darklorddad commited on
Commit
2e61e4b
·
verified ·
1 Parent(s): f90ee2a

Upload 5 files

Browse files
checkpoint-407/config.json ADDED
@@ -0,0 +1,464 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "microsoft/swinv2-large-patch4-window12-192-22k",
3
+ "_num_labels": 200,
4
+ "architectures": [
5
+ "Swinv2ForImageClassification"
6
+ ],
7
+ "attention_probs_dropout_prob": 0.0,
8
+ "depths": [
9
+ 2,
10
+ 2,
11
+ 18,
12
+ 2
13
+ ],
14
+ "drop_path_rate": 0.1,
15
+ "embed_dim": 192,
16
+ "encoder_stride": 32,
17
+ "hidden_act": "gelu",
18
+ "hidden_dropout_prob": 0.0,
19
+ "hidden_size": 1536,
20
+ "id2label": {
21
+ "0": "acadian_flycatcher",
22
+ "1": "american_crow",
23
+ "2": "american_goldfinch",
24
+ "3": "american_pipit",
25
+ "4": "american_redstart",
26
+ "5": "american_three_toed_woodpecker",
27
+ "6": "anna_hummingbird",
28
+ "7": "artic_tern",
29
+ "8": "baird_sparrow",
30
+ "9": "baltimore_oriole",
31
+ "10": "bank_swallow",
32
+ "11": "barn_swallow",
33
+ "12": "bay_breasted_warbler",
34
+ "13": "belted_kingfisher",
35
+ "14": "bewick_wren",
36
+ "15": "black_and_white_warbler",
37
+ "16": "black_billed_cuckoo",
38
+ "17": "black_capped_vireo",
39
+ "18": "black_footed_albatross",
40
+ "19": "black_tern",
41
+ "20": "black_throated_blue_warbler",
42
+ "21": "black_throated_sparrow",
43
+ "22": "blue_grosbeak",
44
+ "23": "blue_headed_vireo",
45
+ "24": "blue_jay",
46
+ "25": "blue_winged_warbler",
47
+ "26": "boat_tailed_grackle",
48
+ "27": "bobolink",
49
+ "28": "bohemian_waxwing",
50
+ "29": "brandt_cormorant",
51
+ "30": "brewer_blackbird",
52
+ "31": "brewer_sparrow",
53
+ "32": "bronzed_cowbird",
54
+ "33": "brown_creeper",
55
+ "34": "brown_pelican",
56
+ "35": "brown_thrasher",
57
+ "36": "cactus_wren",
58
+ "37": "california_gull",
59
+ "38": "canada_warbler",
60
+ "39": "cape_glossy_starling",
61
+ "40": "cape_may_warbler",
62
+ "41": "cardinal",
63
+ "42": "carolina_wren",
64
+ "43": "caspian_tern",
65
+ "44": "cedar_waxwing",
66
+ "45": "cerulean_warbler",
67
+ "46": "chestnut_sided_warbler",
68
+ "47": "chipping_sparrow",
69
+ "48": "chuck_will_widow",
70
+ "49": "clark_nutcracker",
71
+ "50": "clay_colored_sparrow",
72
+ "51": "cliff_swallow",
73
+ "52": "common_raven",
74
+ "53": "common_tern",
75
+ "54": "common_yellowthroat",
76
+ "55": "crested_auklet",
77
+ "56": "dark_eyed_junco",
78
+ "57": "downy_woodpecker",
79
+ "58": "eared_grebe",
80
+ "59": "eastern_towhee",
81
+ "60": "elegant_tern",
82
+ "61": "european_goldfinch",
83
+ "62": "evening_grosbeak",
84
+ "63": "field_sparrow",
85
+ "64": "fish_crow",
86
+ "65": "florida_jay",
87
+ "66": "forsters_tern",
88
+ "67": "fox_sparrow",
89
+ "68": "frigatebird",
90
+ "69": "gadwall",
91
+ "70": "geococcyx",
92
+ "71": "glaucous_winged_gull",
93
+ "72": "golden_winged_warbler",
94
+ "73": "grasshopper_sparrow",
95
+ "74": "gray_catbird",
96
+ "75": "gray_crowned_rosy_finch",
97
+ "76": "gray_kingbird",
98
+ "77": "great_crested_flycatcher",
99
+ "78": "great_grey_shrike",
100
+ "79": "green_jay",
101
+ "80": "green_kingfisher",
102
+ "81": "green_tailed_towhee",
103
+ "82": "green_violetear",
104
+ "83": "groove_billed_ani",
105
+ "84": "harris_sparrow",
106
+ "85": "heermann_gull",
107
+ "86": "henslow_sparrow",
108
+ "87": "herring_gull",
109
+ "88": "hooded_merganser",
110
+ "89": "hooded_oriole",
111
+ "90": "hooded_warbler",
112
+ "91": "horned_grebe",
113
+ "92": "horned_lark",
114
+ "93": "horned_puffin",
115
+ "94": "house_sparrow",
116
+ "95": "house_wren",
117
+ "96": "indigo_bunting",
118
+ "97": "ivory_gull",
119
+ "98": "kentucky_warbler",
120
+ "99": "laysan_albatross",
121
+ "100": "lazuli_bunting",
122
+ "101": "le_conte_sparrow",
123
+ "102": "least_auklet",
124
+ "103": "least_flycatcher",
125
+ "104": "least_tern",
126
+ "105": "lincoln_sparrow",
127
+ "106": "loggerhead_shrike",
128
+ "107": "long_tailed_jaeger",
129
+ "108": "louisiana_waterthrush",
130
+ "109": "magnolia_warbler",
131
+ "110": "mallard",
132
+ "111": "mangrove_cuckoo",
133
+ "112": "marsh_wren",
134
+ "113": "mockingbird",
135
+ "114": "mourning_warbler",
136
+ "115": "myrtle_warbler",
