andrew-healey commited on
Commit
1f23491
·
verified ·
1 Parent(s): f556b3b

Upload folder using huggingface_hub

Browse files
12_head_one_mask_per_head_4_latent_masks/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_head_one_mask_per_head_4_latent_masks", "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, "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": "n_latent_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": 4, "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": 0.0016, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_one_mask_per_head_4_latent_masks"}
12_head_one_mask_per_head_4_latent_masks/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_head_one_mask_per_head_4_latent_masks/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 10.9739
3
+ 0 val perplexity 58331.1328
4
+ 0 train 10.976922 (lr=6.4000e-06) (hash(x)=93356070)
5
+ 10 train 9.610164 (lr=7.0400e-05) (hash(x)=91253010)
6
+ 20 train 8.822256 (lr=1.3440e-04) (hash(x)=74934453)
7
+ 30 train 8.045877 (lr=1.9840e-04) (hash(x)=79278034)
8
+ 40 train 7.623817 (lr=2.6240e-04) (hash(x)=80107892)
9
+ 50 train 7.598309 (lr=3.2640e-04) (hash(x)=70745428)
10
+ 60 train 7.579951 (lr=3.9040e-04) (hash(x)=80077589)
11
+ 70 train 7.124553 (lr=4.5440e-04) (hash(x)=76213766)
12
+ 80 train 7.080781 (lr=5.1840e-04) (hash(x)=83218328)
13
+ 90 train 6.913459 (lr=5.8240e-04) (hash(x)=74855845)
14
+ 100 val loss 6.8669
15
+ 100 val perplexity 959.9525
16
+ 100 train 6.702640 (lr=6.4640e-04) (hash(x)=82814902)
17
+ 110 train 6.686010 (lr=7.1040e-04) (hash(x)=84907741)
18
+ 120 train 6.561395 (lr=7.7440e-04) (hash(x)=82613223)
19
+ 130 train 6.423313 (lr=8.3840e-04) (hash(x)=83540876)
20
+ 140 train 6.406285 (lr=9.0240e-04) (hash(x)=75095216)
21
+ 150 train 6.504651 (lr=9.6640e-04) (hash(x)=97190944)
22
+ 160 train 6.436491 (lr=1.0304e-03) (hash(x)=82117809)
23
+ 170 train 6.334200 (lr=1.0944e-03) (hash(x)=70514724)
24
+ 180 train 6.270076 (lr=1.1584e-03) (hash(x)=81029624)
25
+ 190 train 6.208189 (lr=1.2224e-03) (hash(x)=78996841)
26
+ 200 val loss 6.1355
27
+ 200 val perplexity 461.9548
28
+ 200 train 6.145214 (lr=1.2864e-03) (hash(x)=79845097)
29
+ 210 train 6.047606 (lr=1.3504e-03) (hash(x)=78997683)
30
+ 220 train 5.962054 (lr=1.4144e-03) (hash(x)=74895865)
31
+ 230 train 6.112282 (lr=1.4784e-03) (hash(x)=80933276)
32
+ 240 train 5.803537 (lr=1.5424e-03) (hash(x)=77664606)
33
+ 250 train 5.844026 (lr=1.6000e-03) (hash(x)=75903930)
34
+ 260 train 5.755938 (lr=1.6000e-03) (hash(x)=82996853)
35
+ 270 train 5.677665 (lr=1.5999e-03) (hash(x)=73269568)
36
+ 280 train 5.599977 (lr=1.5998e-03) (hash(x)=82906427)
37
+ 290 train 5.523516 (lr=1.5997e-03) (hash(x)=71797895)
38
+ 300 val loss 5.6848
39
+ 300 val perplexity 294.3557
40
+ 300 train 5.378790 (lr=1.5995e-03) (hash(x)=85232249)
41
+ 310 train 5.319650 (lr=1.5992e-03) (hash(x)=75252489)
42
+ 320 train 5.405895 (lr=1.5990e-03) (hash(x)=74394644)
43
+ 330 train 5.496866 (lr=1.5987e-03) (hash(x)=82772910)
44
+ 340 train 5.542979 (lr=1.5983e-03) (hash(x)=81627464)
45
+ 350 train 5.508110 (lr=1.5979e-03) (hash(x)=84678053)
46
+ 360 train 5.447223 (lr=1.5975e-03) (hash(x)=81884128)
47
+ 370 train 5.454436 (lr=1.5970e-03) (hash(x)=76379242)
48
+ 380 train 5.267288 (lr=1.5965e-03) (hash(x)=98182875)
49
+ 390 train 5.234765 (lr=1.5959e-03) (hash(x)=79710436)
50
+ 400 val loss 5.2849
51
+ 400 val perplexity 197.3264
52
+ 400 train 5.233985 (lr=1.5953e-03) (hash(x)=79841071)
53
+ 410 train 5.188588 (lr=1.5947e-03) (hash(x)=75844151)
54
+ 420 train 5.164733 (lr=1.5940e-03) (hash(x)=73125036)
55
+ 430 train 5.210048 (lr=1.5932e-03) (hash(x)=84214858)
56
+ 440 train 5.068330 (lr=1.5925e-03) (hash(x)=80456994)
57
+ 450 train 5.068047 (lr=1.5917e-03) (hash(x)=76735962)
58
+ 460 train 4.974701 (lr=1.5908e-03) (hash(x)=81845446)
