andrew-healey commited on
Commit
19dcd20
·
verified ·
1 Parent(s): a6b44d6

Upload folder using huggingface_hub

Browse files
12_mini_head_two_masks_4_heads/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_mini_head_two_masks_4_heads", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 12, "n_embd": 264, "head_dim": 22, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "repro_selective_pattern_rankings", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "two_masks", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 64, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.003, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_mini_head_two_masks_4_heads"}
12_mini_head_two_masks_4_heads/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ec6682852ed60a9d42cb4047300f65a04f87328c0ad7a4516be84d11b28f216
3
+ size 964
12_mini_head_two_masks_4_heads/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.2420
3
+ 0 val perplexity 76263.9609
4
+ 0 train 11.240022 (lr=4.1958e-06) (hash(x)=93356070)
5
+ 10 train 9.940840 (lr=4.6154e-05) (hash(x)=91253010)
6
+ 20 train 9.353498 (lr=8.8112e-05) (hash(x)=74934453)
7
+ 30 train 8.665655 (lr=1.3007e-04) (hash(x)=79278034)
8
+ 40 train 7.941760 (lr=1.7203e-04) (hash(x)=80107892)
9
+ 50 train 7.711949 (lr=2.1399e-04) (hash(x)=70745428)
10
+ 60 train 7.755047 (lr=2.5594e-04) (hash(x)=80077589)
11
+ 70 train 7.441437 (lr=2.9790e-04) (hash(x)=76213766)
12
+ 80 train 7.418306 (lr=3.3986e-04) (hash(x)=83218328)
13
+ 90 train 7.206155 (lr=3.8182e-04) (hash(x)=74855845)
14
+ 100 val loss 7.1167
15
+ 100 val perplexity 1232.3829
16
+ 100 train 6.949111 (lr=4.2378e-04) (hash(x)=82814902)
17
+ 110 train 6.907312 (lr=4.6573e-04) (hash(x)=84907741)
18
+ 120 train 6.730330 (lr=5.0769e-04) (hash(x)=82613223)
19
+ 130 train 6.576161 (lr=5.4965e-04) (hash(x)=83540876)
20
+ 140 train 6.539323 (lr=5.9161e-04) (hash(x)=75095216)
21
+ 150 train 6.610376 (lr=6.3357e-04) (hash(x)=97190944)
22
+ 160 train 6.520771 (lr=6.7552e-04) (hash(x)=82117809)
23
+ 170 train 6.438437 (lr=7.1748e-04) (hash(x)=70514724)
24
+ 180 train 6.355740 (lr=7.5944e-04) (hash(x)=81029624)
25
+ 190 train 6.264931 (lr=8.0140e-04) (hash(x)=78996841)
26
+ 200 val loss 6.2150
27
+ 200 val perplexity 500.1796
28
+ 200 train 6.217852 (lr=8.4336e-04) (hash(x)=79845097)
29
+ 210 train 6.087740 (lr=8.8531e-04) (hash(x)=78997683)
30
+ 220 train 6.027409 (lr=9.2727e-04) (hash(x)=74895865)
31
+ 230 train 6.153712 (lr=9.6923e-04) (hash(x)=80933276)
32
+ 240 train 5.840209 (lr=1.0112e-03) (hash(x)=77664606)
33
+ 250 train 5.906199 (lr=1.0531e-03) (hash(x)=75903930)
34
+ 260 train 5.783606 (lr=1.0951e-03) (hash(x)=82996853)
35
+ 270 train 5.730420 (lr=1.1371e-03) (hash(x)=73269568)
36
+ 280 train 5.650702 (lr=1.1790e-03) (hash(x)=82906427)
37
+ 290 train 5.577339 (lr=1.2210e-03) (hash(x)=71797895)
38
+ 300 val loss 5.7350
39
+ 300 val perplexity 309.5132
40
+ 300 train 5.439972 (lr=1.2629e-03) (hash(x)=85232249)
41
+ 310 train 5.385273 (lr=1.3049e-03) (hash(x)=75252489)
42
+ 320 train 5.472575 (lr=1.3469e-03) (hash(x)=74394644)
43
+ 330 train 5.584213 (lr=1.3888e-03) (hash(x)=82772910)
44
+ 340 train 5.627512 (lr=1.4308e-03) (hash(x)=81627464)
45
+ 350 train 5.614381 (lr=1.4727e-03) (hash(x)=84678053)
46
+ 360 train 5.534771 (lr=1.5147e-03) (hash(x)=81884128)
47
+ 370 train 5.558956 (lr=1.5566e-03) (hash(x)=76379242)
48
+ 380 train 5.365591 (lr=1.5986e-03) (hash(x)=98182875)
49
+ 390 train 5.331548 (lr=1.6406e-03) (hash(x)=79710436)
50
+ 400 val loss 5.3888
51
+ 400 val perplexity 218.9469
52
+ 400 train 5.323623 (lr=1.6825e-03) (hash(x)=79841071)
53
+ 410 train 5.288784 (lr=1.7245e-03) (hash(x)=75844151)
54
+ 420 train 5.277255 (lr=1.7664e-03) (hash(x)=73125036)
55
+ 430 train 5.323563 (lr=1.8084e-03) (hash(x)=84214858)
56
+ 440 train 5.205223 (lr=1.8503e-03) (hash(x)=80456994)
