "G-code",
207
- "263": "Scilab",
208
- "264": "Perl",
209
- "265": "Java",
210
- "266": "Julia",
211
- "267": "Turtle",
212
- "268": "Gherkin",
213
- "269": "C++",
214
- "27": "ObjDump",
215
- "270": "Python",
216
- "271": "Io",
217
- "272": "GDB",
218
- "273": "Makefile",
219
- "274": "Common Lisp",
220
- "275": "Apex",
221
- "276": "PostScript",
222
- "277": "Edoid",
223
- "278": "Unity3D Asset",
224
- "279": "OpenEdge ABL",
225
- "28": "SQL",
226
- "280": "Open Policy Agent",
227
- "281": "Sass",
228
- "282": "Rascal",
229
- "283": "Lean",
230
- "284": "Assembly",
231
- "29": "AsciiDoc",
232
- "3": "Chapel",
233
- "30": "Fish",
234
- "31": "Pascal",
235
- "32": "Wavefront Object",
236
- "33": "DIGITAL Command Language",
237
- "34": "Thrift",
238
- "35": "ApacheConf",
239
- "36": "Gerber Image",
240
- "37": "YANG",
241
- "38": "Mathematica",
242
- "39": "OpenStep Property List",
243
- "4": "Ignore List",
244
- "40": "GAP",
245
- "41": "Groovy",
246
- "42": "MATLAB",
247
- "43": "Adobe Font Metrics",
248
- "44": "Bluespec",
249
- "45": "POV-Ray SDL",
250
- "46": "Solidity",
251
- "47": "Dart",
252
- "48": "XS",
253
- "49": "VBScript",
254
- "5": "Objective-C++",
255
- "50": "Fluent",
256
- "51": "SQF",
257
- "52": "Gettext Catalog",
258
- "53": "JAR Manifest",
259
- "54": "Haskell",
260
- "55": "ImageJ Macro",
261
- "56": "Ioke",
262
- "57": "PureScript",
263
- "58": "REALbasic",
264
- "59": "GDScript",
265
- "6": "Squirrel",
266
- "60": "Inno Setup",
267
- "61": "ColdFusion",
268
- "62": "TSV",
269
- "63": "Visual Basic",
270
- "64": "Kvlang",
271
- "65": "CMake",
272
- "66": "Eagle",
273
- "67": "Git Config",
274
- "68": "Blade",
275
- "69": "OCaml",
276
- "7": "AGS Script",
277
- "70": "HCL",
278
- "71": "VHDL",
279
- "72": "LookML",
280
- "73": "GLSL",
281
- "74": "RDoc",
282
- "75": "Metal",
283
- "76": "Isabelle",
284
- "77": "TypeScript",
285
- "78": "Swift",
286
- "79": "Stylus",
287
- "8": "Ruby",
288
- "80": "Ada",
289
- "81": "Crystal",
290
- "82": "Less",
291
- "83": "JQ",
292
- "84": "Verilog",
293
- "85": "XML Property List",
294
- "86": "Hoon",
295
- "87": "Hack",
296
- "88": "PLpgSQL",
297
- "89": "Scala",
298
- "9": "Objective-C",
299
- "90": "FreeMarker",
300
- "91": "Unknown",
301
- "92": "Kotlin",
302
- "93": "GAS",
303
- "94": "Twig",
304
- "95": "EJS",
305
- "96": "JavaScript",
306
- "97": "Gradle",
307
- "98": "ECL",
308
- "99": "Jupyter Notebook"
309
  },
310
  "initializer_cutoff_factor": 2.0,
311
  "initializer_range": 0.02,
312
  "intermediate_size": 1152,
313
  "label2id": {
314
- "AGS Script": "7",
315
- "Ada": "222",
316
- "Adobe Font Metrics": "43",
317
- "ApacheConf": "187",
318
- "Apex": "275",
319
- "AsciiDoc": "182",
320
- "Assembly": "284",
321
- "Blade": "210",
322
- "Bluespec": "192",
323
- "C": "248",
324
- "C#": "175",
325
- "C++": "269",
326
- "CMake": "208",
327
- "CODEOWNERS": "138",
328
- "Chapel": "155",
329
- "CoffeeScript": "254",
330
- "ColdFusion": "205",
331
- "Common Lisp": "274",
332
- "Crystal": "223",
333
- "Csound": "126",
334
