andrew-healey commited on
Commit
7395c41
·
verified ·
1 Parent(s): a22bc25

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_16e-4_head_dim_22_fixed/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_head_baseline_lr_16e-4_head_dim_22_fixed", "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": "none", "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": 0.0016, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_16e-4_fixed"}
12_head_baseline_lr_16e-4_head_dim_22_fixed/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_baseline_lr_16e-4_head_dim_22_fixed/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.4043
3
+ 0 val perplexity 89706.9531
4
+ 0 train 11.388921 (lr=2.2378e-06) (hash(x)=93356070)
5
+ 10 train 10.225544 (lr=2.4615e-05) (hash(x)=91253010)
6
+ 20 train 9.771616 (lr=4.6993e-05) (hash(x)=74934453)
7
+ 30 train 9.346413 (lr=6.9371e-05) (hash(x)=79278034)
8
+ 40 train 8.733868 (lr=9.1748e-05) (hash(x)=80107892)
9
+ 50 train 8.191336 (lr=1.1413e-04) (hash(x)=70745428)
10
+ 60 train 7.902347 (lr=1.3650e-04) (hash(x)=80077589)
11
+ 70 train 7.554003 (lr=1.5888e-04) (hash(x)=76213766)
12
+ 80 train 7.569885 (lr=1.8126e-04) (hash(x)=83218328)
13
+ 90 train 7.409823 (lr=2.0364e-04) (hash(x)=74855845)
14
+ 100 val loss 7.3216
15
+ 100 val perplexity 1512.6713
16
+ 100 train 7.168478 (lr=2.2601e-04) (hash(x)=82814902)
17
+ 110 train 7.130021 (lr=2.4839e-04) (hash(x)=84907741)
18
+ 120 train 6.979678 (lr=2.7077e-04) (hash(x)=82613223)
19
+ 130 train 6.815266 (lr=2.9315e-04) (hash(x)=83540876)
20
+ 140 train 6.763745 (lr=3.1552e-04) (hash(x)=75095216)
21
+ 150 train 6.819985 (lr=3.3790e-04) (hash(x)=97190944)
22
+ 160 train 6.718470 (lr=3.6028e-04) (hash(x)=82117809)
23
+ 170 train 6.621368 (lr=3.8266e-04) (hash(x)=70514724)
24
+ 180 train 6.541959 (lr=4.0503e-04) (hash(x)=81029624)
25
+ 190 train 6.434331 (lr=4.2741e-04) (hash(x)=78996841)
26
+ 200 val loss 6.4002
27
+ 200 val perplexity 601.9530
28
+ 200 train 6.401682 (lr=4.4979e-04) (hash(x)=79845097)
29
+ 210 train 6.270069 (lr=4.7217e-04) (hash(x)=78997683)
30
+ 220 train 6.197242 (lr=4.9455e-04) (hash(x)=74895865)
31
+ 230 train 6.327465 (lr=5.1692e-04) (hash(x)=80933276)
32
+ 240 train 6.023209 (lr=5.3930e-04) (hash(x)=77664606)
33
+ 250 train 6.072259 (lr=5.6168e-04) (hash(x)=75903930)
34
+ 260 train 5.963767 (lr=5.8406e-04) (hash(x)=82996853)
35
+ 270 train 5.892630 (lr=6.0643e-04) (hash(x)=73269568)
36
+ 280 train 5.828564 (lr=6.2881e-04) (hash(x)=82906427)
37
+ 290 train 5.754286 (lr=6.5119e-04) (hash(x)=71797895)
38
+ 300 val loss 5.9032
39
+ 300 val perplexity 366.2070
40
+ 300 train 5.618736 (lr=6.7357e-04) (hash(x)=85232249)
41
+ 310 train 5.563621 (lr=6.9594e-04) (hash(x)=75252489)
42
+ 320 train 5.639262 (lr=7.1832e-04) (hash(x)=74394644)
43
+ 330 train 5.714954 (lr=7.4070e-04) (hash(x)=82772910)
44
+ 340 train 5.768359 (lr=7.6308e-04) (hash(x)=81627464)
45
+ 350 train 5.746442 (lr=7.8545e-04) (hash(x)=84678053)
46
+ 360 train 5.692416 (lr=8.0783e-04) (hash(x)=81884128)
47
+ 370 train 5.691870 (lr=8.3021e-04) (hash(x)=76379242)
48
+ 380 train 5.500958 (lr=8.5259e-04) (hash(x)=98182875)
49
+ 390 train 5.479429 (lr=8.7497e-04) (hash(x)=79710436)
50
+ 400 val loss 5.5305
51
+ 400 val perplexity 252.2672
52
+ 400 train 5.480612 (lr=8.9734e-04) (hash(x)=79841071)
53
+ 410 train 5.414336 (lr=9.1972e-04) (hash(x)=75844151)