57
+ 450 train 5.201133 (lr=1.8923e-03) (hash(x)=76735962)
58
+ 460 train 5.100286 (lr=1.9343e-03) (hash(x)=81845446)
59
+ 470 train 4.915474 (lr=1.9762e-03) (hash(x)=76094689)
60
+ 480 train 5.037648 (lr=2.0182e-03) (hash(x)=83806686)
61
+ 490 train 4.848330 (lr=2.0601e-03) (hash(x)=84690227)
62
+ 500 val loss 5.2096
63
+ 500 val perplexity 183.0159
64
+ 500 train 4.937279 (lr=2.1021e-03) (hash(x)=71851938)
65
+ 510 train 5.087677 (lr=2.1441e-03) (hash(x)=77159346)
66
+ 520 train 5.113108 (lr=2.1860e-03) (hash(x)=80755753)
67
+ 530 train 5.225525 (lr=2.2280e-03) (hash(x)=75379679)
68
+ 540 train 5.150904 (lr=2.2699e-03) (hash(x)=82458619)
69
+ 550 train 5.087330 (lr=2.3119e-03) (hash(x)=86366396)
70
+ 560 train 5.119373 (lr=2.3538e-03) (hash(x)=85095044)
71
+ 570 train 5.070576 (lr=2.3958e-03) (hash(x)=78385159)
72
+ 580 train 5.014266 (lr=2.4378e-03) (hash(x)=79342394)
73
+ 590 train 5.011674 (lr=2.4797e-03) (hash(x)=70782192)
74
+ 600 val loss 5.0329
75
+ 600 val perplexity 153.3793
76
+ 600 train 5.002976 (lr=2.5217e-03) (hash(x)=76130353)
77
+ 610 train 4.940732 (lr=2.5636e-03) (hash(x)=74778440)
78
+ 620 train 4.889901 (lr=2.6056e-03) (hash(x)=79129709)
79
+ 630 train 4.792601 (lr=2.6476e-03) (hash(x)=76469962)
80
+ 640 train 4.781839 (lr=2.6895e-03) (hash(x)=78288049)
81
+ 650 train 4.735378 (lr=2.7315e-03) (hash(x)=76641388)
82
+ 660 train 4.667476 (lr=2.7734e-03) (hash(x)=75937906)
83
+ 670 train 4.647557 (lr=2.8154e-03) (hash(x)=74807157)
84
+ 680 train 4.541842 (lr=2.8573e-03) (hash(x)=77490144)
85
+ 690 train 4.593424 (lr=2.8993e-03) (hash(x)=72900124)
86
+ 700 val loss 4.8597
87
+ 700 val perplexity 128.9861
88
+ 700 train 4.543827 (lr=2.9413e-03) (hash(x)=73424218)
89
+ 710 train 4.841260 (lr=2.9832e-03) (hash(x)=86052345)
90
+ 720 train 4.819865 (lr=3.0000e-03) (hash(x)=77462613)
91
+ 730 train 4.821233 (lr=2.9999e-03) (hash(x)=79027471)
92
+ 740 train 4.781591 (lr=2.9997e-03) (hash(x)=78149992)
93
+ 750 train 4.728841 (lr=2.9994e-03) (hash(x)=89147499)
94
+ 760 train 4.761951 (lr=2.9990e-03) (hash(x)=74931538)
95
+ 770 train 4.708468 (lr=2.9985e-03) (hash(x)=89648204)
96
+ 780 train 4.769874 (lr=2.9979e-03) (hash(x)=83708293)
97
+ 790 train 4.751948 (lr=2.9972e-03) (hash(x)=84336512)
98
+ 800 val loss 4.7356
99
+ 800 val perplexity 113.9357
100
+ 800 train 4.656265 (lr=2.9964e-03) (hash(x)=75025285)
101
+ 810 train 4.571815 (lr=2.9955e-03) (hash(x)=79436536)
102
+ 820 train 4.576073 (lr=2.9945e-03) (hash(x)=90467390)
103
+ 830 train 4.613216 (lr=2.9934e-03) (hash(x)=77292020)
104
+ 840 train 4.539795 (lr=2.9922e-03) (hash(x)=75568927)
105
+ 850 train 4.574992 (lr=2.9909e-03) (hash(x)=79671781)
106
+ 860 train 4.661144 (lr=2.9896e-03) (hash(x)=84280943)
107
+ 870 train 4.644295 (lr=2.9881e-03) (hash(x)=80901294)
108
+ 880 train 4.755160 (lr=2.9865e-03) (hash(x)=81437584)
109
+ 890 train 4.612962 (lr=2.9848e-03) (hash(x)=73893778)
110
+ 900 val loss 4.6341
111
+ 900 val perplexity 102.9395
112
+ 900 train 4.570028 (lr=2.9830e-03) (hash(x)=74987794)
113
+ 910 train 4.584233 (lr=2.9811e-03) (hash(x)=93721374)
114
+ 920 train 4.569137 (lr=2.9792e-03) (hash(x)=79149678)
115
+ 930 train 4.525364 (lr=2.9771e-03) (hash(x)=83179387)
116
+ 940 train 4.532477 (lr=2.9749e-03) (hash(x)=89009978)
117
+ 950 train 4.529596 (lr=2.9726e-03) (hash(x)=76483735)
118
+ 960 train 4.451268 (lr=2.9703e-03) (hash(x)=78638850)
119
+ 970 train 4.492554 (lr=2.9678e-03) (hash(x)=90415206)
120
+ 980 train 4.428101 (lr=2.9652e-03) (hash(x)=90819195)
121
+ 990 train 4.553214 (lr=2.9626e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.5453
123
+ 1000 val perplexity 94.1918
124
+ 1000 train 4.590935 (lr=2.9598e-03) (hash(x)=82436789)
125
+ 1010 train 4.657196 (lr=2.9570e-03) (hash(x)=63022149)
126
+ 1020 train 4.560525 (lr=2.9540e-03) (hash(x)=83501199)
127
