oliviermills commited on
Commit
4179360
·
verified ·
1 Parent(s): bd591b0

Training log

Browse files
gliner2.5-small-passage/training_log.json ADDED
@@ -0,0 +1,260 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "method": "gliner-finetune",
3
+ "args": {
4
+ "model": "fastino/gliner2.5-small-v1",
5
+ "output_repo": "baobabtech/evalexplorer-classify-adapters/gliner2.5-small-passage",
6
+ "dataset": "baobabtech/evalexplorer-data",
7
+ "config": "classify_codes",
8
+ "input": "passage",
9
+ "max_words": 1000,
10
+ "title_words": 120,
11
+ "chunk_size": 384,
12
+ "chunk_overlap": 64,
13
+ "epochs": 5,
14
+ "max_steps": -1,
15
+ "limit_train": null,
16
+ "batch_size": 16,
17
+ "encoder_lr": 1e-05,
18
+ "task_lr": 0.0005,
19
+ "seed": 42,
20
+ "eval_split": "test",
21
+ "eval_limit": null,
22
+ "eval_batch_size": 16
23
+ },
24
+ "output_repo": "baobabtech/evalexplorer-classify-adapters/gliner2.5-small-passage",
25
+ "train_rows": 1148,
26
+ "accelerator": "1\u00d7 NVIDIA A100-SXM4-80GB (79 GB)",
27
+ "job_id": "6ac019bb404719ba37627115",
28
+ "metrics": {
29
+ "train_runtime": 184.5655586719513,
30
+ "best_eval_loss": 132.10815599229602
31
+ },
32
+ "log_history": [
33
+ {
34
+ "step": 10,
35
+ "epoch": 0.1267605633802817,
36
+ "loss": 326.4710388183594
37
+ },
38
+ {
39
+ "step": 20,
40
+ "epoch": 0.2676056338028169,
41
+ "loss": 304.33172607421875
42
+ },
43
+ {
44
+ "step": 30,
45
+ "epoch": 0.4084507042253521,
46
+ "loss": 217.84324645996094
47
+ },
48
+ {
49
+ "step": 40,
50
+ "epoch": 0.5492957746478874,
51
+ "loss": 150.05361938476562
52
+ },
53
+ {
54
+ "step": 50,
55
+ "epoch": 0.6901408450704225,
56
+ "loss": 167.84349060058594
57
+ },
58
+ {
59
+ "step": 60,
60
+ "epoch": 0.8309859154929577,
61
+ "loss": 201.2783660888672
62
+ },
63
+ {
64
+ "step": 70,
65
+ "epoch": 0.971830985915493,
66
+ "loss": 134.7239990234375
67
+ },
68
+ {
69
+ "step": 71,
70
+ "epoch": 0,
71
+ "eval_loss": 197.0072953965929
72
+ },
73
+ {
74
+ "step": 80,
75
+ "epoch": 1.1126760563380282,
76
+ "loss": 146.74496459960938
77
+ },
78
+ {
79
+ "step": 90,
80
+ "epoch": 1.2535211267605635,
81
+ "loss": 152.97711181640625
82
+ },
83
+ {
84
+ "step": 100,
85
+ "epoch": 1.3943661971830985,
86
+ "loss": 120.85997772216797
87
+ },
88
+ {
89
+ "step": 110,
90
+ "epoch": 1.5352112676056338,
91
+ "loss": 129.96192932128906
92
+ },
93
+ {
94
+ "step": 120,
95
+ "epoch": 1.676056338028169,
96
+ "loss": 118.16130828857422
97
+ },
98
+ {
99
+ "step": 130,
100
+ "epoch": 1.8169014084507042,
101
+ "loss": 130.75257873535156
102
+ },
103
+ {
104
+ "step": 140,
105
+ "epoch": 1.9577464788732395,
106
+ "loss": 132.6935577392578
107
+ },
108
+ {
109
+ "step": 142,
110
+ "epoch": 1,
111
+ "eval_loss": 132.10815599229602
112
+ },
113
+ {
114
+ "step": 150,
115
+ "epoch": 2.0985915492957745,
116
+ "loss": 127.60694885253906
117
+ },
118
+ {
119