59
+ 470 train 4.774404 (lr=1.5899e-03) (hash(x)=76094689)
60
+ 480 train 4.869439 (lr=1.5890e-03) (hash(x)=83806686)
61
+ 490 train 4.698429 (lr=1.5880e-03) (hash(x)=84690227)
62
+ 500 val loss 5.0409
63
+ 500 val perplexity 154.6122
64
+ 500 train 4.787300 (lr=1.5870e-03) (hash(x)=71851938)
65
+ 510 train 4.896961 (lr=1.5859e-03) (hash(x)=77159346)
66
+ 520 train 4.948387 (lr=1.5848e-03) (hash(x)=80755753)
67
+ 530 train 4.980748 (lr=1.5837e-03) (hash(x)=75379679)
68
+ 540 train 4.953137 (lr=1.5825e-03) (hash(x)=82458619)
69
+ 550 train 4.898540 (lr=1.5813e-03) (hash(x)=86366396)
70
+ 560 train 4.988710 (lr=1.5800e-03) (hash(x)=85095044)
71
+ 570 train 4.901937 (lr=1.5787e-03) (hash(x)=78385159)
72
+ 580 train 4.845851 (lr=1.5774e-03) (hash(x)=79342394)
73
+ 590 train 4.907160 (lr=1.5760e-03) (hash(x)=70782192)
74
+ 600 val loss 4.8935
75
+ 600 val perplexity 133.4138
76
+ 600 train 4.874977 (lr=1.5746e-03) (hash(x)=76130353)
77
+ 610 train 4.786975 (lr=1.5731e-03) (hash(x)=74778440)
78
+ 620 train 4.773256 (lr=1.5716e-03) (hash(x)=79129709)
79
+ 630 train 4.693193 (lr=1.5701e-03) (hash(x)=76469962)
80
+ 640 train 4.705766 (lr=1.5685e-03) (hash(x)=78288049)
81
+ 650 train 4.659735 (lr=1.5668e-03) (hash(x)=76641388)
82
+ 660 train 4.573209 (lr=1.5652e-03) (hash(x)=75937906)
83
+ 670 train 4.605877 (lr=1.5635e-03) (hash(x)=74807157)
84
+ 680 train 4.520327 (lr=1.5617e-03) (hash(x)=77490144)
85
+ 690 train 4.568202 (lr=1.5600e-03) (hash(x)=72900124)
86
+ 700 val loss 4.8356
87
+ 700 val perplexity 125.9128
88
+ 700 train 4.544733 (lr=1.5581e-03) (hash(x)=73424218)
89
+ 710 train 4.837429 (lr=1.5563e-03) (hash(x)=86052345)
90
+ 720 train 4.829470 (lr=1.5544e-03) (hash(x)=77462613)
91
+ 730 train 4.857572 (lr=1.5524e-03) (hash(x)=79027471)
92
+ 740 train 4.848959 (lr=1.5504e-03) (hash(x)=78149992)
93
+ 750 train 4.770066 (lr=1.5484e-03) (hash(x)=89147499)
94
+ 760 train 4.848103 (lr=1.5464e-03) (hash(x)=74931538)
95
+ 770 train 4.833956 (lr=1.5443e-03) (hash(x)=89648204)
96
+ 780 train 4.865717 (lr=1.5421e-03) (hash(x)=83708293)
97
+ 790 train 4.825516 (lr=1.5400e-03) (hash(x)=84336512)
98
+ 800 val loss 4.8041
99
+ 800 val perplexity 122.0127
100
+ 800 train 4.738462 (lr=1.5378e-03) (hash(x)=75025285)
101
+ 810 train 4.630217 (lr=1.5355e-03) (hash(x)=79436536)
102
+ 820 train 4.718338 (lr=1.5332e-03) (hash(x)=90467390)
103
+ 830 train 4.738597 (lr=1.5309e-03) (hash(x)=77292020)
104
+ 840 train 4.693678 (lr=1.5285e-03) (hash(x)=75568927)
105
+ 850 train 4.718541 (lr=1.5261e-03) (hash(x)=79671781)
106
+ 860 train 4.846093 (lr=1.5237e-03) (hash(x)=84280943)
107
+ 870 train 4.892892 (lr=1.5212e-03) (hash(x)=80901294)
108
+ 880 train 5.067603 (lr=1.5187e-03) (hash(x)=81437584)
109
+ 890 train 4.853504 (lr=1.5162e-03) (hash(x)=73893778)
110
+ 900 val loss 4.8819
111
+ 900 val perplexity 131.8840
112
+ 900 train 4.801009 (lr=1.5136e-03) (hash(x)=74987794)
113
+ 910 train 4.815507 (lr=1.5109e-03) (hash(x)=93721374)
114
+ 920 train 4.915287 (lr=1.5083e-03) (hash(x)=79149678)
115
+ 930 train 4.839033 (lr=1.5056e-03) (hash(x)=83179387)
116
+ 940 train 4.840887 (lr=1.5029e-03) (hash(x)=89009978)
117
+ 950 train 4.837934 (lr=1.5001e-03) (hash(x)=76483735)
118
+ 960 train 4.808829 (lr=1.4973e-03) (hash(x)=78638850)
119
+ 970 train 4.884855 (lr=1.4944e-03) (hash(x)=90415206)
120
+ 980 train 4.797330 (lr=1.4916e-03) (hash(x)=90819195)
121
+ 990 train 4.931005 (lr=1.4886e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.9117
123
+ 1000 val perplexity 135.8718
124
+ 1000 train 4.953598 (lr=1.4857e-03) (hash(x)=82436789)
125
+ 1010 train 4.964676 (lr=1.4827e-03) (hash(x)=63022149)
126
+ 1020 train 4.960149 (lr=1.4797e-03) (hash(x)=83501199)
127
+ 1030 train 4.999383 (lr=1.4767e-03) (hash(x)=77093326)
128
+ 1040 train 4.839481 (lr=1.4736e-03) (hash(x)=73437559)