137
+ "116": "nashville_warbler",
138
+ "117": "nelson_sharp_tailed_sparrow",
139
+ "118": "nighthawk",
140
+ "119": "northern_flicker",
141
+ "120": "northern_fulmar",
142
+ "121": "northern_waterthrush",
143
+ "122": "olive_sided_flycatcher",
144
+ "123": "orange_crowned_warbler",
145
+ "124": "orchard_oriole",
146
+ "125": "ovenbird",
147
+ "126": "pacific_loon",
148
+ "127": "painted_bunting",
149
+ "128": "palm_warbler",
150
+ "129": "parakeet_auklet",
151
+ "130": "pelagic_cormorant",
152
+ "131": "philadelphia_vireo",
153
+ "132": "pied_billed_grebe",
154
+ "133": "pied_kingfisher",
155
+ "134": "pigeon_guillemot",
156
+ "135": "pileated_woodpecker",
157
+ "136": "pine_grosbeak",
158
+ "137": "pine_warbler",
159
+ "138": "pomarine_jaeger",
160
+ "139": "prairie_warbler",
161
+ "140": "prothonotary_warbler",
162
+ "141": "purple_finch",
163
+ "142": "red_bellied_woodpecker",
164
+ "143": "red_breasted_merganser",
165
+ "144": "red_cockaded_woodpecker",
166
+ "145": "red_eyed_vireo",
167
+ "146": "red_faced_cormorant",
168
+ "147": "red_headed_woodpecker",
169
+ "148": "red_legged_kittiwake",
170
+ "149": "red_winged_blackbird",
171
+ "150": "rhinoceros_auklet",
172
+ "151": "ring_billed_gull",
173
+ "152": "ringed_kingfisher",
174
+ "153": "rock_wren",
175
+ "154": "rose_breasted_grosbeak",
176
+ "155": "ruby_throated_hummingbird",
177
+ "156": "rufous_hummingbird",
178
+ "157": "rusty_blackbird",
179
+ "158": "sage_thrasher",
180
+ "159": "savannah_sparrow",
181
+ "160": "sayornis",
182
+ "161": "scarlet_tanager",
183
+ "162": "scissor_tailed_flycatcher",
184
+ "163": "scott_oriole",
185
+ "164": "seaside_sparrow",
186
+ "165": "shiny_cowbird",
187
+ "166": "slaty_backed_gull",
188
+ "167": "song_sparrow",
189
+ "168": "sooty_albatross",
190
+ "169": "spotted_catbird",
191
+ "170": "summer_tanager",
192
+ "171": "swainson_warbler",
193
+ "172": "tennessee_warbler",
194
+ "173": "tree_sparrow",
195
+ "174": "tree_swallow",
196
+ "175": "tropical_kingbird",
197
+ "176": "vermilion_flycatcher",
198
+ "177": "vesper_sparrow",
199
+ "178": "warbling_vireo",
200
+ "179": "western_grebe",
201
+ "180": "western_gull",
202
+ "181": "western_meadowlark",
203
+ "182": "western_wood_pewee",
204
+ "183": "whip_poor_will",
205
+ "184": "white_breasted_kingfisher",
206
+ "185": "white_breasted_nuthatch",
207
+ "186": "white_crowned_sparrow",
208
+ "187": "white_eyed_vireo",
209
+ "188": "white_necked_raven",
210
+ "189": "white_pelican",
211
+ "190": "white_throated_sparrow",
212
+ "191": "wilson_warbler",
213
+ "192": "winter_wren",
214
+ "193": "worm_eating_warbler",
215
+ "194": "yellow_bellied_flycatcher",
216
+ "195": "yellow_billed_cuckoo",
217
+ "196": "yellow_breasted_chat",
218
+ "197": "yellow_headed_blackbird",
219
+ "198": "yellow_throated_vireo",
220
+ "199": "yellow_warbler"
221
+ },
222
+ "image_size": 192,
223
+ "initializer_range": 0.02,
224
+ "label2id": {
225
+ "acadian_flycatcher": 0,
226
+ "american_crow": 1,
227
+ "american_goldfinch": 2,
228
+ "american_pipit": 3,
229
+ "american_redstart": 4,
230
+ "american_three_toed_woodpecker": 5,
231
+ "anna_hummingbird": 6,
232
+ "artic_tern": 7,
233
+ "baird_sparrow": 8,
234
+ "baltimore_oriole": 9,
235
+ "bank_swallow": 10,
236
+ "barn_swallow": 11,
237
+ "bay_breasted_warbler": 12,
238
+ "belted_kingfisher": 13,
239
+ "bewick_wren": 14,
240
+ "black_and_white_warbler": 15,
241
+ "black_billed_cuckoo": 16,
242
+ "black_capped_vireo": 17,
243
+ "black_footed_albatross": 18,
244
+ "black_tern": 19,
245
+ "black_throated_blue_warbler": 20,
246
+ "black_throated_sparrow": 21,
247
+ "blue_grosbeak": 22,
248
+ "blue_headed_vireo": 23,
249
+ "blue_jay": 24,
250
+ "blue_winged_warbler": 25,
251
+ "boat_tailed_grackle": 26,
252
+ "bobolink": 27,
253
+ "bohemian_waxwing": 28,
254
+ "brandt_cormorant": 29,
255
+ "brewer_blackbird": 30,
256
+ "brewer_sparrow": 31,
257
+ "bronzed_cowbird": 32,
258
+ "brown_creeper": 33,
259
+ "brown_pelican": 34,
260
+ "brown_thrasher": 35,
261
+ "cactus_wren": 36,
262
+ "california_gull": 37,
263
+ "canada_warbler": 38,
264
+ "cape_glossy_starling": 39,
265
+ "cape_may_warbler": 40,