54
+ 420 train 5.443130 (lr=9.4210e-04) (hash(x)=73125036)
55
+ 430 train 5.449091 (lr=9.6448e-04) (hash(x)=84214858)
56
+ 440 train 5.343349 (lr=9.8685e-04) (hash(x)=80456994)
57
+ 450 train 5.323442 (lr=1.0092e-03) (hash(x)=76735962)
58
+ 460 train 5.248092 (lr=1.0316e-03) (hash(x)=81845446)
59
+ 470 train 5.064189 (lr=1.0540e-03) (hash(x)=76094689)
60
+ 480 train 5.185876 (lr=1.0764e-03) (hash(x)=83806686)
61
+ 490 train 4.992406 (lr=1.0987e-03) (hash(x)=84690227)
62
+ 500 val loss 5.3241
63
+ 500 val perplexity 205.2183
64
+ 500 train 5.071612 (lr=1.1211e-03) (hash(x)=71851938)
65
+ 510 train 5.168808 (lr=1.1435e-03) (hash(x)=77159346)
66
+ 520 train 5.236441 (lr=1.1659e-03) (hash(x)=80755753)
67
+ 530 train 5.333798 (lr=1.1883e-03) (hash(x)=75379679)
68
+ 540 train 5.249462 (lr=1.2106e-03) (hash(x)=82458619)
69
+ 550 train 5.227415 (lr=1.2330e-03) (hash(x)=86366396)
70
+ 560 train 5.243289 (lr=1.2554e-03) (hash(x)=85095044)
71
+ 570 train 5.201520 (lr=1.2778e-03) (hash(x)=78385159)
72
+ 580 train 5.150285 (lr=1.3001e-03) (hash(x)=79342394)
73
+ 590 train 5.142563 (lr=1.3225e-03) (hash(x)=70782192)
74
+ 600 val loss 5.1417
75
+ 600 val perplexity 171.0023
76
+ 600 train 5.119190 (lr=1.3449e-03) (hash(x)=76130353)
77
+ 610 train 5.041281 (lr=1.3673e-03) (hash(x)=74778440)
78
+ 620 train 5.027589 (lr=1.3897e-03) (hash(x)=79129709)
79
+ 630 train 4.917737 (lr=1.4120e-03) (hash(x)=76469962)
80
+ 640 train 4.910306 (lr=1.4344e-03) (hash(x)=78288049)
81
+ 650 train 4.868297 (lr=1.4568e-03) (hash(x)=76641388)
82
+ 660 train 4.813710 (lr=1.4792e-03) (hash(x)=75937906)
83
+ 670 train 4.817692 (lr=1.5015e-03) (hash(x)=74807157)
84
+ 680 train 4.693060 (lr=1.5239e-03) (hash(x)=77490144)
85
+ 690 train 4.725158 (lr=1.5463e-03) (hash(x)=72900124)
86
+ 700 val loss 4.9833
87
+ 700 val perplexity 145.9506
88
+ 700 train 4.678411 (lr=1.5687e-03) (hash(x)=73424218)
89
+ 710 train 4.971233 (lr=1.5910e-03) (hash(x)=86052345)
90
+ 720 train 4.937612 (lr=1.6000e-03) (hash(x)=77462613)
91
+ 730 train 4.933874 (lr=1.5999e-03) (hash(x)=79027471)
92
+ 740 train 4.887502 (lr=1.5998e-03) (hash(x)=78149992)
93
+ 750 train 4.829276 (lr=1.5997e-03) (hash(x)=89147499)
94
+ 760 train 4.870240 (lr=1.5995e-03) (hash(x)=74931538)
95
+ 770 train 4.806106 (lr=1.5992e-03) (hash(x)=89648204)
96
+ 780 train 4.855033 (lr=1.5989e-03) (hash(x)=83708293)
97
+ 790 train 4.911362 (lr=1.5985e-03) (hash(x)=84336512)
98
+ 800 val loss 4.8276
99
+ 800 val perplexity 124.9129
100
+ 800 train 4.744928 (lr=1.5981e-03) (hash(x)=75025285)
101
+ 810 train 4.671612 (lr=1.5976e-03) (hash(x)=79436536)
102
+ 820 train 4.670561 (lr=1.5971e-03) (hash(x)=90467390)
103
+ 830 train 4.684519 (lr=1.5965e-03) (hash(x)=77292020)
104
+ 840 train 4.614301 (lr=1.5959e-03) (hash(x)=75568927)
105
+ 850 train 4.652792 (lr=1.5952e-03) (hash(x)=79671781)
106
+ 860 train 4.754268 (lr=1.5944e-03) (hash(x)=84280943)
107
+ 870 train 4.718310 (lr=1.5936e-03) (hash(x)=80901294)
108
+ 880 train 4.832251 (lr=1.5928e-03) (hash(x)=81437584)
109
+ 890 train 4.678856 (lr=1.5919e-03) (hash(x)=73893778)
110
+ 900 val loss 4.7112
111
+ 900 val perplexity 111.1862
112
+ 900 train 4.644197 (lr=1.5909e-03) (hash(x)=74987794)
113
+ 910 train 4.655973 (lr=1.5899e-03) (hash(x)=93721374)
114
+ 920 train 4.632627 (lr=1.5889e-03) (hash(x)=79149678)
115