+ 1030 train 4.538416 (lr=2.9510e-03) (hash(x)=77093326)
128
+ 1040 train 4.420394 (lr=2.9478e-03) (hash(x)=73437559)
129
+ 1050 train 4.505728 (lr=2.9446e-03) (hash(x)=84550388)
130
+ 1060 train 4.508118 (lr=2.9412e-03) (hash(x)=86184566)
131
+ 1070 train 4.501207 (lr=2.9378e-03) (hash(x)=85644922)
132
+ 1080 train 4.435485 (lr=2.9343e-03) (hash(x)=86179801)
133
+ 1090 train 4.411906 (lr=2.9307e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.4789
135
+ 1100 val perplexity 88.1387
136
+ 1100 train 4.498798 (lr=2.9270e-03) (hash(x)=96780388)
137
+ 1110 train 4.446856 (lr=2.9231e-03) (hash(x)=87223122)
138
+ 1120 train 4.418639 (lr=2.9192e-03) (hash(x)=77292786)
139
+ 1130 train 4.433734 (lr=2.9152e-03) (hash(x)=88761375)
140
+ 1140 train 4.346690 (lr=2.9112e-03) (hash(x)=80956468)
141
+ 1150 train 4.517535 (lr=2.9070e-03) (hash(x)=84725462)
142
+ 1160 train 4.511478 (lr=2.9027e-03) (hash(x)=76098113)
143
+ 1170 train 4.485915 (lr=2.8983e-03) (hash(x)=84387685)
144
+ 1180 train 4.434893 (lr=2.8939e-03) (hash(x)=76120321)
145
+ 1190 train 4.394454 (lr=2.8893e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4131
147
+ 1200 val perplexity 82.5270
148
+ 1200 train 4.301891 (lr=2.8847e-03) (hash(x)=79340644)
149
+ 1210 train 4.397011 (lr=2.8800e-03) (hash(x)=88345025)
150
+ 1220 train 4.299796 (lr=2.8751e-03) (hash(x)=86406230)
151
+ 1230 train 4.292518 (lr=2.8702e-03) (hash(x)=80608975)
152
+ 1240 train 4.354300 (lr=2.8652e-03) (hash(x)=81635225)
153
+ 1250 train 4.183466 (lr=2.8601e-03) (hash(x)=82126376)
154
+ 1260 train 4.354262 (lr=2.8550e-03) (hash(x)=83737972)
155
+ 1270 train 4.270424 (lr=2.8497e-03) (hash(x)=81132345)
156
+ 1280 train 4.367188 (lr=2.8443e-03) (hash(x)=85850404)
157
+ 1290 train 4.364059 (lr=2.8389e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.3715
159
+ 1300 val perplexity 79.1662
160
+ 1300 train 4.347380 (lr=2.8333e-03) (hash(x)=77545187)
161
+ 1310 train 4.398432 (lr=2.8277e-03) (hash(x)=86412685)
162
+ 1320 train 4.294516 (lr=2.8220e-03) (hash(x)=86429640)
163
+ 1330 train 4.423416 (lr=2.8162e-03) (hash(x)=79645281)
164
+ 1340 train 4.379417 (lr=2.8103e-03) (hash(x)=82668541)
165
+ 1350 train 4.242847 (lr=2.8044e-03) (hash(x)=73627248)
166
+ 1360 train 4.345039 (lr=2.7983e-03) (hash(x)=74845133)
167
+ 1370 train 4.248496 (lr=2.7922e-03) (hash(x)=77850497)
168
+ 1380 train 4.280531 (lr=2.7860e-03) (hash(x)=76143954)
169
+ 1390 train 4.291858 (lr=2.7797e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.3550
171
+ 1400 val perplexity 77.8637
172
+ 1400 train 4.266463 (lr=2.7733e-03) (hash(x)=76010938)
173
+ 1410 train 4.309113 (lr=2.7668e-03) (hash(x)=81260300)
174
+ 1420 train 4.427899 (lr=2.7603e-03) (hash(x)=83803629)
175
+ 1430 train 4.254822 (lr=2.7536e-03) (hash(x)=83840110)
176
+ 1440 train 4.444929 (lr=2.7469e-03) (hash(x)=97711831)
177
+ 1450 train 4.321461 (lr=2.7401e-03) (hash(x)=81110139)
178
+ 1460 train 4.251062 (lr=2.7333e-03) (hash(x)=90942839)
179
+ 1470 train 4.286316 (lr=2.7263e-03) (hash(x)=80721564)
180
+ 1480 train 4.386154 (lr=2.7193e-03) (hash(x)=77852059)
181
+ 1490 train 4.280328 (lr=2.7121e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3149
183
+ 1500 val perplexity 74.8061
184
+ 1500 train 4.302010 (lr=2.7050e-03) (hash(x)=84496142)
185
+ 1510 train 4.242334 (lr=2.6977e-03) (hash(x)=70186729)
186
+ 1520 train 4.265534 (lr=2.6903e-03) (hash(x)=74854227)
187
+ 1530 train 4.174006 (lr=2.6829e-03) (hash(x)=77468161)
188
+ 1540 train 4.189521 (lr=2.6754e-03) (hash(x)=87853059)
189
+ 1550 train 4.126344 (lr=2.6678e-03) (hash(x)=71225436)
190
+ 1560 train 4.228718 (lr=2.6602e-03) (hash(x)=84072783)
191
+ 1570 train 4.275251 (lr=2.6525e-03) (hash(x)=72723098)
192
+ 1580 train 4.293262 (lr=2.6447e-03) (hash(x)=90409866)
193
+ 1590 train 4.366618 (lr=2.6368e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.2893