129
+ 1050 train 4.971480 (lr=1.4704e-03) (hash(x)=84550388)
130
+ 1060 train 5.104366 (lr=1.4673e-03) (hash(x)=86184566)
131
+ 1070 train 4.972236 (lr=1.4641e-03) (hash(x)=85644922)
132
+ 1080 train 4.952349 (lr=1.4609e-03) (hash(x)=86179801)
133
+ 1090 train 4.906353 (lr=1.4576e-03) (hash(x)=82868303)
134
+ 1100 val loss 5.1277
135
+ 1100 val perplexity 168.6351
136
+ 1100 train 5.145108 (lr=1.4543e-03) (hash(x)=96780388)
137
+ 1110 train 4.999079 (lr=1.4510e-03) (hash(x)=87223122)
138
+ 1120 train 4.976474 (lr=1.4476e-03) (hash(x)=77292786)
139
+ 1130 train 4.942958 (lr=1.4443e-03) (hash(x)=88761375)
140
+ 1140 train 4.888082 (lr=1.4408e-03) (hash(x)=80956468)
141
+ 1150 train 4.999834 (lr=1.4374e-03) (hash(x)=84725462)
142
+ 1160 train 4.976130 (lr=1.4339e-03) (hash(x)=76098113)
143
+ 1170 train 5.036952 (lr=1.4304e-03) (hash(x)=84387685)
144
+ 1180 train 4.934584 (lr=1.4268e-03) (hash(x)=76120321)
145
+ 1190 train 4.858549 (lr=1.4232e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.8897
147
+ 1200 val perplexity 132.9076
148
+ 1200 train 4.764554 (lr=1.4196e-03) (hash(x)=79340644)
149
+ 1210 train 4.864171 (lr=1.4160e-03) (hash(x)=88345025)
150
+ 1220 train 4.852883 (lr=1.4123e-03) (hash(x)=86406230)
151
+ 1230 train 4.818523 (lr=1.4086e-03) (hash(x)=80608975)
152
+ 1240 train 4.856926 (lr=1.4049e-03) (hash(x)=81635225)
153
+ 1250 train 4.656309 (lr=1.4011e-03) (hash(x)=82126376)
154
+ 1260 train 4.806645 (lr=1.3973e-03) (hash(x)=83737972)
155
+ 1270 train 4.830136 (lr=1.3935e-03) (hash(x)=81132345)
156
+ 1280 train 4.853323 (lr=1.3896e-03) (hash(x)=85850404)
157
+ 1290 train 4.929839 (lr=1.3857e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.9413
159
+ 1300 val perplexity 139.9499
160
+ 1300 train 4.902924 (lr=1.3818e-03) (hash(x)=77545187)
161
+ 1310 train 4.910340 (lr=1.3778e-03) (hash(x)=86412685)
162
+ 1320 train 4.768957 (lr=1.3739e-03) (hash(x)=86429640)
163
+ 1330 train 4.979923 (lr=1.3699e-03) (hash(x)=79645281)
164
+ 1340 train 4.924356 (lr=1.3658e-03) (hash(x)=82668541)
165
+ 1350 train 4.724071 (lr=1.3618e-03) (hash(x)=73627248)
166
+ 1360 train 4.803435 (lr=1.3577e-03) (hash(x)=74845133)
167
+ 1370 train 4.742704 (lr=1.3536e-03) (hash(x)=77850497)
168
+ 1380 train 4.846462 (lr=1.3494e-03) (hash(x)=76143954)
169
+ 1390 train 4.817306 (lr=1.3453e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.9686
171
+ 1400 val perplexity 143.8291
172
+ 1400 train 4.891697 (lr=1.3411e-03) (hash(x)=76010938)
173
+ 1410 train 4.925680 (lr=1.3368e-03) (hash(x)=81260300)
174
+ 1420 train 5.089675 (lr=1.3326e-03) (hash(x)=83803629)
175
+ 1430 train 4.920062 (lr=1.3283e-03) (hash(x)=83840110)
176
+ 1440 train 5.129083 (lr=1.3240e-03) (hash(x)=97711831)
177
+ 1450 train 5.126706 (lr=1.3197e-03) (hash(x)=81110139)
178
+ 1460 train 5.085930 (lr=1.3153e-03) (hash(x)=90942839)
179
+ 1470 train 5.244080 (lr=1.3109e-03) (hash(x)=80721564)
180
+ 1480 train 5.259272 (lr=1.3065e-03) (hash(x)=77852059)
181
+ 1490 train 5.143749 (lr=1.3021e-03) (hash(x)=80011365)
182
+ 1500 val loss 5.0854
183
+ 1500 val perplexity 161.6505
184
+ 1500 train 5.062841 (lr=1.2976e-03) (hash(x)=84496142)
185
+ 1510 train 4.976560 (lr=1.2932e-03) (hash(x)=70186729)
186
+ 1520 train 4.965950 (lr=1.2887e-03) (hash(x)=74854227)
187
+ 1530 train 4.917572 (lr=1.2841e-03) (hash(x)=77468161)
188
+ 1540 train 5.089735 (lr=1.2796e-03) (hash(x)=87853059)
189
+ 1550 train 4.964570 (lr=1.2750e-03) (hash(x)=71225436)
190
+ 1560 train 5.227436 (lr=1.2704e-03) (hash(x)=84072783)
191
+ 1570 train 5.272835 (lr=1.2658e-03) (hash(x)=72723098)
192
+ 1580 train 5.195763 (lr=1.2612e-03) (hash(x)=90409866)
193
+ 1590 train 5.448853 (lr=1.2565e-03) (hash(x)=87481378)
194
+ 1600 val loss 5.2051
195
+ 1600 val perplexity 182.2057
196
+ 1600 train 5.248561 (lr=1.2518e-03) (hash(x)=77643862)