266
+ "cardinal": 41,
267
+ "carolina_wren": 42,
268
+ "caspian_tern": 43,
269
+ "cedar_waxwing": 44,
270
+ "cerulean_warbler": 45,
271
+ "chestnut_sided_warbler": 46,
272
+ "chipping_sparrow": 47,
273
+ "chuck_will_widow": 48,
274
+ "clark_nutcracker": 49,
275
+ "clay_colored_sparrow": 50,
276
+ "cliff_swallow": 51,
277
+ "common_raven": 52,
278
+ "common_tern": 53,
279
+ "common_yellowthroat": 54,
280
+ "crested_auklet": 55,
281
+ "dark_eyed_junco": 56,
282
+ "downy_woodpecker": 57,
283
+ "eared_grebe": 58,
284
+ "eastern_towhee": 59,
285
+ "elegant_tern": 60,
286
+ "european_goldfinch": 61,
287
+ "evening_grosbeak": 62,
288
+ "field_sparrow": 63,
289
+ "fish_crow": 64,
290
+ "florida_jay": 65,
291
+ "forsters_tern": 66,
292
+ "fox_sparrow": 67,
293
+ "frigatebird": 68,
294
+ "gadwall": 69,
295
+ "geococcyx": 70,
296
+ "glaucous_winged_gull": 71,
297
+ "golden_winged_warbler": 72,
298
+ "grasshopper_sparrow": 73,
299
+ "gray_catbird": 74,
300
+ "gray_crowned_rosy_finch": 75,
301
+ "gray_kingbird": 76,
302
+ "great_crested_flycatcher": 77,
303
+ "great_grey_shrike": 78,
304
+ "green_jay": 79,
305
+ "green_kingfisher": 80,
306
+ "green_tailed_towhee": 81,
307
+ "green_violetear": 82,
308
+ "groove_billed_ani": 83,
309
+ "harris_sparrow": 84,
310
+ "heermann_gull": 85,
311
+ "henslow_sparrow": 86,
312
+ "herring_gull": 87,
313
+ "hooded_merganser": 88,
314
+ "hooded_oriole": 89,
315
+ "hooded_warbler": 90,
316
+ "horned_grebe": 91,
317
+ "horned_lark": 92,
318
+ "horned_puffin": 93,
319
+ "house_sparrow": 94,
320
+ "house_wren": 95,
321
+ "indigo_bunting": 96,
322
+ "ivory_gull": 97,
323
+ "kentucky_warbler": 98,
324
+ "laysan_albatross": 99,
325
+ "lazuli_bunting": 100,
326
+ "le_conte_sparrow": 101,
327
+ "least_auklet": 102,
328
+ "least_flycatcher": 103,
329
+ "least_tern": 104,
330
+ "lincoln_sparrow": 105,
331
+ "loggerhead_shrike": 106,
332
+ "long_tailed_jaeger": 107,
333
+ "louisiana_waterthrush": 108,
334
+ "magnolia_warbler": 109,
335
+ "mallard": 110,
336
+ "mangrove_cuckoo": 111,
337
+ "marsh_wren": 112,
338
+ "mockingbird": 113,
339
+ "mourning_warbler": 114,
340
+ "myrtle_warbler": 115,
341
+ "nashville_warbler": 116,
342
+ "nelson_sharp_tailed_sparrow": 117,
343
+ "nighthawk": 118,
344
+ "northern_flicker": 119,
345
+ "northern_fulmar": 120,
346
+ "northern_waterthrush": 121,
347
+ "olive_sided_flycatcher": 122,
348
+ "orange_crowned_warbler": 123,
349
+ "orchard_oriole": 124,
350
+ "ovenbird": 125,
351
+ "pacific_loon": 126,
352
+ "painted_bunting": 127,
353
+ "palm_warbler": 128,
354
+ "parakeet_auklet": 129,
355
+ "pelagic_cormorant": 130,
356
+ "philadelphia_vireo": 131,
357
+ "pied_billed_grebe": 132,
358
+ "pied_kingfisher": 133,
359
+ "pigeon_guillemot": 134,
360
+ "pileated_woodpecker": 135,
361
+ "pine_grosbeak": 136,
362
+ "pine_warbler": 137,
363
+ "pomarine_jaeger": 138,
364
+ "prairie_warbler": 139,
365
+ "prothonotary_warbler": 140,
366
+ "purple_finch": 141,
367
+ "red_bellied_woodpecker": 142,
368
+ "red_breasted_merganser": 143,
369
+ "red_cockaded_woodpecker": 144,
370
+ "red_eyed_vireo": 145,
371
+ "red_faced_cormorant": 146,
372
+ "red_headed_woodpecker": 147,
373
+ "red_legged_kittiwake": 148,
374
+ "red_winged_blackbird": 149,
375
+ "rhinoceros_auklet": 150,
376
+ "ring_billed_gull": 151,
377
+ "ringed_kingfisher": 152,
378
+ "rock_wren": 153,
379
+ "rose_breasted_grosbeak": 154,
380
+ "ruby_throated_hummingbird": 155,
381
+ "rufous_hummingbird": 156,
382
+ "rusty_blackbird": 157,
383
+ "sage_thrasher": 158,
384
+ "savannah_sparrow": 159,
385
+ "sayornis": 160,
386
+ "scarlet_tanager": 161,
387
+ "scissor_tailed_flycatcher": 162,
388
+ "scott_oriole": 163,
389
+ "seaside_sparrow": 164,
390
+ "shiny_cowbird": 165,
391
+ "slaty_backed_gull": 166,
392
+ "song_sparrow": 167,
393
+ "sooty_albatross": 168,
394
+ "spotted_catbird": 169,
395
+ "summer_tanager": 170,
396
+ "swainson_warbler": 171,
397
+ "tennessee_warbler": 172,
398
+ "tree_sparrow": 173,
399
+ "tree_swallow": 174,
400
+ "tropical_kingbird": 175,
401
+ "vermilion_flycatcher": 176,
402
+ "vesper_sparrow": 177,
403
+ "warbling_vireo": 178,
404