195
+ 1600 val perplexity 72.9171
196
+ 1600 train 4.350384 (lr=2.6289e-03) (hash(x)=77643862)
197
+ 1610 train 4.237774 (lr=2.6208e-03) (hash(x)=88862575)
198
+ 1620 train 4.254552 (lr=2.6128e-03) (hash(x)=84612581)
199
+ 1630 train 4.292946 (lr=2.6046e-03) (hash(x)=87075989)
200
+ 1640 train 4.173844 (lr=2.5964e-03) (hash(x)=88277361)
201
+ 1650 train 4.291949 (lr=2.5881e-03) (hash(x)=78750236)
202
+ 1660 train 4.207770 (lr=2.5797e-03) (hash(x)=82604581)
203
+ 1670 train 4.050609 (lr=2.5713e-03) (hash(x)=68482265)
204
+ 1680 train 4.176988 (lr=2.5628e-03) (hash(x)=75088835)
205
+ 1690 train 4.124127 (lr=2.5542e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.2899
207
+ 1700 val perplexity 72.9570
208
+ 1700 train 4.281199 (lr=2.5455e-03) (hash(x)=79986754)
209
+ 1710 train 4.269662 (lr=2.5368e-03) (hash(x)=83657930)
210
+ 1720 train 4.216094 (lr=2.5281e-03) (hash(x)=81754135)
211
+ 1730 train 4.300531 (lr=2.5192e-03) (hash(x)=78904427)
212
+ 1740 train 4.202200 (lr=2.5103e-03) (hash(x)=85920177)
213
+ 1750 train 4.256034 (lr=2.5014e-03) (hash(x)=86573211)
214
+ 1760 train 4.207814 (lr=2.4924e-03) (hash(x)=81737128)
215
+ 1770 train 4.202887 (lr=2.4833e-03) (hash(x)=77400968)
216
+ 1780 train 4.218459 (lr=2.4741e-03) (hash(x)=73545497)
217
+ 1790 train 4.469220 (lr=2.4649e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.2461
219
+ 1800 val perplexity 69.8295
220
+ 1800 train 4.311965 (lr=2.4556e-03) (hash(x)=87819781)
221
+ 1810 train 4.172852 (lr=2.4463e-03) (hash(x)=86870770)
222
+ 1820 train 4.177966 (lr=2.4369e-03) (hash(x)=82522211)
223
+ 1830 train 4.267137 (lr=2.4275e-03) (hash(x)=61947437)
224
+ 1840 train 4.142032 (lr=2.4180e-03) (hash(x)=79865406)
225
+ 1850 train 4.106224 (lr=2.4084e-03) (hash(x)=79828721)
226
+ 1860 train 4.251416 (lr=2.3988e-03) (hash(x)=80869571)
227
+ 1870 train 4.235191 (lr=2.3891e-03) (hash(x)=73780971)
228
+ 1880 train 4.198192 (lr=2.3794e-03) (hash(x)=79249549)
229
+ 1890 train 4.255785 (lr=2.3696e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.2232
231
+ 1900 val perplexity 68.2546
232
+ 1900 train 4.306304 (lr=2.3598e-03) (hash(x)=82456430)
233
+ 1910 train 4.094693 (lr=2.3499e-03) (hash(x)=82222135)
234
+ 1920 train 4.126866 (lr=2.3400e-03) (hash(x)=70033249)
235
+ 1930 train 4.119643 (lr=2.3300e-03) (hash(x)=72887360)
236
+ 1940 train 4.164147 (lr=2.3200e-03) (hash(x)=83251100)
237
+ 1950 train 4.198253 (lr=2.3099e-03) (hash(x)=79660266)
238
+ 1960 train 4.090333 (lr=2.2998e-03) (hash(x)=78919068)
239
+ 1970 train 4.184448 (lr=2.2896e-03) (hash(x)=94690431)
240
+ 1980 train 4.011884 (lr=2.2793e-03) (hash(x)=88153756)
241
+ 1990 train 4.072132 (lr=2.2691e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.2075
243
+ 2000 val perplexity 67.1864
244
+ 2000 train 4.168382 (lr=2.2588e-03) (hash(x)=81308591)
245
+ 2010 train 4.218012 (lr=2.2484e-03) (hash(x)=80205479)
246
+ 2020 train 4.125780 (lr=2.2380e-03) (hash(x)=81008704)
247
+ 2030 train 4.203201 (lr=2.2275e-03) (hash(x)=82914358)
248
+ 2040 train 4.213617 (lr=2.2170e-03) (hash(x)=88064399)
249
+ 2050 train 4.219054 (lr=2.2065e-03) (hash(x)=82889390)
250
+ 2060 train 4.286422 (lr=2.1959e-03) (hash(x)=80913554)
251
+ 2070 train 4.125287 (lr=2.1853e-03) (hash(x)=78121791)
252
+ 2080 train 4.262558 (lr=2.1746e-03) (hash(x)=75356657)
253
+ 2090 train 4.092182 (lr=2.1639e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.1861
255
+ 2100 val perplexity 65.7654
256
+ 2100 train 4.012233 (lr=2.1532e-03) (hash(x)=68928225)
257
+ 2110 train 4.062612 (lr=2.1424e-03) (hash(x)=93734745)
258
+ 2120 train 4.057480 (lr=2.1316e-03) (hash(x)=72878164)
259
+ 2130 train 4.190535 (lr=2.1208e-03) (hash(x)=71508226)
260
+ 2140 train 4.141170 (lr=2.1099e-03) (hash(x)=84168671)
261
+ 2150 train 4.079234 (lr=2.0990e-03) (hash(x)=74673639)
262