197
+ 1610 train 5.362939 (lr=1.2471e-03) (hash(x)=88862575)
198
+ 1620 train 5.495556 (lr=1.2424e-03) (hash(x)=84612581)
199
+ 1630 train 5.236500 (lr=1.2376e-03) (hash(x)=87075989)
200
+ 1640 train 5.054854 (lr=1.2329e-03) (hash(x)=88277361)
201
+ 1650 train 5.103882 (lr=1.2281e-03) (hash(x)=78750236)
202
+ 1660 train 5.009445 (lr=1.2233e-03) (hash(x)=82604581)
203
+ 1670 train 4.902416 (lr=1.2184e-03) (hash(x)=68482265)
204
+ 1680 train 4.951748 (lr=1.2136e-03) (hash(x)=75088835)
205
+ 1690 train 4.847054 (lr=1.2087e-03) (hash(x)=66060989)
206
+ 1700 val loss 5.0148
207
+ 1700 val perplexity 150.6326
208
+ 1700 train 5.003978 (lr=1.2038e-03) (hash(x)=79986754)
209
+ 1710 train 4.999313 (lr=1.1989e-03) (hash(x)=83657930)
210
+ 1720 train 4.953660 (lr=1.1940e-03) (hash(x)=81754135)
211
+ 1730 train 4.977748 (lr=1.1890e-03) (hash(x)=78904427)
212
+ 1740 train 5.002439 (lr=1.1841e-03) (hash(x)=85920177)
213
+ 1750 train 5.095030 (lr=1.1791e-03) (hash(x)=86573211)
214
+ 1760 train 4.953336 (lr=1.1741e-03) (hash(x)=81737128)
215
+ 1770 train 4.917145 (lr=1.1691e-03) (hash(x)=77400968)
216
+ 1780 train 4.988267 (lr=1.1641e-03) (hash(x)=73545497)
217
+ 1790 train 5.172650 (lr=1.1590e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.9545
219
+ 1800 val perplexity 141.8129
220
+ 1800 train 4.987381 (lr=1.1539e-03) (hash(x)=87819781)
221
+ 1810 train 4.896414 (lr=1.1489e-03) (hash(x)=86870770)
222
+ 1820 train 4.849457 (lr=1.1438e-03) (hash(x)=82522211)
223
+ 1830 train 5.021136 (lr=1.1387e-03) (hash(x)=61947437)
224
+ 1840 train 4.912839 (lr=1.1335e-03) (hash(x)=79865406)
225
+ 1850 train 4.788253 (lr=1.1284e-03) (hash(x)=79828721)
226
+ 1860 train 4.926324 (lr=1.1232e-03) (hash(x)=80869571)
227
+ 1870 train 4.916054 (lr=1.1181e-03) (hash(x)=73780971)
228
+ 1880 train 4.866315 (lr=1.1129e-03) (hash(x)=79249549)
229
+ 1890 train 4.914686 (lr=1.1077e-03) (hash(x)=81041904)
230
+ 1900 val loss 5.0041
231
+ 1900 val perplexity 149.0259
232
+ 1900 train 5.132901 (lr=1.1025e-03) (hash(x)=82456430)
233
+ 1910 train 4.823038 (lr=1.0973e-03) (hash(x)=82222135)
234
+ 1920 train 4.826719 (lr=1.0920e-03) (hash(x)=70033249)
235
+ 1930 train 4.874485 (lr=1.0868e-03) (hash(x)=72887360)
236
+ 1940 train 4.905589 (lr=1.0815e-03) (hash(x)=83251100)
237
+ 1950 train 4.941963 (lr=1.0763e-03) (hash(x)=79660266)
238
+ 1960 train 4.838334 (lr=1.0710e-03) (hash(x)=78919068)
239
+ 1970 train 4.945966 (lr=1.0657e-03) (hash(x)=94690431)
240
+ 1980 train 4.835110 (lr=1.0604e-03) (hash(x)=88153756)
241
+ 1990 train 4.854059 (lr=1.0551e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.9645
243
+ 2000 val perplexity 143.2416
244
+ 2000 train 4.949703 (lr=1.0497e-03) (hash(x)=81308591)
245
+ 2010 train 5.138031 (lr=1.0444e-03) (hash(x)=80205479)
246
+ 2020 train 4.984600 (lr=1.0391e-03) (hash(x)=81008704)
247
+ 2030 train 5.063983 (lr=1.0337e-03) (hash(x)=82914358)
248
+ 2040 train 4.998124 (lr=1.0284e-03) (hash(x)=88064399)
249
+ 2050 train 4.994970 (lr=1.0230e-03) (hash(x)=82889390)
250
+ 2060 train 4.974585 (lr=1.0176e-03) (hash(x)=80913554)
251
+ 2070 train 4.861185 (lr=1.0122e-03) (hash(x)=78121791)
252
+ 2080 train 5.049816 (lr=1.0068e-03) (hash(x)=75356657)
253
+ 2090 train 4.819481 (lr=1.0014e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.9400
255
+ 2100 val perplexity 139.7712
256
+ 2100 train 4.770241 (lr=9.9602e-04) (hash(x)=68928225)
257
+ 2110 train 4.802988 (lr=9.9060e-04) (hash(x)=93734745)
258
+ 2120 train 4.866114 (lr=9.8518e-04) (hash(x)=72878164)
259
+ 2130 train 5.008235 (lr=9.7975e-04) (hash(x)=71508226)
260
+ 2140 train 4.899097 (lr=9.7432e-04) (hash(x)=84168671)
261
+ 2150 train 4.843262 (lr=9.6888e-04) (hash(x)=74673639)
262
+ 2160 train 4.998524 (lr=9.6344e-04) (hash(x)=75470031)
263
+ 2170 train 4.979599 (lr=9.5799e-04) (hash(x)=74307890)