- "Csound Document": "131",
335
- "D": "167",
336
- "DIGITAL Command Language": "185",
337
- "Dart": "195",
338
- "Dockerfile": "260",
339
- "ECL": "238",
340
- "EJS": "235",
341
- "Eagle": "209",
342
- "Edoid": "277",
343
- "Erlang": "247",
344
- "Fish": "183",
345
- "Fluent": "197",
346
- "Fortran Free Form": "243",
347
- "FreeBasic": "164",
348
- "FreeMarker": "231",
349
- "G-code": "262",
350
- "GAP": "40",
351
- "GAS": "233",
352
- "GDB": "272",
353
- "GDScript": "203",
354
- "GLSL": "215",
355
- "Gerber Image": "188",
356
- "Gettext Catalog": "199",
357
- "Gherkin": "268",
358
- "Git Config": "67",
359
- "Gnuplot": "176",
360
- "Go": "257",
361
- "Go Module": "178",
362
- "Gradle": "237",
363
- "Graphviz (DOT)": "246",
364
- "Groovy": "190",
365
- "HCL": "214",
366
- "HTML+Razor": "242",
367
- "Hack": "228",
368
- "Haskell": "200",
369
- "Haxe": "177",
370
- "Hoon": "227",
371
- "Ignore List": "158",
372
- "ImageJ Macro": "201",
373
- "Inform 7": "153",
374
- "Inno Setup": "204",
375
- "Io": "271",
376
- "Ioke": "56",
377
- "Isabelle": "218",
378
- "J": "252",
379
- "JAR Manifest": "53",
380
- "JQ": "83",
381
- "Java": "265",
382
- "JavaScript": "236",
383
- "Julia": "266",
384
- "Jupyter Notebook": "239",
385
- "KiCad Layout": "156",
386
- "Kit": "245",
387
- "Kotlin": "232",
388
- "Kvlang": "64",
389
- "LLVM": "249",
390
- "Lean": "283",
391
- "Less": "224",
392
- "Lex": "261",
393
- "Logtalk": "244",
394
- "LookML": "212",
395
- "Lua": "154",
396
- "MATLAB": "191",
397
- "Makefile": "273",
398
- "Mathematica": "38",
399
- "Metal": "217",
400
- "Modelica": "169",
401
- "Nim": "157",
402
- "OCaml": "211",
403
- "ObjDump": "180",
404
- "Objective-C": "162",
405
- "Objective-C++": "159",
406
- "Open Policy Agent": "280",
407
- "OpenEdge ABL": "279",
408
- "OpenStep Property List": "39",
409
- "OpenType Feature File": "240",
410
- "PHP": "251",
411
- "PLpgSQL": "229",
412
- "POV-Ray SDL": "193",
413
- "Pascal": "31",
414
- "Perl": "264",
415
- "Pickle": "170",
416
- "PicoLisp": "17",
417
- "PlantUML": "255",
418
- "PostScript": "276",
419
- "PowerShell": "173",
420
- "Prolog": "102",
421
- "Protocol Buffer Text Format": "241",
422
- "PureScript": "202",
423
- "Python": "270",
424
- "R": "259",
425
- "RDoc": "216",
426
- "REALbasic": "58",
427
- "Ragel in Ruby Host": "253",
428
- "Rascal": "282",
429
- "Raw token data": "168",
430
- "Redcode": "127",
431
- "Ruby": "161",
432
- "Rust": "163",
433
- "SCSS": "171",
434
- "SQF": "198",
435
- "SQL": "181",
436
- "Sass": "281",
437
- "Scala": "230",
438
- "Scheme": "165",
439
- "Scilab": "263",
440
- "Shell": "172",
441
- "Smali": "135",
442
- "Solidity": "194",
443
- "Squirrel": "160",
444
- "Starlark": "166",
445
- "Stylus": "221",
446
- "Swift": "220",
447
- "TSV": "206",
448
- "TSX": "174",
449
- "Thrift": "186",
450
- "Turtle": "267",
451
- "Twig": "234",
452
- "TypeScript": "219",
453
- "Unity3D Asset": "278",
454
- "Unix Assembly": "100",
455
- "Unknown": "91",
456
- "VBScript": "196",
457
- "VCL": "128",
458
- "VHDL": "213",
459