+ 2160 train 4.232322 (lr=2.0881e-03) (hash(x)=75470031)
263
+ 2170 train 4.187548 (lr=2.0771e-03) (hash(x)=74307890)
264
+ 2180 train 4.301085 (lr=2.0661e-03) (hash(x)=77214245)
265
+ 2190 train 4.087014 (lr=2.0550e-03) (hash(x)=88628359)
266
+ 2200 val loss 4.1617
267
+ 2200 val perplexity 64.1782
268
+ 2200 train 4.102144 (lr=2.0440e-03) (hash(x)=74779126)
269
+ 2210 train 4.144354 (lr=2.0329e-03) (hash(x)=79817976)
270
+ 2220 train 4.088383 (lr=2.0217e-03) (hash(x)=75258996)
271
+ 2230 train 4.068884 (lr=2.0106e-03) (hash(x)=88164047)
272
+ 2240 train 4.052829 (lr=1.9994e-03) (hash(x)=78580686)
273
+ 2250 train 3.987047 (lr=1.9882e-03) (hash(x)=76510617)
274
+ 2260 train 4.069172 (lr=1.9770e-03) (hash(x)=76043966)
275
+ 2270 train 4.039061 (lr=1.9657e-03) (hash(x)=73062098)
276
+ 2280 train 4.267662 (lr=1.9544e-03) (hash(x)=81885909)
277
+ 2290 train 4.204568 (lr=1.9431e-03) (hash(x)=81481741)
278
+ 2300 val loss 4.1417
279
+ 2300 val perplexity 62.9121
280
+ 2300 train 4.061571 (lr=1.9318e-03) (hash(x)=82104275)
281
+ 2310 train 4.111306 (lr=1.9205e-03) (hash(x)=82137309)
282
+ 2320 train 4.089928 (lr=1.9091e-03) (hash(x)=86741167)
283
+ 2330 train 4.096393 (lr=1.8977e-03) (hash(x)=84796263)
284
+ 2340 train 4.079998 (lr=1.8863e-03) (hash(x)=169966529)
285
+ 2350 train 3.950702 (lr=1.8749e-03) (hash(x)=80006061)
286
+ 2360 train 4.006207 (lr=1.8635e-03) (hash(x)=71090999)
287
+ 2370 train 4.136633 (lr=1.8520e-03) (hash(x)=86641850)
288
+ 2380 train 4.034856 (lr=1.8406e-03) (hash(x)=84573253)
289
+ 2390 train 4.067786 (lr=1.8291e-03) (hash(x)=74483764)
290
+ 2400 val loss 4.1361
291
+ 2400 val perplexity 62.5592
292
+ 2400 train 4.043357 (lr=1.8176e-03) (hash(x)=78327659)
293
+ 2410 train 4.107500 (lr=1.8061e-03) (hash(x)=77273627)
294
+ 2420 train 3.977954 (lr=1.7946e-03) (hash(x)=76938049)
295
+ 2430 train 3.971897 (lr=1.7830e-03) (hash(x)=98449442)
296
+ 2440 train 3.837660 (lr=1.7715e-03) (hash(x)=78157797)
297
+ 2450 train 3.878682 (lr=1.7600e-03) (hash(x)=80637582)
298
+ 2460 train 3.864181 (lr=1.7484e-03) (hash(x)=80023854)
299
+ 2470 train 3.930397 (lr=1.7368e-03) (hash(x)=86678884)
300
+ 2480 train 4.154822 (lr=1.7253e-03) (hash(x)=84086469)
301
+ 2490 train 4.176461 (lr=1.7137e-03) (hash(x)=83131141)
302
+ 2500 val loss 4.1134
303
+ 2500 val perplexity 61.1557
304
+ 2500 train 4.093238 (lr=1.7021e-03) (hash(x)=82583497)
305
+ 2510 train 4.147510 (lr=1.6906e-03) (hash(x)=79164326)
306
+ 2520 train 4.058905 (lr=1.6790e-03) (hash(x)=76456503)
307
+ 2530 train 4.116585 (lr=1.6674e-03) (hash(x)=82001228)
308
+ 2540 train 4.148282 (lr=1.6558e-03) (hash(x)=75615595)
309
+ 2550 train 4.096762 (lr=1.6442e-03) (hash(x)=78397869)
310
+ 2560 train 4.098824 (lr=1.6326e-03) (hash(x)=75547032)
311
+ 2570 train 4.012700 (lr=1.6210e-03) (hash(x)=89201025)
312
+ 2580 train 4.000011 (lr=1.6094e-03) (hash(x)=89856704)
313
+ 2590 train 4.007967 (lr=1.5979e-03) (hash(x)=82175682)
314
+ 2600 val loss 4.0956
315
+ 2600 val perplexity 60.0731
316
+ 2600 train 4.020466 (lr=1.5863e-03) (hash(x)=78312826)
317
+ 2610 train 4.004652 (lr=1.5747e-03) (hash(x)=77066588)
318
+ 2620 train 3.843831 (lr=1.5632e-03) (hash(x)=78666061)
319
+ 2630 train 3.847181 (lr=1.5516e-03) (hash(x)=93762143)
320
+ 2640 train 3.865944 (lr=1.5400e-03) (hash(x)=83191587)
321
+ 2650 train 3.777612 (lr=1.5285e-03) (hash(x)=87169585)
322
+ 2660 train 4.005943 (lr=1.5170e-03) (hash(x)=86426388)
323
+ 2670 train 4.142155 (lr=1.5054e-03) (hash(x)=76692638)
324
+ 2680 train 4.046444 (lr=1.4939e-03) (hash(x)=77446063)
325
+ 2690 train 4.050703 (lr=1.4824e-03) (hash(x)=79809050)
326
+ 2700 val loss 4.0875
327
+ 2700 val perplexity 59.5929
328
+ 2700 train 3.978821 (lr=1.4709e-03) (hash(x)=83116823)
329
+ 2710 train 4.094662 (lr=1.4594e-03) (hash(x)=75622148)
330
+ 2720 train 4.025250 (lr=1.4480e-03) (hash(x)=80690512)
331