264
+ 2180 train 5.014943 (lr=9.5253e-04) (hash(x)=77214245)
265
+ 2190 train 4.856992 (lr=9.4708e-04) (hash(x)=88628359)
266
+ 2200 val loss 5.0871
267
+ 2200 val perplexity 161.9215
268
+ 2200 train 4.994662 (lr=9.4161e-04) (hash(x)=74779126)
269
+ 2210 train 4.982049 (lr=9.3615e-04) (hash(x)=79817976)
270
+ 2220 train 4.853894 (lr=9.3068e-04) (hash(x)=75258996)
271
+ 2230 train 4.870675 (lr=9.2521e-04) (hash(x)=88164047)
272
+ 2240 train 4.870108 (lr=9.1974e-04) (hash(x)=78580686)
273
+ 2250 train 4.916204 (lr=9.1426e-04) (hash(x)=76510617)
274
+ 2260 train 4.909360 (lr=9.0878e-04) (hash(x)=76043966)
275
+ 2270 train 4.868372 (lr=9.0330e-04) (hash(x)=73062098)
276
+ 2280 train 5.070706 (lr=8.9782e-04) (hash(x)=81885909)
277
+ 2290 train 4.994043 (lr=8.9234e-04) (hash(x)=81481741)
278
+ 2300 val loss 5.0012
279
+ 2300 val perplexity 148.5848
280
+ 2300 train 4.925090 (lr=8.8685e-04) (hash(x)=82104275)
281
+ 2310 train 4.976020 (lr=8.8137e-04) (hash(x)=82137309)
282
+ 2320 train 5.011151 (lr=8.7589e-04) (hash(x)=86741167)
283
+ 2330 train 5.017084 (lr=8.7040e-04) (hash(x)=84796263)
284
+ 2340 train 4.915669 (lr=8.6492e-04) (hash(x)=169966529)
285
+ 2350 train 4.765397 (lr=8.5944e-04) (hash(x)=80006061)
286
+ 2360 train 4.905653 (lr=8.5396e-04) (hash(x)=71090999)
287
+ 2370 train 5.001131 (lr=8.4848e-04) (hash(x)=86641850)
288
+ 2380 train 4.904800 (lr=8.4300e-04) (hash(x)=84573253)
289
+ 2390 train 5.014882 (lr=8.3753e-04) (hash(x)=74483764)
290
+ 2400 val loss 5.0810
291
+ 2400 val perplexity 160.9333
292
+ 2400 train 5.002321 (lr=8.3205e-04) (hash(x)=78327659)
293
+ 2410 train 5.038434 (lr=8.2658e-04) (hash(x)=77273627)
294
+ 2420 train 4.921847 (lr=8.2112e-04) (hash(x)=76938049)
295
+ 2430 train 5.038579 (lr=8.1565e-04) (hash(x)=98449442)
296
+ 2440 train 4.883294 (lr=8.1020e-04) (hash(x)=78157797)
297
+ 2450 train 5.032661 (lr=8.0474e-04) (hash(x)=80637582)
298
+ 2460 train 5.121312 (lr=7.9929e-04) (hash(x)=80023854)
299
+ 2470 train 5.132766 (lr=7.9384e-04) (hash(x)=86678884)
300
+ 2480 train 5.308314 (lr=7.8840e-04) (hash(x)=84086469)
301
+ 2490 train 5.312792 (lr=7.8296e-04) (hash(x)=83131141)
302
+ 2500 val loss 5.2412
303
+ 2500 val perplexity 188.8961
304
+ 2500 train 5.226176 (lr=7.7753e-04) (hash(x)=82583497)
305
+ 2510 train 5.240191 (lr=7.7211e-04) (hash(x)=79164326)
306
+ 2520 train 5.102191 (lr=7.6669e-04) (hash(x)=76456503)
307
+ 2530 train 5.142907 (lr=7.6128e-04) (hash(x)=82001228)
308
+ 2540 train 5.238275 (lr=7.5587e-04) (hash(x)=75615595)
309
+ 2550 train 5.243271 (lr=7.5048e-04) (hash(x)=78397869)
310
+ 2560 train 5.119880 (lr=7.4509e-04) (hash(x)=75547032)
311
+ 2570 train 5.095049 (lr=7.3970e-04) (hash(x)=89201025)
312
+ 2580 train 5.045280 (lr=7.3433e-04) (hash(x)=89856704)
313
+ 2590 train 5.037935 (lr=7.2896e-04) (hash(x)=82175682)
314
+ 2600 val loss 5.1597
315
+ 2600 val perplexity 174.1126
316
+ 2600 train 5.048204 (lr=7.2361e-04) (hash(x)=78312826)
317
+ 2610 train 5.067132 (lr=7.1826e-04) (hash(x)=77066588)
318
+ 2620 train 4.977514 (lr=7.1292e-04) (hash(x)=78666061)
319
+ 2630 train 4.965766 (lr=7.0759e-04) (hash(x)=93762143)
320
+ 2640 train 4.940514 (lr=7.0227e-04) (hash(x)=83191587)
321
+ 2650 train 4.924569 (lr=6.9696e-04) (hash(x)=87169585)
322
+ 2660 train 5.178645 (lr=6.9166e-04) (hash(x)=86426388)
323
+ 2670 train 5.249987 (lr=6.8638e-04) (hash(x)=76692638)
324
+ 2680 train 5.177525 (lr=6.8110e-04) (hash(x)=77446063)
325
+ 2690 train 5.251211 (lr=6.7584e-04) (hash(x)=79809050)
326
+ 2700 val loss 5.2251
327
+ 2700 val perplexity 185.8853
328
+ 2700 train 5.121318 (lr=6.7059e-04) (hash(x)=83116823)
329
+ 2710 train 5.212320 (lr=6.6534e-04) (hash(x)=75622148)
330
+ 2720 train 5.092994 (lr=6.6012e-04) (hash(x)=80690512)
331
+ 2730 train 5.165248 (lr=6.5490e-04) (hash(x)=78009984)