+ "western_grebe": 179,
405
+ "western_gull": 180,
406
+ "western_meadowlark": 181,
407
+ "western_wood_pewee": 182,
408
+ "whip_poor_will": 183,
409
+ "white_breasted_kingfisher": 184,
410
+ "white_breasted_nuthatch": 185,
411
+ "white_crowned_sparrow": 186,
412
+ "white_eyed_vireo": 187,
413
+ "white_necked_raven": 188,
414
+ "white_pelican": 189,
415
+ "white_throated_sparrow": 190,
416
+ "wilson_warbler": 191,
417
+ "winter_wren": 192,
418
+ "worm_eating_warbler": 193,
419
+ "yellow_bellied_flycatcher": 194,
420
+ "yellow_billed_cuckoo": 195,
421
+ "yellow_breasted_chat": 196,
422
+ "yellow_headed_blackbird": 197,
423
+ "yellow_throated_vireo": 198,
424
+ "yellow_warbler": 199
425
+ },
426
+ "layer_norm_eps": 1e-05,
427
+ "mlp_ratio": 4.0,
428
+ "model_type": "swinv2",
429
+ "num_channels": 3,
430
+ "num_heads": [
431
+ 6,
432
+ 12,
433
+ 24,
434
+ 48
435
+ ],
436
+ "num_layers": 4,
437
+ "out_features": [
438
+ "stage4"
439
+ ],
440
+ "out_indices": [
441
+ 4
442
+ ],
443
+ "patch_size": 4,
444
+ "path_norm": true,
445
+ "pretrained_window_sizes": [
446
+ 0,
447
+ 0,
448
+ 0,
449
+ 0
450
+ ],
451
+ "problem_type": "single_label_classification",
452
+ "qkv_bias": true,
453
+ "stage_names": [
454
+ "stem",
455
+ "stage1",
456
+ "stage2",
457
+ "stage3",
458
+ "stage4"
459
+ ],
460
+ "torch_dtype": "float32",
461
+ "transformers_version": "4.48.0",
462
+ "use_absolute_embeddings": false,
463
+ "window_size": 12
464
+ }
checkpoint-407/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27e08700996cd058870900682c6f868e69b0d8722470e1404fcb4421a181d039
3
+ size 14244
checkpoint-407/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8194bdd9d7fc1a7249d65a99fb906b2aa97bcc94d185c15e3e3624015920cfce
3
+ size 1064
checkpoint-407/trainer_state.json ADDED
@@ -0,0 +1,646 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.37125059962272644,
3
+ "best_model_checkpoint": "Model-SwinV2-Large-\\checkpoint-407",
4
+ "epoch": 10.980132450331126,
5
+ "eval_steps": 7,
6
+ "global_step": 407,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.18543046357615894,
13
+ "grad_norm": 7.775666236877441,
14
+ "learning_rate": 8.108108108108109e-07,
15
+ "loss": 5.4649,
16
+ "step": 7
17
+ },
18
+ {
19
+ "epoch": 0.3708609271523179,
20
+ "grad_norm": 9.706766128540039,
21
+ "learning_rate": 1.756756756756757e-06,
22
+ "loss": 5.4623,
23
+ "step": 14
24
+ },
25
+ {
26
+ "epoch": 0.5562913907284768,
27
+ "grad_norm": 7.272828578948975,
28
+ "learning_rate": 2.702702702702703e-06,
29
+ "loss": 5.4689,
30
+ "step": 21
31
+ },
32
+ {
33
+ "epoch": 0.7417218543046358,
34
+ "grad_norm": 9.961599349975586,
35
+ "learning_rate": 3.648648648648649e-06,
36
+ "loss": 5.3771,
37
+ "step": 28
38
+ },
39
+ {
40
+ "epoch": 0.9271523178807947,
41
+ "grad_norm": 9.451801300048828,
42
+ "learning_rate": 4.594594594594595e-06,
43
+ "loss": 5.3897,
44
+ "step": 35
45
+ },
46
+ {
47
+ "epoch": 0.9801324503311258,
48
+ "eval_accuracy": 0.009016393442622951,
49
+ "eval_f1_macro": 0.008969822352444303,
50
+ "eval_f1_micro": 0.009016393442622951,
51
+ "eval_f1_weighted": 0.008562213642681454,
52
+ "eval_loss": 5.330114841461182,
53
+ "eval_precision_macro": 0.012256033182503772,
54
+ "eval_precision_micro": 0.009016393442622951,
55
+ "eval_precision_weighted": 0.011259550478451154,
56
+ "eval_recall_macro": 0.009261904761904761,
57
+ "eval_recall_micro": 0.009016393442622951,
58
+ "eval_recall_weighted": 0.009016393442622951,
59
+ "eval_runtime": 84.595,
60
+ "eval_samples_per_second": 14.422,
61
+ "eval_steps_per_second": 0.236,
62
+ "step": 37
63
+ },
64
+ {
65
+ "epoch": 1.1324503311258278,
66
+ "grad_norm": 8.083910942077637,
67
+ "learning_rate": 5.5405405405405415e-06,
68
+ "loss": 5.5232,
69
+ "step": 42
70
+ },
71
+ {
72
+ "epoch": 1.3178807947019868,
73
+ "grad_norm": 16.997234344482422,
74
+ "learning_rate": 6.486486486486487e-06,
75
+ "loss": 5.3019,
76
+ "step": 49
77
+ },
78
+ {
79
+ "epoch": 1.5033112582781456,
80
+ "grad_norm": 8.079388618469238,
81
+ "learning_rate": 7.297297297297298e-06,
82
+ "loss": 5.2404,
83
+ "step": 56
84
+ },
85
+ {
86
+ "epoch": 1.6887417218543046,
87
+ "grad_norm": 7.1500725746154785,