- "Verilog": "225",
460
- "Vim Script": "256",
461
- "Vim Snippet": "258",
462
- "Visual Basic": "207",
463
- "Visual Basic .NET": "250",
464
- "Vue": "179",
465
- "Wavefront Object": "184",
466
- "XML Property List": "226",
467
- "XS": "48",
468
- "YANG": "189"
469
  },
470
  "layer_norm_eps": 1e-05,
471
  "local_attention": 128,
 
21
  "hidden_activation": "gelu",
22
  "hidden_size": 768,
23
  "id2label": {
24
+ "0": 7,
25
+ "1": 17,
26
+ "10": 58,
27
+ "100": 231,
28
+ "101": 232,
29
+ "102": 233,
30
+ "103": 234,
31
+ "104": 235,
32
+ "105": 236,
33
+ "106": 237,
34
+ "107": 238,
35
+ "108": 239,
36
+ "109": 240,
37
+ "11": 64,
38
+ "110": 241,
39
+ "111": 242,
40
+ "112": 243,
41
+ "113": 244,
42
+ "114": 245,
43
+ "115": 246,
44
+ "116": 247,
45
+ "117": 248,
46
+ "118": 249,
47
+ "119": 250,
48
+ "12": 67,
49
+ "120": 251,
50
+ "121": 252,
51
+ "122": 253,
52
+ "123": 254,
53
+ "124": 255,
54
+ "125": 256,
55
+ "126": 257,
56
+ "127": 258,
57
+ "128": 259,
58
+ "129": 260,
59
+ "13": 83,
60
+ "130": 261,
61
+ "131": 262,
62
+ "132": 263,
63
+ "133": 264,
64
+ "134": 265,
65
+ "135": 266,
66
+ "136": 267,
67
+ "137": 268,
68
+ "138": 269,
69
+ "139": 270,
70
+ "14": 91,
71
+ "140": 271,
72
+ "141": 272,
73
+ "142": 273,
74
+ "143": 274,
75
+ "144": 275,
76
+ "145": 276,
77
+ "146": 277,
78
+ "147": 278,
79
+ "148": 279,
80
+ "149": 280,
81
+ "15": 100,
82
+ "150": 281,
83
+ "151": 282,
84
+ "152": 283,
85
+ "153": 284,
86
+ "154": 154,
87
+ "155": 155,
88
+ "156": 156,
89
+ "157": 157,
90
+ "158": 158,
91
+ "159": 159,
92
+ "16": 102,
93
+ "160": 160,
94
+ "161": 161,
95
+ "162": 162,
96
+ "163": 163,
97
+ "164": 164,
98
+ "165": 165,
99
+ "166": 166,
100
+ "167": 167,
101
+ "168": 168,
102
+ "169": 169,
103
+ "17": 126,
104
+ "170": 170,
105
+ "171": 171,
106
+ "172": 172,
107
+ "173": 173,
108
+ "174": 174,
109
+ "175": 175,
110
+ "176": 176,
111
+ "177": 177,
112
+ "178": 178,
113
+ "179": 179,
114
+ "18": 127,
115
+ "180": 180,
116
+ "181": 181,
117
+ "182": 182,
118
+ "183": 183,
119
+ "184": 184,
120
+ "185": 185,
121
+ "186": 186,
122
+ "187": 187,
123
+ "188": 188,
124
+ "189": 189,
125
+ "19": 128,
126
+ "190": 190,
127
+ "191": 191,
128
+ "192": 192,
129
+ "193": 193,
130
+ "194": 194,
131
+ "195": 195,
132
+ "196": 196,
133
+ "197": 197,
134
+ "198": 198,
135
+ "199": 199,
136
+ "2": 31,
137
+ "20": 131,
138
+ "200": 200,
139
+ "201": 201,
140
+ "202": 202,
141
+ "203": 203,
142
+ "204": 204,
143
+ "205": 205,
144
+ "206": 206,
145
+ "207": 207,
146
+ "208": 208,
147
+ "209": 209,
148
+ "21": 135,
149
+ "210": 210,
150
+ "211": 211,
151
+ "212": 212,
152
+ "213": 213,
153
+ "214": 214,
154
+ "215": 215,
155
+ "216": 216,
156
+ "217": 217,
157
+ "218": 218,
158
+ "219": 219,
159
+ "22": 138,
160
+ "220": 220,
161
+ "221": 221,
162
+ "222": 222,
163
+ "223": 223,
164
+ "224": 224,
165
+ "225": 225,
166
+ "226": 226,
167
+ "227": 227,
168
+ "228": 228,
169
+ "229": 229,
170
+ "23": 153,
171
+ "230": 230,
172
+ "231": 231,
173
+ "232": 232,
174
+ "233": 233,
175
+ "234": 234,
176