332
+ 2740 train 5.108691 (lr=6.4970e-04) (hash(x)=74205488)
333
+ 2750 train 5.072241 (lr=6.4451e-04) (hash(x)=91013332)
334
+ 2760 train 5.026791 (lr=6.3934e-04) (hash(x)=76954961)
335
+ 2770 train 5.037196 (lr=6.3417e-04) (hash(x)=79390317)
336
+ 2780 train 4.858578 (lr=6.2903e-04) (hash(x)=70168783)
337
+ 2790 train 5.030842 (lr=6.2390e-04) (hash(x)=76028417)
338
+ 2800 val loss 5.1289
339
+ 2800 val perplexity 168.8340
340
+ 2800 train 4.917107 (lr=6.1878e-04) (hash(x)=77656050)
341
+ 2810 train 4.991508 (lr=6.1368e-04) (hash(x)=94248216)
342
+ 2820 train 5.043240 (lr=6.0859e-04) (hash(x)=78305078)
343
+ 2830 train 4.993572 (lr=6.0352e-04) (hash(x)=79948848)
344
+ 2840 train 4.946194 (lr=5.9846e-04) (hash(x)=85341024)
345
+ 2850 train 5.175757 (lr=5.9342e-04) (hash(x)=78735170)
346
+ 2860 train 5.416435 (lr=5.8840e-04) (hash(x)=71616419)
347
+ 2870 train 5.227557 (lr=5.8340e-04) (hash(x)=78656517)
348
+ 2880 train 5.193727 (lr=5.7841e-04) (hash(x)=80073987)
349
+ 2890 train 5.086908 (lr=5.7344e-04) (hash(x)=76894809)
350
+ 2900 val loss 5.1505
351
+ 2900 val perplexity 172.5213
352
+ 2900 train 5.081022 (lr=5.6849e-04) (hash(x)=80499838)
353
+ 2910 train 5.065008 (lr=5.6355e-04) (hash(x)=72673354)
354
+ 2920 train 5.023953 (lr=5.5864e-04) (hash(x)=84265768)
355
+ 2930 train 5.059848 (lr=5.5374e-04) (hash(x)=79612060)
356
+ 2940 train 4.919847 (lr=5.4886e-04) (hash(x)=74970087)
357
+ 2950 train 5.094071 (lr=5.4400e-04) (hash(x)=84166818)
358
+ 2960 train 5.084225 (lr=5.3916e-04) (hash(x)=89410221)
359
+ 2970 train 5.117095 (lr=5.3434e-04) (hash(x)=75672566)
360
+ 2980 train 4.959161 (lr=5.2954e-04) (hash(x)=81760314)
361
+ 2990 train 5.048182 (lr=5.2476e-04) (hash(x)=80605200)
362
+ 3000 val loss 5.2789
363
+ 3000 val perplexity 196.1458
364
+ 3000 train 5.095525 (lr=5.2000e-04) (hash(x)=83804735)
365
+ 3010 train 4.888479 (lr=5.1526e-04) (hash(x)=77015303)
366
+ 3020 train 4.970644 (lr=5.1054e-04) (hash(x)=81464523)
367
+ 3030 train 5.138883 (lr=5.0585e-04) (hash(x)=80168230)
368
+ 3040 train 5.052806 (lr=5.0117e-04) (hash(x)=84817006)
369
+ 3050 train 5.078069 (lr=4.9652e-04) (hash(x)=71601811)
370
+ 3060 train 4.996464 (lr=4.9189e-04) (hash(x)=85499733)
371
+ 3070 train 5.083865 (lr=4.8729e-04) (hash(x)=78670408)
372
+ 3080 train 5.007542 (lr=4.8270e-04) (hash(x)=77120468)
373
+ 3090 train 5.072447 (lr=4.7814e-04) (hash(x)=77927426)
374
+ 3100 val loss 5.0787
375
+ 3100 val perplexity 160.5659
376
+ 3100 train 5.052860 (lr=4.7360e-04) (hash(x)=83998606)
377
+ 3110 train 4.938472 (lr=4.6909e-04) (hash(x)=77990218)
378
+ 3120 train 4.964771 (lr=4.6460e-04) (hash(x)=81623970)
379
+ 3130 train 4.929324 (lr=4.6013e-04) (hash(x)=75014781)
380
+ 3140 train 4.946321 (lr=4.5569e-04) (hash(x)=72591250)
381
+ 3150 train 5.020030 (lr=4.5127e-04) (hash(x)=81421847)
382
+ 3160 train 4.985006 (lr=4.4688e-04) (hash(x)=75112631)
383
+ 3170 train 5.136054 (lr=4.4251e-04) (hash(x)=87518033)
384
+ 3180 train 4.958146 (lr=4.3817e-04) (hash(x)=78648348)
385
+ 3190 train 5.047131 (lr=4.3385e-04) (hash(x)=78270029)
386
+ 3200 val loss 5.1084
387
+ 3200 val perplexity 165.4026
388
+ 3200 train 5.195908 (lr=4.2956e-04) (hash(x)=83380714)
389
+ 3210 train 5.260527 (lr=4.2529e-04) (hash(x)=67804991)
390
+ 3220 train 5.077404 (lr=4.2105e-04) (hash(x)=76968804)
391
+ 3230 train 5.064898 (lr=4.1684e-04) (hash(x)=75798670)
392
+ 3240 train 5.108447 (lr=4.1266e-04) (hash(x)=83192811)
393
+ 3250 train 5.024596 (lr=4.0850e-04) (hash(x)=81659789)
394
+ 3260 train 5.032725 (lr=4.0437e-04) (hash(x)=80622502)
395
+ 3270 train 5.080941 (lr=4.0027e-04) (hash(x)=85436511)
396
+ 3280 train 4.964487 (lr=3.9619e-04) (hash(x)=80589180)
397
+ 3290 train 5.120704 (lr=3.9214e-04) (hash(x)=85747193)
398
+ 3300 val loss 5.1377
399
+ 3300 val perplexity 170.3293
400