88
+ "learning_rate": 8.243243243243245e-06,
89
+ "loss": 5.218,
90
+ "step": 63
91
+ },
92
+ {
93
+ "epoch": 1.8741721854304636,
94
+ "grad_norm": 6.12971830368042,
95
+ "learning_rate": 9.18918918918919e-06,
96
+ "loss": 5.1645,
97
+ "step": 70
98
+ },
99
+ {
100
+ "epoch": 1.980132450331126,
101
+ "eval_accuracy": 0.05245901639344262,
102
+ "eval_f1_macro": 0.03662301941302143,
103
+ "eval_f1_micro": 0.05245901639344262,
104
+ "eval_f1_weighted": 0.0388490193249229,
105
+ "eval_loss": 5.068942070007324,
106
+ "eval_precision_macro": 0.03409158543573505,
107
+ "eval_precision_micro": 0.05245901639344262,
108
+ "eval_precision_weighted": 0.035601775905091275,
109
+ "eval_recall_macro": 0.04867261904761905,
110
+ "eval_recall_micro": 0.05245901639344262,
111
+ "eval_recall_weighted": 0.05245901639344262,
112
+ "eval_runtime": 60.2034,
113
+ "eval_samples_per_second": 20.265,
114
+ "eval_steps_per_second": 0.332,
115
+ "step": 74
116
+ },
117
+ {
118
+ "epoch": 2.0794701986754967,
119
+ "grad_norm": 10.855502128601074,
120
+ "learning_rate": 1.0135135135135136e-05,
121
+ "loss": 5.2745,
122
+ "step": 77
123
+ },
124
+ {
125
+ "epoch": 2.2649006622516556,
126
+ "grad_norm": 8.226344108581543,
127
+ "learning_rate": 1.0945945945945948e-05,
128
+ "loss": 5.0241,
129
+ "step": 84
130
+ },
131
+ {
132
+ "epoch": 2.4503311258278146,
133
+ "grad_norm": 6.742478370666504,
134
+ "learning_rate": 1.1891891891891893e-05,
135
+ "loss": 4.9665,
136
+ "step": 91
137
+ },
138
+ {
139
+ "epoch": 2.6357615894039736,
140
+ "grad_norm": 8.76828384399414,
141
+ "learning_rate": 1.2837837837837838e-05,
142
+ "loss": 4.7954,
143
+ "step": 98
144
+ },
145
+ {
146
+ "epoch": 2.821192052980132,
147
+ "grad_norm": 16.411344528198242,
148
+ "learning_rate": 1.3783783783783786e-05,
149
+ "loss": 4.6041,
150
+ "step": 105
151
+ },
152
+ {
153
+ "epoch": 2.980132450331126,
154
+ "eval_accuracy": 0.26311475409836066,
155
+ "eval_f1_macro": 0.20009963945415252,
156
+ "eval_f1_micro": 0.26311475409836066,
157
+ "eval_f1_weighted": 0.21148759565581807,
158
+ "eval_loss": 3.9028923511505127,
159
+ "eval_precision_macro": 0.22816811623497163,
160
+ "eval_precision_micro": 0.26311475409836066,
161
+ "eval_precision_weighted": 0.23705714211336962,
162
+ "eval_recall_macro": 0.24582738095238096,
163
+ "eval_recall_micro": 0.26311475409836066,
164
+ "eval_recall_weighted": 0.26311475409836066,
165
+ "eval_runtime": 110.6815,
166
+ "eval_samples_per_second": 11.023,
167
+ "eval_steps_per_second": 0.181,
168
+ "step": 111
169
+ },
170
+ {
171
+ "epoch": 3.0264900662251657,
172
+ "grad_norm": 14.454651832580566,
173
+ "learning_rate": 1.472972972972973e-05,
174
+ "loss": 4.4593,
175
+ "step": 112
176
+ },
177
+ {
178
+ "epoch": 3.2119205298013247,
179
+ "grad_norm": 33.457191467285156,
180
+ "learning_rate": 1.5675675675675676e-05,
181
+ "loss": 3.7871,
182
+ "step": 119
183
+ },
184
+ {
185
+ "epoch": 3.3973509933774833,
186
+ "grad_norm": 20.33750343322754,
187
+ "learning_rate": 1.662162162162162e-05,
188
+ "loss": 3.3659,
189
+ "step": 126
190
+ },
191
+ {
192
+ "epoch": 3.5827814569536423,
193
+ "grad_norm": 17.52394676208496,
194
+ "learning_rate": 1.7432432432432434e-05,
195
+ "loss": 2.9398,
196
+ "step": 133
197
+ },
198
+ {
199
+ "epoch": 3.7682119205298013,
200
+ "grad_norm": 38.1245002746582,
201
+ "learning_rate": 1.837837837837838e-05,
202
+ "loss": 2.5717,
203
+ "step": 140
204
+ },
205
+ {
206
+ "epoch": 3.9536423841059603,
207
+ "grad_norm": 40.0474967956543,
208
+ "learning_rate": 1.9324324324324324e-05,
209
+ "loss": 2.0853,
210
+ "step": 147
211
+ },
212
+ {
213
+ "epoch": 3.980132450331126,
214
+ "eval_accuracy": 0.6860655737704918,
215
+ "eval_f1_macro": 0.6476607878873493,
216
+ "eval_f1_micro": 0.6860655737704918,
217
+ "eval_f1_weighted": 0.6600757648502281,
218
+ "eval_loss": 1.2792028188705444,
219
+ "eval_precision_macro": 0.6980460318405365,
220
+ "eval_precision_micro": 0.6860655737704918,
221
+ "eval_precision_weighted": 0.7044202635761663,
222
+ "eval_recall_macro": 0.6700654761904761,
223
+ "eval_recall_micro": 0.6860655737704918,
224