+ "235": 235,
177
+ "236": 236,
178
+ "237": 237,
179
+ "238": 238,
180
+ "239": 239,
181
+ "24": 154,
182
+ "240": 240,
183
+ "241": 241,
184
+ "242": 242,
185
+ "243": 243,
186
+ "244": 244,
187
+ "245": 245,
188
+ "246": 246,
189
+ "247": 247,
190
+ "248": 248,
191
+ "249": 249,
192
+ "25": 155,
193
+ "250": 250,
194
+ "251": 251,
195
+ "252": 252,
196
+ "253": 253,
197
+ "254": 254,
198
+ "255": 255,
199
+ "256": 256,
200
+ "257": 257,
201
+ "258": 258,
202
+ "259": 259,
203
+ "26": 156,
204
+ "260": 260,
205
+ "261": 261,
206
+ "262": 262,
207
+ "263": 263,
208
+ "264": 264,
209
+ "265": 265,
210
+ "266": 266,
211
+ "267": 267,
212
+ "268": 268,
213
+ "269": 269,
214
+ "27": 157,
215
+ "270": 270,
216
+ "271": 271,
217
+ "272": 272,
218
+ "273": 273,
219
+ "274": 274,
220
+ "275": 275,
221
+ "276": 276,
222
+ "277": 277,
223
+ "278": 278,
224
+ "279": 279,
225
+ "28": 158,
226
+ "280": 280,
227
+ "281": 281,
228
+ "282": 282,
229
+ "283": 283,
230
+ "284": 284,
231
+ "29": 159,
232
+ "3": 38,
233
+ "30": 160,
234
+ "31": 161,
235
+ "32": 162,
236
+ "33": 163,
237
+ "34": 164,
238
+ "35": 165,
239
+ "36": 166,
240
+ "37": 167,
241
+ "38": 168,
242
+ "39": 169,
243
+ "4": 39,
244
+ "40": 171,
245
+ "41": 172,
246
+ "42": 173,
247
+ "43": 174,
248
+ "44": 175,
249
+ "45": 176,
250
+ "46": 177,
251
+ "47": 178,
252
+ "48": 179,
253
+ "49": 180,
254
+ "5": 40,
255
+ "50": 181,
256
+ "51": 182,
257
+ "52": 183,
258
+ "53": 184,
259
+ "54": 185,
260
+ "55": 186,
261
+ "56": 187,
262
+ "57": 188,
263
+ "58": 189,
264
+ "59": 190,
265
+ "6": 43,
266
+ "60": 191,
267
+ "61": 192,
268
+ "62": 193,
269
+ "63": 194,
270
+ "64": 195,
271
+ "65": 196,
272
+ "66": 197,
273
+ "67": 198,
274
+ "68": 199,
275
+ "69": 200,
276
+ "7": 48,
277
+ "70": 201,
278
+ "71": 202,
279
+ "72": 203,
280
+ "73": 204,
281
+ "74": 205,
282
+ "75": 206,
283
+ "76": 207,
284
+ "77": 208,
285
+ "78": 209,
286
+ "79": 210,
287
+ "8": 53,
288
+ "80": 211,
289
+ "81": 212,
290
+ "82": 213,
291
+ "83": 214,
292
+ "84": 215,
293
+ "85": 216,
294
+ "86": 217,
295
+ "87": 218,
296
+ "88": 219,
297
+ "89": 220,
298
+ "9": 56,
299
+ "90": 221,
300
+ "91": 222,
301
+ "92": 223,
302
+ "93": 224,
303
+ "94": 225,
304
+ "95": 226,
305
+ "96": 227,
306
+ "97": 228,
307
+ "98": 229,
308
+ "99": 230
309
  },
310
  "initializer_cutoff_factor": 2.0,
311
  "initializer_range": 0.02,
312
  "intermediate_size": 1152,
313
  "label2id": {
314
+ "7": "0",
315
+ "17": "1",
316
+ "31": "2",
317
+ "38": "3",
318
+ "39": "4",
319
+ "40": "5",
320
+ "43": "6",
321
+ "48": "7",
322
+ "53": "8",
323
+ "56": "9",
324
+ "58": "10",
325
+ "64": "11",
326
+ "67": "12",
327
+ "83": "13",
328
+ "91": "14",
329
+ "100": "15",
330
+ "102": "16",
331
+ "126": "17",
332
+ "127": "18",
333
+ "128": "19",
334
+ "131": "20",
335
+ "135": "21",
336
+ "138": "22",
337
+ "153": "23",
338
+ "154": "154",
339
+ "155": "155",
340
+ "156": "156",
341
+ "157": "157",
342
+ "158": "158",
343
+ "159": "159",
344
+ "160": "160",
345
+ "161": "161",
346
+ "162": "162",
347