+ 3300 train 5.039641 (lr=3.8813e-04) (hash(x)=74635692)
401
+ 3310 train 5.037121 (lr=3.8414e-04) (hash(x)=82045455)
402
+ 3320 train 4.919912 (lr=3.8017e-04) (hash(x)=73249173)
403
+ 3330 train 4.857579 (lr=3.7624e-04) (hash(x)=71492338)
404
+ 3340 train 5.128398 (lr=3.7234e-04) (hash(x)=76374071)
405
+ 3350 train 5.116617 (lr=3.6847e-04) (hash(x)=80474064)
406
+ 3360 train 5.052750 (lr=3.6462e-04) (hash(x)=84390892)
407
+ 3370 train 5.155038 (lr=3.6081e-04) (hash(x)=83399949)
408
+ 3380 train 5.177409 (lr=3.5702e-04) (hash(x)=77648059)
409
+ 3390 train 5.149597 (lr=3.5327e-04) (hash(x)=80152701)
410
+ 3400 val loss 5.1255
411
+ 3400 val perplexity 168.2625
412
+ 3400 train 5.167397 (lr=3.4955e-04) (hash(x)=81952545)
413
+ 3410 train 5.134211 (lr=3.4585e-04) (hash(x)=80908993)
414
+ 3420 train 5.155999 (lr=3.4219e-04) (hash(x)=79610037)
415
+ 3430 train 5.113538 (lr=3.3856e-04) (hash(x)=87624382)
416
+ 3440 train 5.099593 (lr=3.3496e-04) (hash(x)=82336381)
417
+ 3450 train 5.080705 (lr=3.3140e-04) (hash(x)=87687835)
418
+ 3460 train 5.098285 (lr=3.2786e-04) (hash(x)=76376135)
419
+ 3470 train 4.995934 (lr=3.2436e-04) (hash(x)=79331391)
420
+ 3480 train 4.951308 (lr=3.2089e-04) (hash(x)=107002681)
421
+ 3490 train 4.996736 (lr=3.1745e-04) (hash(x)=77465514)
422
+ 3500 val loss 5.1570
423
+ 3500 val perplexity 173.6477
424
+ 3500 train 5.037126 (lr=3.1404e-04) (hash(x)=88237229)
425
+ 3510 train 5.088996 (lr=3.1067e-04) (hash(x)=81011739)
426
+ 3520 train 5.017595 (lr=3.0733e-04) (hash(x)=64643427)
427
+ 3530 train 5.262295 (lr=3.0402e-04) (hash(x)=78029539)
428
+ 3540 train 5.235551 (lr=3.0075e-04) (hash(x)=83188968)
429
+ 3550 train 5.238494 (lr=2.9751e-04) (hash(x)=86104185)
430
+ 3560 train 5.183664 (lr=2.9430e-04) (hash(x)=81993629)
431
+ 3570 train 5.103960 (lr=2.9113e-04) (hash(x)=72719368)
432
+ 3580 train 5.135702 (lr=2.8799e-04) (hash(x)=90091487)
433
+ 3590 train 4.994220 (lr=2.8489e-04) (hash(x)=77393152)
434
+ 3600 val loss 5.0959
435
+ 3600 val perplexity 163.3433
436
+ 3600 train 5.076273 (lr=2.8182e-04) (hash(x)=73103504)
437
+ 3610 train 5.094037 (lr=2.7878e-04) (hash(x)=84429400)
438
+ 3620 train 5.155484 (lr=2.7578e-04) (hash(x)=77021795)
439
+ 3630 train 5.066847 (lr=2.7282e-04) (hash(x)=92830605)
440
+ 3640 train 5.239064 (lr=2.6989e-04) (hash(x)=78313175)
441
+ 3650 train 5.033025 (lr=2.6700e-04) (hash(x)=85395549)
442
+ 3660 train 4.860571 (lr=2.6414e-04) (hash(x)=78114459)
443
+ 3670 train 4.947598 (lr=2.6132e-04) (hash(x)=74968316)
444
+ 3680 train 4.878042 (lr=2.5853e-04) (hash(x)=73358737)
445
+ 3690 train 5.058198 (lr=2.5578e-04) (hash(x)=76399442)
446
+ 3700 val loss 5.0955
447
+ 3700 val perplexity 163.2838
448
+ 3700 train 4.966326 (lr=2.5306e-04) (hash(x)=74039273)
449
+ 3710 train 5.167697 (lr=2.5038e-04) (hash(x)=79299680)
450
+ 3720 train 5.077923 (lr=2.4774e-04) (hash(x)=94701498)
451
+ 3730 train 5.129753 (lr=2.4514e-04) (hash(x)=75352071)
452
+ 3740 train 5.040686 (lr=2.4257e-04) (hash(x)=77909487)
453
+ 3750 train 5.087613 (lr=2.4004e-04) (hash(x)=73986730)
454
+ 3760 train 5.085809 (lr=2.3754e-04) (hash(x)=79325763)
455
+ 3770 train 5.127954 (lr=2.3509e-04) (hash(x)=72457818)
456
+ 3780 train 5.273745 (lr=2.3267e-04) (hash(x)=71775590)
457
+ 3790 train 5.187685 (lr=2.3029e-04) (hash(x)=82638943)
458
+ 3800 val loss 5.1196
459
+ 3800 val perplexity 167.2635
460
+ 3800 train 5.118042 (lr=2.2794e-04) (hash(x)=79965893)
461
+ 3810 train 5.013916 (lr=2.2564e-04) (hash(x)=72598235)
462
+ 3820 train 5.028393 (lr=2.2337e-04) (hash(x)=83113889)
463
+ 3830 train 5.099103 (lr=2.2114e-04) (hash(x)=74434590)
464
+ 3840 train 5.043058 (lr=2.1894e-04) (hash(x)=82860348)
465
+ 3850 train 4.922311 (lr=2.1679e-04) (hash(x)=78067565)
466
+ 3860 train 4.982595 (lr=2.1468e-04) (hash(x)=82592498)
467