+ "eval_recall_weighted": 0.6860655737704918,
225
+ "eval_runtime": 47.8473,
226
+ "eval_samples_per_second": 25.498,
227
+ "eval_steps_per_second": 0.418,
228
+ "step": 148
229
+ },
230
+ {
231
+ "epoch": 4.158940397350993,
232
+ "grad_norm": 14.780089378356934,
233
+ "learning_rate": 2.0270270270270273e-05,
234
+ "loss": 1.7539,
235
+ "step": 154
236
+ },
237
+ {
238
+ "epoch": 4.344370860927152,
239
+ "grad_norm": 23.0028076171875,
240
+ "learning_rate": 2.1216216216216218e-05,
241
+ "loss": 1.4574,
242
+ "step": 161
243
+ },
244
+ {
245
+ "epoch": 4.529801324503311,
246
+ "grad_norm": 14.302562713623047,
247
+ "learning_rate": 2.2162162162162166e-05,
248
+ "loss": 1.3244,
249
+ "step": 168
250
+ },
251
+ {
252
+ "epoch": 4.71523178807947,
253
+ "grad_norm": 15.798663139343262,
254
+ "learning_rate": 2.3108108108108108e-05,
255
+ "loss": 1.2262,
256
+ "step": 175
257
+ },
258
+ {
259
+ "epoch": 4.900662251655629,
260
+ "grad_norm": 18.089494705200195,
261
+ "learning_rate": 2.4054054054054056e-05,
262
+ "loss": 1.2735,
263
+ "step": 182
264
+ },
265
+ {
266
+ "epoch": 4.9801324503311255,
267
+ "eval_accuracy": 0.8327868852459016,
268
+ "eval_f1_macro": 0.8214352639534529,
269
+ "eval_f1_micro": 0.8327868852459016,
270
+ "eval_f1_weighted": 0.8270994593768409,
271
+ "eval_loss": 0.6065749526023865,
272
+ "eval_precision_macro": 0.8606998556998556,
273
+ "eval_precision_micro": 0.8327868852459016,
274
+ "eval_precision_weighted": 0.8636802746433894,
275
+ "eval_recall_macro": 0.8253214285714285,
276
+ "eval_recall_micro": 0.8327868852459016,
277
+ "eval_recall_weighted": 0.8327868852459016,
278
+ "eval_runtime": 83.6381,
279
+ "eval_samples_per_second": 14.587,
280
+ "eval_steps_per_second": 0.239,
281
+ "step": 185
282
+ },
283
+ {
284
+ "epoch": 5.105960264900662,
285
+ "grad_norm": 16.889223098754883,
286
+ "learning_rate": 2.5e-05,
287
+ "loss": 1.0632,
288
+ "step": 189
289
+ },
290
+ {
291
+ "epoch": 5.291390728476821,
292
+ "grad_norm": 12.659591674804688,
293
+ "learning_rate": 2.594594594594595e-05,
294
+ "loss": 0.9143,
295
+ "step": 196
296
+ },
297
+ {
298
+ "epoch": 5.47682119205298,
299
+ "grad_norm": 10.869107246398926,
300
+ "learning_rate": 2.689189189189189e-05,
301
+ "loss": 0.9349,
302
+ "step": 203
303
+ },
304
+ {
305
+ "epoch": 5.662251655629139,
306
+ "grad_norm": 15.735921859741211,
307
+ "learning_rate": 2.7837837837837836e-05,
308
+ "loss": 0.8886,
309
+ "step": 210
310
+ },
311
+ {
312
+ "epoch": 5.847682119205298,
313
+ "grad_norm": 13.42103385925293,
314
+ "learning_rate": 2.8783783783783785e-05,
315
+ "loss": 0.8957,
316
+ "step": 217
317
+ },
318
+ {
319
+ "epoch": 5.9801324503311255,
320
+ "eval_accuracy": 0.8754098360655738,
321
+ "eval_f1_macro": 0.8704101537007883,
322
+ "eval_f1_micro": 0.8754098360655738,
323
+ "eval_f1_weighted": 0.8727622563744599,
324
+ "eval_loss": 0.4349415898323059,
325
+ "eval_precision_macro": 0.8963481005709266,
326
+ "eval_precision_micro": 0.8754098360655738,
327
+ "eval_precision_weighted": 0.8973616240811536,
328
+ "eval_recall_macro": 0.8722559523809523,
329
+ "eval_recall_micro": 0.8754098360655738,
330
+ "eval_recall_weighted": 0.8754098360655738,
331
+ "eval_runtime": 86.4338,
332
+ "eval_samples_per_second": 14.115,
333
+ "eval_steps_per_second": 0.231,
334
+ "step": 222
335
+ },
336
+ {
337
+ "epoch": 6.052980132450331,
338
+ "grad_norm": 13.915326118469238,
339
+ "learning_rate": 2.9729729729729733e-05,
340
+ "loss": 0.8343,
341
+ "step": 224
342
+ },
343
+ {
344
+ "epoch": 6.23841059602649,
345
+ "grad_norm": 11.141881942749023,
346
+ "learning_rate": 3.0675675675675675e-05,
347
+ "loss": 0.742,
348
+ "step": 231
349
+ },
350
+ {
351
+ "epoch": 6.423841059602649,
352
+ "grad_norm": 13.977461814880371,
353
+ "learning_rate": 3.162162162162162e-05,
354
+ "loss": 0.7757,
355
+ "step": 238
356
+ },
357
+ {
358
+ "epoch": 6.609271523178808,
359
+ "grad_norm": 12.792037010192871,
360
+ "learning_rate": 3.256756756756757e-05,
361
+ "loss": 0.7528,
362
+ "step": 245
363
+ },
364
+ {
365
+ "epoch": 6.7947019867549665,
366
+ "grad_norm": 9.550854682922363,
367
+ "learning_rate": 3.351351351351352e-05,