+ 2730 train 4.093600 (lr=1.4365e-03) (hash(x)=78009984)
332
+ 2740 train 4.047515 (lr=1.4251e-03) (hash(x)=74205488)
333
+ 2750 train 4.054039 (lr=1.4137e-03) (hash(x)=91013332)
334
+ 2760 train 3.972038 (lr=1.4023e-03) (hash(x)=76954961)
335
+ 2770 train 3.982512 (lr=1.3909e-03) (hash(x)=79390317)
336
+ 2780 train 3.854488 (lr=1.3795e-03) (hash(x)=70168783)
337
+ 2790 train 4.017802 (lr=1.3682e-03) (hash(x)=76028417)
338
+ 2800 val loss 4.0678
339
+ 2800 val perplexity 58.4290
340
+ 2800 train 3.839256 (lr=1.3569e-03) (hash(x)=77656050)
341
+ 2810 train 3.882767 (lr=1.3456e-03) (hash(x)=94248216)
342
+ 2820 train 3.948845 (lr=1.3343e-03) (hash(x)=78305078)
343
+ 2830 train 3.860161 (lr=1.3230e-03) (hash(x)=79948848)
344
+ 2840 train 3.730049 (lr=1.3118e-03) (hash(x)=85341024)
345
+ 2850 train 4.038423 (lr=1.3006e-03) (hash(x)=78735170)
346
+ 2860 train 4.285612 (lr=1.2894e-03) (hash(x)=71616419)
347
+ 2870 train 4.057415 (lr=1.2783e-03) (hash(x)=78656517)
348
+ 2880 train 4.069396 (lr=1.2671e-03) (hash(x)=80073987)
349
+ 2890 train 4.002001 (lr=1.2560e-03) (hash(x)=76894809)
350
+ 2900 val loss 4.0433
351
+ 2900 val perplexity 57.0158
352
+ 2900 train 3.980704 (lr=1.2450e-03) (hash(x)=80499838)
353
+ 2910 train 4.028818 (lr=1.2339e-03) (hash(x)=72673354)
354
+ 2920 train 4.011950 (lr=1.2229e-03) (hash(x)=84265768)
355
+ 2930 train 4.038533 (lr=1.2119e-03) (hash(x)=79612060)
356
+ 2940 train 3.891789 (lr=1.2010e-03) (hash(x)=74970087)
357
+ 2950 train 4.052449 (lr=1.1901e-03) (hash(x)=84166818)
358
+ 2960 train 4.041331 (lr=1.1792e-03) (hash(x)=89410221)
359
+ 2970 train 3.976947 (lr=1.1684e-03) (hash(x)=75672566)
360
+ 2980 train 3.832167 (lr=1.1576e-03) (hash(x)=81760314)
361
+ 2990 train 3.807815 (lr=1.1468e-03) (hash(x)=80605200)
362
+ 3000 val loss 4.0445
363
+ 3000 val perplexity 57.0854
364
+ 3000 train 3.800022 (lr=1.1361e-03) (hash(x)=83804735)
365
+ 3010 train 3.726972 (lr=1.1254e-03) (hash(x)=77015303)
366
+ 3020 train 3.799004 (lr=1.1147e-03) (hash(x)=81464523)
367
+ 3030 train 3.995228 (lr=1.1041e-03) (hash(x)=80168230)
368
+ 3040 train 3.938375 (lr=1.0935e-03) (hash(x)=84817006)
369
+ 3050 train 4.044604 (lr=1.0830e-03) (hash(x)=71601811)
370
+ 3060 train 3.966243 (lr=1.0725e-03) (hash(x)=85499733)
371
+ 3070 train 4.022532 (lr=1.0620e-03) (hash(x)=78670408)
372
+ 3080 train 3.917745 (lr=1.0516e-03) (hash(x)=77120468)
373
+ 3090 train 4.019437 (lr=1.0412e-03) (hash(x)=77927426)
374
+ 3100 val loss 4.0217
375
+ 3100 val perplexity 55.7982
376
+ 3100 train 4.036481 (lr=1.0309e-03) (hash(x)=83998606)
377
+ 3110 train 3.936145 (lr=1.0207e-03) (hash(x)=77990218)
378
+ 3120 train 3.850933 (lr=1.0104e-03) (hash(x)=81623970)
379
+ 3130 train 3.826051 (lr=1.0002e-03) (hash(x)=75014781)
380
+ 3140 train 3.876135 (lr=9.9011e-04) (hash(x)=72591250)
381
+ 3150 train 3.889138 (lr=9.8002e-04) (hash(x)=81421847)
382
+ 3160 train 3.940847 (lr=9.6999e-04) (hash(x)=75112631)
383
+ 3170 train 4.031105 (lr=9.6000e-04) (hash(x)=87518033)
384
+ 3180 train 3.894937 (lr=9.5007e-04) (hash(x)=78648348)
385
+ 3190 train 4.000477 (lr=9.4019e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0002
387
+ 3200 val perplexity 54.6086
388
+ 3200 train 4.073890 (lr=9.3036e-04) (hash(x)=83380714)
389
+ 3210 train 4.261600 (lr=9.2058e-04) (hash(x)=67804991)
390
+ 3220 train 3.986545 (lr=9.1085e-04) (hash(x)=76968804)
391
+ 3230 train 3.917281 (lr=9.0118e-04) (hash(x)=75798670)
392
+ 3240 train 3.994160 (lr=8.9157e-04) (hash(x)=83192811)
393
+ 3250 train 3.931210 (lr=8.8201e-04) (hash(x)=81659789)
394
+ 3260 train 3.977590 (lr=8.7251e-04) (hash(x)=80622502)
395
+ 3270 train 3.945882 (lr=8.6307e-04) (hash(x)=85436511)
396
+ 3280 train 3.889329 (lr=8.5368e-04) (hash(x)=80589180)
397
+ 3290 train 4.027271 (lr=8.4435e-04) (hash(x)=85747193)
398
+ 3300 val loss 3.9962
399
+ 3300 val perplexity 54.3918