+ 3870 train 4.939715 (lr=2.1260e-04) (hash(x)=81820733)
468
+ 3880 train 4.981380 (lr=2.1056e-04) (hash(x)=87709040)
469
+ 3890 train 5.103719 (lr=2.0856e-04) (hash(x)=70379093)
470
+ 3900 val loss 5.0707
471
+ 3900 val perplexity 159.2844
472
+ 3900 train 5.135643 (lr=2.0660e-04) (hash(x)=76597431)
473
+ 3910 train 5.047150 (lr=2.0468e-04) (hash(x)=90490716)
474
+ 3920 train 5.100316 (lr=2.0280e-04) (hash(x)=81970659)
475
+ 3930 train 5.081352 (lr=2.0096e-04) (hash(x)=81496334)
476
+ 3940 train 5.018997 (lr=1.9915e-04) (hash(x)=75717605)
477
+ 3950 train 5.046584 (lr=1.9739e-04) (hash(x)=82749357)
478
+ 3960 train 5.071330 (lr=1.9566e-04) (hash(x)=89355157)
479
+ 3970 train 4.998615 (lr=1.9398e-04) (hash(x)=78980403)
480
+ 3980 train 4.989475 (lr=1.9233e-04) (hash(x)=76627217)
481
+ 3990 train 4.948875 (lr=1.9073e-04) (hash(x)=72412879)
482
+ 4000 val loss 5.0257
483
+ 4000 val perplexity 152.2695
484
+ 4000 train 4.967622 (lr=1.8917e-04) (hash(x)=83018142)
485
+ 4010 train 4.973487 (lr=1.8764e-04) (hash(x)=81272436)
486
+ 4020 train 4.927087 (lr=1.8616e-04) (hash(x)=85497482)
487
+ 4030 train 5.046237 (lr=1.8471e-04) (hash(x)=85241734)
488
+ 4040 train 4.995852 (lr=1.8331e-04) (hash(x)=77925307)
489
+ 4050 train 4.964125 (lr=1.8194e-04) (hash(x)=84826179)
490
+ 4060 train 5.035626 (lr=1.8062e-04) (hash(x)=83606764)
491
+ 4070 train 5.033113 (lr=1.7934e-04) (hash(x)=80567590)
492
+ 4080 train 5.001812 (lr=1.7810e-04) (hash(x)=76860998)
493
+ 4090 train 5.020388 (lr=1.7689e-04) (hash(x)=74902328)
494
+ 4100 val loss 5.0432
495
+ 4100 val perplexity 154.9714
496
+ 4100 train 5.058005 (lr=1.7573e-04) (hash(x)=82832041)
497
+ 4110 train 4.987716 (lr=1.7461e-04) (hash(x)=79143262)
498
+ 4120 train 4.970866 (lr=1.7354e-04) (hash(x)=77038149)
499
+ 4130 train 5.171425 (lr=1.7250e-04) (hash(x)=86339074)
500
+ 4140 train 5.169609 (lr=1.7150e-04) (hash(x)=76686216)
501
+ 4150 train 5.001726 (lr=1.7055e-04) (hash(x)=70522682)
502
+ 4160 train 4.984271 (lr=1.6963e-04) (hash(x)=90958555)
503
+ 4170 train 5.057621 (lr=1.6876e-04) (hash(x)=91463532)
504
+ 4180 train 5.011285 (lr=1.6793e-04) (hash(x)=81959329)
505
+ 4190 train 4.960071 (lr=1.6713e-04) (hash(x)=83146752)
506
+ 4200 val loss 5.0336
507
+ 4200 val perplexity 153.4902
508
+ 4200 train 4.937669 (lr=1.6639e-04) (hash(x)=78361715)
509
+ 4210 train 5.092117 (lr=1.6568e-04) (hash(x)=87364889)
510
+ 4220 train 4.888556 (lr=1.6501e-04) (hash(x)=70465156)
511
+ 4230 train 5.053319 (lr=1.6439e-04) (hash(x)=84524081)
512
+ 4240 train 5.027927 (lr=1.6380e-04) (hash(x)=77824868)
513
+ 4250 train 5.128632 (lr=1.6326e-04) (hash(x)=81710711)
514
+ 4260 train 5.044909 (lr=1.6276e-04) (hash(x)=76362728)
515
+ 4270 train 4.991559 (lr=1.6230e-04) (hash(x)=83115208)
516
+ 4280 train 4.903317 (lr=1.6188e-04) (hash(x)=87218314)
517
+ 4290 train 5.043246 (lr=1.6151e-04) (hash(x)=74582673)
518
+ 4300 val loss 5.0066
519
+ 4300 val perplexity 149.3924
520
+ 4300 train 4.963599 (lr=1.6117e-04) (hash(x)=77379615)
521
+ 4310 train 4.894029 (lr=1.6088e-04) (hash(x)=78669579)
522
+ 4320 train 5.032390 (lr=1.6063e-04) (hash(x)=83066608)
523
+ 4330 train 5.037401 (lr=1.6042e-04) (hash(x)=83037340)
524
+ 4340 train 5.090609 (lr=1.6026e-04) (hash(x)=82849771)
525
+ 4350 train 4.978410 (lr=1.6013e-04) (hash(x)=76693985)
526
+ 4360 train 5.035444 (lr=1.6005e-04) (hash(x)=77745394)
527
+ 4370 train 5.052416 (lr=1.6001e-04) (hash(x)=79954388)
528
+ 4374 val loss 5.0415
529
+ 4374 val perplexity 154.7094
12_head_one_mask_per_head_4_latent_masks/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4fe0b756bd75b9b2c2a088f6d1bae431b200b23c8558ae1edb49178d4b688ced
3
+ size 505705586
12_head_one_mask_per_head_4_latent_masks/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e87929bb257953d567b3f09b309b84b5e1949f7e6a9f1ad8bc79ed7529b5302b
3
+ size 1005133878