368
+ "loss": 0.7589,
369
+ "step": 252
370
+ },
371
+ {
372
+ "epoch": 6.9801324503311255,
373
+ "grad_norm": 56.45753860473633,
374
+ "learning_rate": 3.445945945945946e-05,
375
+ "loss": 0.7522,
376
+ "step": 259
377
+ },
378
+ {
379
+ "epoch": 6.9801324503311255,
380
+ "eval_accuracy": 0.8885245901639345,
381
+ "eval_f1_macro": 0.8810641592669457,
382
+ "eval_f1_micro": 0.8885245901639345,
383
+ "eval_f1_weighted": 0.8837497312389461,
384
+ "eval_loss": 0.41575273871421814,
385
+ "eval_precision_macro": 0.9128655927405928,
386
+ "eval_precision_micro": 0.8885245901639345,
387
+ "eval_precision_weighted": 0.9129847338453896,
388
+ "eval_recall_macro": 0.8841249999999999,
389
+ "eval_recall_micro": 0.8885245901639345,
390
+ "eval_recall_weighted": 0.8885245901639345,
391
+ "eval_runtime": 87.2308,
392
+ "eval_samples_per_second": 13.986,
393
+ "eval_steps_per_second": 0.229,
394
+ "step": 259
395
+ },
396
+ {
397
+ "epoch": 7.185430463576159,
398
+ "grad_norm": 9.251741409301758,
399
+ "learning_rate": 3.5405405405405403e-05,
400
+ "loss": 0.71,
401
+ "step": 266
402
+ },
403
+ {
404
+ "epoch": 7.370860927152318,
405
+ "grad_norm": 12.447789192199707,
406
+ "learning_rate": 3.635135135135135e-05,
407
+ "loss": 0.6271,
408
+ "step": 273
409
+ },
410
+ {
411
+ "epoch": 7.556291390728477,
412
+ "grad_norm": 15.672873497009277,
413
+ "learning_rate": 3.7162162162162165e-05,
414
+ "loss": 0.704,
415
+ "step": 280
416
+ },
417
+ {
418
+ "epoch": 7.741721854304636,
419
+ "grad_norm": 21.882080078125,
420
+ "learning_rate": 3.8108108108108106e-05,
421
+ "loss": 0.6397,
422
+ "step": 287
423
+ },
424
+ {
425
+ "epoch": 7.927152317880795,
426
+ "grad_norm": 12.572905540466309,
427
+ "learning_rate": 3.9054054054054055e-05,
428
+ "loss": 0.6157,
429
+ "step": 294
430
+ },
431
+ {
432
+ "epoch": 7.9801324503311255,
433
+ "eval_accuracy": 0.8860655737704918,
434
+ "eval_f1_macro": 0.8812874675633964,
435
+ "eval_f1_micro": 0.8860655737704918,
436
+ "eval_f1_weighted": 0.883605155706993,
437
+ "eval_loss": 0.37819206714630127,
438
+ "eval_precision_macro": 0.9077945526695527,
439
+ "eval_precision_micro": 0.8860655737704918,
440
+ "eval_precision_weighted": 0.9074083043550256,
441
+ "eval_recall_macro": 0.8824166666666667,
442
+ "eval_recall_micro": 0.8860655737704918,
443
+ "eval_recall_weighted": 0.8860655737704918,
444
+ "eval_runtime": 86.9903,
445
+ "eval_samples_per_second": 14.025,
446
+ "eval_steps_per_second": 0.23,
447
+ "step": 296
448
+ },
449
+ {
450
+ "epoch": 8.132450331125828,
451
+ "grad_norm": 9.28415298461914,
452
+ "learning_rate": 4e-05,
453
+ "loss": 0.632,
454
+ "step": 301
455
+ },
456
+ {
457
+ "epoch": 8.317880794701987,
458
+ "grad_norm": 9.938474655151367,
459
+ "learning_rate": 4.0945945945945945e-05,
460
+ "loss": 0.5854,
461
+ "step": 308
462
+ },
463
+ {
464
+ "epoch": 8.503311258278146,
465
+ "grad_norm": 9.133801460266113,
466
+ "learning_rate": 4.189189189189189e-05,
467
+ "loss": 0.4822,
468
+ "step": 315
469
+ },
470
+ {
471
+ "epoch": 8.688741721854305,
472
+ "grad_norm": 9.311871528625488,
473
+ "learning_rate": 4.283783783783784e-05,
474
+ "loss": 0.5033,
475
+ "step": 322
476
+ },
477
+ {
478
+ "epoch": 8.874172185430464,
479
+ "grad_norm": 8.707598686218262,
480
+ "learning_rate": 4.378378378378379e-05,
481
+ "loss": 0.6243,
482
+ "step": 329
483
+ },
484
+ {
485
+ "epoch": 8.980132450331126,
486
+ "eval_accuracy": 0.8795081967213115,
487
+ "eval_f1_macro": 0.8727452907859256,
488
+ "eval_f1_micro": 0.8795081967213115,
489
+ "eval_f1_weighted": 0.8751382461299096,
490
+ "eval_loss": 0.4089548587799072,
491
+ "eval_precision_macro": 0.8947101648351647,
492
+ "eval_precision_micro": 0.8795081967213115,
493
+ "eval_precision_weighted": 0.8962854563469319,
494
+ "eval_recall_macro": 0.8770833333333334,
495
+ "eval_recall_micro": 0.8795081967213115,
496
+ "eval_recall_weighted": 0.8795081967213115,
497
+ "eval_runtime": 85.9321,
498
+ "eval_samples_per_second": 14.197,
499
+ "eval_steps_per_second": 0.233,
500
+ "step": 333
501
+ },
502
+ {
503
+ "epoch": 9.079470198675496,
504
+ "grad_norm": 10.794309616088867,
505