+ "163": "163",
348
+ "164": "164",
349
+ "165": "165",
350
+ "166": "166",
351
+ "167": "167",
352
+ "168": "168",
353
+ "169": "169",
354
+ "170": "170",
355
+ "171": "171",
356
+ "172": "172",
357
+ "173": "173",
358
+ "174": "174",
359
+ "175": "175",
360
+ "176": "176",
361
+ "177": "177",
362
+ "178": "178",
363
+ "179": "179",
364
+ "180": "180",
365
+ "181": "181",
366
+ "182": "182",
367
+ "183": "183",
368
+ "184": "184",
369
+ "185": "185",
370
+ "186": "186",
371
+ "187": "187",
372
+ "188": "188",
373
+ "189": "189",
374
+ "190": "190",
375
+ "191": "191",
376
+ "192": "192",
377
+ "193": "193",
378
+ "194": "194",
379
+ "195": "195",
380
+ "196": "196",
381
+ "197": "197",
382
+ "198": "198",
383
+ "199": "199",
384
+ "200": "200",
385
+ "201": "201",
386
+ "202": "202",
387
+ "203": "203",
388
+ "204": "204",
389
+ "205": "205",
390
+ "206": "206",
391
+ "207": "207",
392
+ "208": "208",
393
+ "209": "209",
394
+ "210": "210",
395
+ "211": "211",
396
+ "212": "212",
397
+ "213": "213",
398
+ "214": "214",
399
+ "215": "215",
400
+ "216": "216",
401
+ "217": "217",
402
+ "218": "218",
403
+ "219": "219",
404
+ "220": "220",
405
+ "221": "221",
406
+ "222": "222",
407
+ "223": "223",
408
+ "224": "224",
409
+ "225": "225",
410
+ "226": "226",
411
+ "227": "227",
412
+ "228": "228",
413
+ "229": "229",
414
+ "230": "230",
415
+ "231": "231",
416
+ "232": "232",
417
+ "233": "233",
418
+ "234": "234",
419
+ "235": "235",
420
+ "236": "236",
421
+ "237": "237",
422
+ "238": "238",
423
+ "239": "239",
424
+ "240": "240",
425
+ "241": "241",
426
+ "242": "242",
427
+ "243": "243",
428
+ "244": "244",
429
+ "245": "245",
430
+ "246": "246",
431
+ "247": "247",
432
+ "248": "248",
433
+ "249": "249",
434
+ "250": "250",
435
+ "251": "251",
436
+ "252": "252",
437
+ "253": "253",
438
+ "254": "254",
439
+ "255": "255",
440
+ "256": "256",
441
+ "257": "257",
442
+ "258": "258",
443
+ "259": "259",
444
+ "260": "260",
445
+ "261": "261",
446
+ "262": "262",
447
+ "263": "263",
448
+ "264": "264",
449
+ "265": "265",
450
+ "266": "266",
451
+ "267": "267",
452
+ "268": "268",
453
+ "269": "269",
454
+ "270": "270",
455
+ "271": "271",
456
+ "272": "272",
457
+ "273": "273",
458
+ "274": "274",
459
+ "275": "275",
460
+ "276": "276",
461
+ "277": "277",
462
+ "278": "278",
463
+ "279": "279",
464
+ "280": "280",
465
+ "281": "281",
466
+ "282": "282",
467
+ "283": "283",
468
+ "284": "284"
469
  },
470
  "layer_norm_eps": 1e-05,
471
  "local_attention": 128,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:95ef29622e699f95a585e6ef1263168d7fb2113107f8d1bec1f3fc3186c5517b
3
  size 599310308
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe6037a95b51ab55eca90629e34cd687da59993e01ed8b9f1cb15075868a4514
3
  size 599310308
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:78ada9de9dfea8d6abc067a6c0e8e2154382080a2f4def52676217d8bb1155b4
3
  size 5905
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ad0181d394bb3507dc8d789366400bf5358f02e775c63324094924065e12b79
3
  size 5905