400
+ 3300 train 3.892448 (lr=8.3508e-04) (hash(x)=74635692)
401
+ 3310 train 3.875926 (lr=8.2588e-04) (hash(x)=82045455)
402
+ 3320 train 3.752017 (lr=8.1673e-04) (hash(x)=73249173)
403
+ 3330 train 3.711201 (lr=8.0764e-04) (hash(x)=71492338)
404
+ 3340 train 4.043011 (lr=7.9862e-04) (hash(x)=76374071)
405
+ 3350 train 4.012915 (lr=7.8966e-04) (hash(x)=80474064)
406
+ 3360 train 3.922348 (lr=7.8076e-04) (hash(x)=84390892)
407
+ 3370 train 3.969140 (lr=7.7192e-04) (hash(x)=83399949)
408
+ 3380 train 4.020267 (lr=7.6315e-04) (hash(x)=77648059)
409
+ 3390 train 3.993971 (lr=7.5445e-04) (hash(x)=80152701)
410
+ 3400 val loss 3.9755
411
+ 3400 val perplexity 53.2791
412
+ 3400 train 4.040460 (lr=7.4581e-04) (hash(x)=81952545)
413
+ 3410 train 3.998460 (lr=7.3724e-04) (hash(x)=80908993)
414
+ 3420 train 4.043931 (lr=7.2874e-04) (hash(x)=79610037)
415
+ 3430 train 3.895556 (lr=7.2030e-04) (hash(x)=87624382)
416
+ 3440 train 3.969567 (lr=7.1193e-04) (hash(x)=82336381)
417
+ 3450 train 3.945179 (lr=7.0363e-04) (hash(x)=87687835)
418
+ 3460 train 3.948816 (lr=6.9541e-04) (hash(x)=76376135)
419
+ 3470 train 3.859250 (lr=6.8725e-04) (hash(x)=79331391)
420
+ 3480 train 3.761140 (lr=6.7916e-04) (hash(x)=107002681)
421
+ 3490 train 3.849377 (lr=6.7114e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.9726
423
+ 3500 val perplexity 53.1212
424
+ 3500 train 3.805276 (lr=6.6320e-04) (hash(x)=88237229)
425
+ 3510 train 3.879308 (lr=6.5533e-04) (hash(x)=81011739)
426
+ 3520 train 3.926079 (lr=6.4753e-04) (hash(x)=64643427)
427
+ 3530 train 4.205634 (lr=6.3981e-04) (hash(x)=78029539)
428
+ 3540 train 4.028028 (lr=6.3216e-04) (hash(x)=83188968)
429
+ 3550 train 4.053922 (lr=6.2458e-04) (hash(x)=86104185)
430
+ 3560 train 3.976454 (lr=6.1708e-04) (hash(x)=81993629)
431
+ 3570 train 3.948395 (lr=6.0966e-04) (hash(x)=72719368)
432
+ 3580 train 3.958232 (lr=6.0231e-04) (hash(x)=90091487)
433
+ 3590 train 3.843460 (lr=5.9504e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9530
435
+ 3600 val perplexity 52.0908
436
+ 3600 train 3.954013 (lr=5.8785e-04) (hash(x)=73103504)
437
+ 3610 train 3.956662 (lr=5.8074e-04) (hash(x)=84429400)
438
+ 3620 train 3.935569 (lr=5.7370e-04) (hash(x)=77021795)
439
+ 3630 train 3.886144 (lr=5.6675e-04) (hash(x)=92830605)
440
+ 3640 train 4.086278 (lr=5.5987e-04) (hash(x)=78313175)
441
+ 3650 train 3.882939 (lr=5.5308e-04) (hash(x)=85395549)
442
+ 3660 train 3.681696 (lr=5.4636e-04) (hash(x)=78114459)
443
+ 3670 train 3.804089 (lr=5.3973e-04) (hash(x)=74968316)
444
+ 3680 train 3.731938 (lr=5.3318e-04) (hash(x)=73358737)
445
+ 3690 train 3.827537 (lr=5.2671e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9541
447
+ 3700 val perplexity 52.1487
448
+ 3700 train 3.840573 (lr=5.2033e-04) (hash(x)=74039273)
449
+ 3710 train 4.014916 (lr=5.1402e-04) (hash(x)=79299680)
450
+ 3720 train 3.929354 (lr=5.0780e-04) (hash(x)=94701498)
451
+ 3730 train 4.006836 (lr=5.0167e-04) (hash(x)=75352071)
452
+ 3740 train 3.886767 (lr=4.9562e-04) (hash(x)=77909487)
453
+ 3750 train 3.988847 (lr=4.8965e-04) (hash(x)=73986730)
454
+ 3760 train 3.953511 (lr=4.8377e-04) (hash(x)=79325763)
455
+ 3770 train 3.997174 (lr=4.7798e-04) (hash(x)=72457818)
456
+ 3780 train 4.125257 (lr=4.7227e-04) (hash(x)=71775590)
457
+ 3790 train 4.015434 (lr=4.6665e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9363
459
+ 3800 val perplexity 51.2305
460
+ 3800 train 3.990609 (lr=4.6112e-04) (hash(x)=79965893)
461
+ 3810 train 3.927805 (lr=4.5567e-04) (hash(x)=72598235)
462
+ 3820 train 3.909876 (lr=4.5031e-04) (hash(x)=83113889)
463
+ 3830 train 3.968087 (lr=4.4504e-04) (hash(x)=74434590)
464
+ 3840 train 3.889962 (lr=4.3986e-04) (hash(x)=82860348)
465
+ 3850 train 3.824477 (lr=4.3477e-04) (hash(x)=78067565)
466
+ 3860 train 3.890043 (lr=4.2977e-04) (hash(x)=82592498)
467