+ "learning_rate": 4.472972972972973e-05,
506
+ "loss": 0.5572,
507
+ "step": 336
508
+ },
509
+ {
510
+ "epoch": 9.264900662251655,
511
+ "grad_norm": 10.890398979187012,
512
+ "learning_rate": 4.567567567567568e-05,
513
+ "loss": 0.5487,
514
+ "step": 343
515
+ },
516
+ {
517
+ "epoch": 9.450331125827814,
518
+ "grad_norm": 10.409049987792969,
519
+ "learning_rate": 4.662162162162162e-05,
520
+ "loss": 0.6332,
521
+ "step": 350
522
+ },
523
+ {
524
+ "epoch": 9.635761589403973,
525
+ "grad_norm": 9.35953426361084,
526
+ "learning_rate": 4.756756756756757e-05,
527
+ "loss": 0.5324,
528
+ "step": 357
529
+ },
530
+ {
531
+ "epoch": 9.821192052980132,
532
+ "grad_norm": 6.873753547668457,
533
+ "learning_rate": 4.851351351351351e-05,
534
+ "loss": 0.539,
535
+ "step": 364
536
+ },
537
+ {
538
+ "epoch": 9.980132450331126,
539
+ "eval_accuracy": 0.8811475409836066,
540
+ "eval_f1_macro": 0.8763333539488136,
541
+ "eval_f1_micro": 0.8811475409836066,
542
+ "eval_f1_weighted": 0.878023237317979,
543
+ "eval_loss": 0.39534232020378113,
544
+ "eval_precision_macro": 0.9026571067821066,
545
+ "eval_precision_micro": 0.8811475409836066,
546
+ "eval_precision_weighted": 0.9016379544863151,
547
+ "eval_recall_macro": 0.8773511904761905,
548
+ "eval_recall_micro": 0.8811475409836066,
549
+ "eval_recall_weighted": 0.8811475409836066,
550
+ "eval_runtime": 87.5111,
551
+ "eval_samples_per_second": 13.941,
552
+ "eval_steps_per_second": 0.229,
553
+ "step": 370
554
+ },
555
+ {
556
+ "epoch": 10.026490066225165,
557
+ "grad_norm": 13.179052352905273,
558
+ "learning_rate": 4.945945945945946e-05,
559
+ "loss": 0.5554,
560
+ "step": 371
561
+ },
562
+ {
563
+ "epoch": 10.211920529801324,
564
+ "grad_norm": 7.148564338684082,
565
+ "learning_rate": 4.995495495495496e-05,
566
+ "loss": 0.4887,
567
+ "step": 378
568
+ },
569
+ {
570
+ "epoch": 10.397350993377483,
571
+ "grad_norm": 9.368721961975098,
572
+ "learning_rate": 4.984984984984985e-05,
573
+ "loss": 0.5459,
574
+ "step": 385
575
+ },
576
+ {
577
+ "epoch": 10.582781456953642,
578
+ "grad_norm": 10.710000038146973,
579
+ "learning_rate": 4.9744744744744746e-05,
580
+ "loss": 0.5324,
581
+ "step": 392
582
+ },
583
+ {
584
+ "epoch": 10.7682119205298,
585
+ "grad_norm": 11.22227668762207,
586
+ "learning_rate": 4.963963963963964e-05,
587
+ "loss": 0.5425,
588
+ "step": 399
589
+ },
590
+ {
591
+ "epoch": 10.95364238410596,
592
+ "grad_norm": 8.322479248046875,
593
+ "learning_rate": 4.953453453453454e-05,
594
+ "loss": 0.5839,
595
+ "step": 406
596
+ },
597
+ {
598
+ "epoch": 10.980132450331126,
599
+ "eval_accuracy": 0.8918032786885246,
600
+ "eval_f1_macro": 0.8875326968782852,
601
+ "eval_f1_micro": 0.8918032786885246,
602
+ "eval_f1_weighted": 0.889380159243949,
603
+ "eval_loss": 0.37125059962272644,
604
+ "eval_precision_macro": 0.9082038517038518,
605
+ "eval_precision_micro": 0.8918032786885246,
606
+ "eval_precision_weighted": 0.9081954816585965,
607
+ "eval_recall_macro": 0.8885952380952381,
608
+ "eval_recall_micro": 0.8918032786885246,
609
+ "eval_recall_weighted": 0.8918032786885246,
610
+ "eval_runtime": 130.2164,
611
+ "eval_samples_per_second": 9.369,
612
+ "eval_steps_per_second": 0.154,
613
+ "step": 407
614
+ }
615
+ ],
616
+ "logging_steps": 7,
617
+ "max_steps": 3700,
618
+ "num_input_tokens_seen": 0,
619
+ "num_train_epochs": 100,
620
+ "save_steps": 7,
621
+ "stateful_callbacks": {
622
+ "EarlyStoppingCallback": {
623
+ "args": {
624
+ "early_stopping_patience": 5,
625
+ "early_stopping_threshold": 0.01
626
+ },
627
+ "attributes": {
628
+ "early_stopping_patience_counter": 3
629
+ }
630
+ },
631
+ "TrainerControl": {
632
+ "args": {
633
+ "should_epoch_stop": false,
634
+ "should_evaluate": false,
635
+ "should_log": false,
636
+ "should_save": true,
637
+ "should_training_stop": false
638
+ },
639
+ "attributes": {}
640
+ }
641
+ },
642
+ "total_flos": 6.804145744376758e+18,
643
+ "train_batch_size": 32,
644
+ "trial_name": null,
645
+ "trial_params": null
646
+ }
checkpoint-407/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bbc9d5bfc360c6b73e096edb34a7e0db5f03fc4b70bd6ec3ae2f6bdf95a8727
3
+ size 5304