+ "step": 160,
120
+ "epoch": 2.23943661971831,
121
+ "loss": 109.15935516357422
122
+ },
123
+ {
124
+ "step": 170,
125
+ "epoch": 2.380281690140845,
126
+ "loss": 110.58048248291016
127
+ },
128
+ {
129
+ "step": 180,
130
+ "epoch": 2.52112676056338,
131
+ "loss": 97.95825958251953
132
+ },
133
+ {
134
+ "step": 190,
135
+ "epoch": 2.6619718309859155,
136
+ "loss": 106.41092681884766
137
+ },
138
+ {
139
+ "step": 200,
140
+ "epoch": 2.802816901408451,
141
+ "loss": 101.91635131835938
142
+ },
143
+ {
144
+ "step": 210,
145
+ "epoch": 2.943661971830986,
146
+ "loss": 111.9019775390625
147
+ },
148
+ {
149
+ "step": 213,
150
+ "epoch": 2,
151
+ "eval_loss": 137.78013186984592
152
+ },
153
+ {
154
+ "step": 220,
155
+ "epoch": 3.084507042253521,
156
+ "loss": 106.5567855834961
157
+ },
158
+ {
159
+ "step": 230,
160
+ "epoch": 3.2253521126760565,
161
+ "loss": 102.8129653930664
162
+ },
163
+ {
164
+ "step": 240,
165
+ "epoch": 3.3661971830985915,
166
+ "loss": 125.02168273925781
167
+ },
168
+ {
169
+ "step": 250,
170
+ "epoch": 3.507042253521127,
171
+ "loss": 91.05559539794922
172
+ },
173
+ {
174
+ "step": 260,
175
+ "epoch": 3.647887323943662,
176
+ "loss": 98.93767547607422
177
+ },
178
+ {
179
+ "step": 270,
180
+ "epoch": 3.788732394366197,
181
+ "loss": 93.60192108154297
182
+ },
183
+ {
184
+ "step": 280,
185
+ "epoch": 3.9295774647887325,
186
+ "loss": 110.9609146118164
187
+ },
188
+ {
189
+ "step": 284,
190
+ "epoch": 3,
191
+ "eval_loss": 148.7078013949924
192
+ },
193
+ {
194
+ "step": 290,
195
+ "epoch": 4.070422535211268,
196
+ "loss": 89.94481658935547
197
+ },
198
+ {
199
+ "step": 300,
200
+ "epoch": 4.211267605633803,
201
+ "loss": 85.77145385742188
202
+ },
203
+ {
204
+ "step": 310,
205
+ "epoch": 4.352112676056338,
206
+ "loss": 98.24652099609375
207
+ },
208
+ {
209
+ "step": 320,
210
+ "epoch": 4.492957746478873,
211
+ "loss": 87.61077880859375
212
+ },
213
+ {
214
+ "step": 330,
215
+ "epoch": 4.633802816901408,
216
+ "loss": 93.1797103881836
217
+ },
218
+ {
219
+ "step": 340,
220
+ "epoch": 4.774647887323944,
221
+ "loss": 92.88860321044922
222
+ },
223
+ {
224
+ "step": 350,
225
+ "epoch": 4.915492957746479,
226
+ "loss": 102.67058563232422
227
+ },
228
+ {
229
+ "step": 355,
230
+ "epoch": 4,
231
+ "eval_loss": 143.14172829522028
232
+ }
233
+ ],
234
+ "details": [
235
+ [
236
+ "Encoder / task learning rate",
237
+ "1e-05 / 0.0005"
238
+ ],
239
+ [
240
+ "Epochs, batch",
241
+ "5 epochs (max steps -1), batch 16"
242
+ ],
243
+ [
244
+ "Input",
245
+ "one passage per document: title page + own abstract or executive summary"
246
+ ],
247
+ [
248
+ "Training texts",
249
+ "1148 (863 with country spans)"
250
+ ],
251
+ [
252
+ "Best validation loss",
253
+ 132.10815599229602
254
+ ],
255
+ [
256
+ "Fine-tuning",
257
+ "full model (encoder + task heads)"
258
+ ]
259
+ ]
260
+ }