+ 3870 train 3.820526 (lr=4.2486e-04) (hash(x)=81820733)
468
+ 3880 train 3.837957 (lr=4.2004e-04) (hash(x)=87709040)
469
+ 3890 train 3.880893 (lr=4.1530e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9299
471
+ 3900 val perplexity 50.9039
472
+ 3900 train 4.004045 (lr=4.1066e-04) (hash(x)=76597431)
473
+ 3910 train 3.884222 (lr=4.0611e-04) (hash(x)=90490716)
474
+ 3920 train 3.991057 (lr=4.0166e-04) (hash(x)=81970659)
475
+ 3930 train 3.975126 (lr=3.9729e-04) (hash(x)=81496334)
476
+ 3940 train 3.960249 (lr=3.9302e-04) (hash(x)=75717605)
477
+ 3950 train 3.903774 (lr=3.8884e-04) (hash(x)=82749357)
478
+ 3960 train 3.931346 (lr=3.8475e-04) (hash(x)=89355157)
479
+ 3970 train 3.885843 (lr=3.8076e-04) (hash(x)=78980403)
480
+ 3980 train 3.857227 (lr=3.7685e-04) (hash(x)=76627217)
481
+ 3990 train 3.860527 (lr=3.7305e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9247
483
+ 4000 val perplexity 50.6376
484
+ 4000 train 3.870709 (lr=3.6933e-04) (hash(x)=83018142)
485
+ 4010 train 3.889404 (lr=3.6572e-04) (hash(x)=81272436)
486
+ 4020 train 3.814106 (lr=3.6219e-04) (hash(x)=85497482)
487
+ 4030 train 3.886128 (lr=3.5876e-04) (hash(x)=85241734)
488
+ 4040 train 3.852141 (lr=3.5543e-04) (hash(x)=77925307)
489
+ 4050 train 3.858808 (lr=3.5219e-04) (hash(x)=84826179)
490
+ 4060 train 3.913105 (lr=3.4905e-04) (hash(x)=83606764)
491
+ 4070 train 3.826577 (lr=3.4600e-04) (hash(x)=80567590)
492
+ 4080 train 3.875301 (lr=3.4305e-04) (hash(x)=76860998)
493
+ 4090 train 3.917786 (lr=3.4019e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9120
495
+ 4100 val perplexity 50.0003
496
+ 4100 train 3.958393 (lr=3.3744e-04) (hash(x)=82832041)
497
+ 4110 train 3.838257 (lr=3.3477e-04) (hash(x)=79143262)
498
+ 4120 train 3.864722 (lr=3.3221e-04) (hash(x)=77038149)
499
+ 4130 train 4.044418 (lr=3.2974e-04) (hash(x)=86339074)
500
+ 4140 train 4.108546 (lr=3.2737e-04) (hash(x)=76686216)
501
+ 4150 train 3.890633 (lr=3.2510e-04) (hash(x)=70522682)
502
+ 4160 train 3.849080 (lr=3.2292e-04) (hash(x)=90958555)
503
+ 4170 train 3.873763 (lr=3.2085e-04) (hash(x)=91463532)
504
+ 4180 train 3.862164 (lr=3.1887e-04) (hash(x)=81959329)
505
+ 4190 train 3.820952 (lr=3.1699e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9139
507
+ 4200 val perplexity 50.0957
508
+ 4200 train 3.807653 (lr=3.1520e-04) (hash(x)=78361715)
509
+ 4210 train 3.951216 (lr=3.1352e-04) (hash(x)=87364889)
510
+ 4220 train 3.806162 (lr=3.1193e-04) (hash(x)=70465156)
511
+ 4230 train 3.961999 (lr=3.1044e-04) (hash(x)=84524081)
512
+ 4240 train 3.913294 (lr=3.0905e-04) (hash(x)=77824868)
513
+ 4250 train 3.931151 (lr=3.0776e-04) (hash(x)=81710711)
514
+ 4260 train 3.912488 (lr=3.0657e-04) (hash(x)=76362728)
515
+ 4270 train 3.903061 (lr=3.0548e-04) (hash(x)=83115208)
516
+ 4280 train 3.816491 (lr=3.0449e-04) (hash(x)=87218314)
517
+ 4290 train 3.978020 (lr=3.0359e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9088
519
+ 4300 val perplexity 49.8407
520
+ 4300 train 3.849984 (lr=3.0280e-04) (hash(x)=77379615)
521
+ 4310 train 3.828372 (lr=3.0210e-04) (hash(x)=78669579)
522
+ 4320 train 3.939689 (lr=3.0150e-04) (hash(x)=83066608)
523
+ 4330 train 3.907775 (lr=3.0101e-04) (hash(x)=83037340)
524
+ 4340 train 3.944016 (lr=3.0061e-04) (hash(x)=82849771)
525
+ 4350 train 3.878971 (lr=3.0031e-04) (hash(x)=76693985)
526
+ 4360 train 3.970914 (lr=3.0011e-04) (hash(x)=77745394)
527
+ 4370 train 3.931363 (lr=3.0001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9020
529
+ 4374 val perplexity 49.5008
12_mini_head_two_masks_4_heads/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b163a5c930474e0be2392d28239030d42da2b7b14125ebc30374c9cff638679
3
+ size 98537090
12_mini_head_two_masks_4_heads/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d6933615d1cc42239a57cc72c2f16f97e97e393194b5ea90cfb81ec9852b9d0
3
+ size 190793606