+ 930 train 4.600400 (lr=1.5878e-03) (hash(x)=83179387)
116
+ 940 train 4.590433 (lr=1.5866e-03) (hash(x)=89009978)
117
+ 950 train 4.581834 (lr=1.5854e-03) (hash(x)=76483735)
118
+ 960 train 4.507612 (lr=1.5841e-03) (hash(x)=78638850)
119
+ 970 train 4.544681 (lr=1.5828e-03) (hash(x)=90415206)
120
+ 980 train 4.488647 (lr=1.5815e-03) (hash(x)=90819195)
121
+ 990 train 4.615076 (lr=1.5800e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.6005
123
+ 1000 val perplexity 99.5361
124
+ 1000 train 4.646739 (lr=1.5786e-03) (hash(x)=82436789)
125
+ 1010 train 4.698070 (lr=1.5770e-03) (hash(x)=63022149)
126
+ 1020 train 4.623026 (lr=1.5755e-03) (hash(x)=83501199)
127
+ 1030 train 4.588602 (lr=1.5738e-03) (hash(x)=77093326)
128
+ 1040 train 4.468604 (lr=1.5722e-03) (hash(x)=73437559)
129
+ 1050 train 4.557301 (lr=1.5704e-03) (hash(x)=84550388)
130
+ 1060 train 4.556844 (lr=1.5687e-03) (hash(x)=86184566)
131
+ 1070 train 4.554273 (lr=1.5668e-03) (hash(x)=85644922)
132
+ 1080 train 4.485902 (lr=1.5650e-03) (hash(x)=86179801)
133
+ 1090 train 4.459916 (lr=1.5630e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.5421
135
+ 1100 val perplexity 93.8891
136
+ 1100 train 4.576309 (lr=1.5610e-03) (hash(x)=96780388)
137
+ 1110 train 4.507275 (lr=1.5590e-03) (hash(x)=87223122)
138
+ 1120 train 4.469243 (lr=1.5569e-03) (hash(x)=77292786)
139
+ 1130 train 4.501046 (lr=1.5548e-03) (hash(x)=88761375)
140
+ 1140 train 4.405317 (lr=1.5526e-03) (hash(x)=80956468)
141
+ 1150 train 4.573056 (lr=1.5504e-03) (hash(x)=84725462)
142
+ 1160 train 4.571260 (lr=1.5481e-03) (hash(x)=76098113)
143
+ 1170 train 4.532968 (lr=1.5458e-03) (hash(x)=84387685)
144
+ 1180 train 4.480906 (lr=1.5434e-03) (hash(x)=76120321)
145
+ 1190 train 4.436551 (lr=1.5410e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4557
147
+ 1200 val perplexity 86.1193
148
+ 1200 train 4.340444 (lr=1.5385e-03) (hash(x)=79340644)
149
+ 1210 train 4.431742 (lr=1.5360e-03) (hash(x)=88345025)
150
+ 1220 train 4.343075 (lr=1.5334e-03) (hash(x)=86406230)
151
+ 1230 train 4.337785 (lr=1.5308e-03) (hash(x)=80608975)
152
+ 1240 train 4.387195 (lr=1.5281e-03) (hash(x)=81635225)
153
+ 1250 train 4.223104 (lr=1.5254e-03) (hash(x)=82126376)
154
+ 1260 train 4.400750 (lr=1.5226e-03) (hash(x)=83737972)
155
+ 1270 train 4.309894 (lr=1.5198e-03) (hash(x)=81132345)
156
+ 1280 train 4.401238 (lr=1.5170e-03) (hash(x)=85850404)
157
+ 1290 train 4.411984 (lr=1.5141e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.4162
159
+ 1300 val perplexity 82.7792
160
+ 1300 train 4.382085 (lr=1.5111e-03) (hash(x)=77545187)
161
+ 1310 train 4.436922 (lr=1.5081e-03) (hash(x)=86412685)
162
+ 1320 train 4.325025 (lr=1.5051e-03) (hash(x)=86429640)
163
+ 1330 train 4.465931 (lr=1.5020e-03) (hash(x)=79645281)
164
+ 1340 train 4.422534 (lr=1.4989e-03) (hash(x)=82668541)
165
+ 1350 train 4.280288 (lr=1.4957e-03) (hash(x)=73627248)
166
+ 1360 train 4.376854 (lr=1.4924e-03) (hash(x)=74845133)
167
+ 1370 train 4.288740 (lr=1.4892e-03) (hash(x)=77850497)
168
+ 1380 train 4.315447 (lr=1.4859e-03) (hash(x)=76143954)
169
+ 1390 train 4.321943 (lr=1.4825e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.3914
171
+ 1400 val perplexity 80.7523
172
+ 1400 train 4.302134 (lr=1.4791e-03) (hash(x)=76010938)
173
+ 1410 train 4.343997 (lr=1.4756e-03) (hash(x)=81260300)
174
+ 1420 train 4.461286 (lr=1.4721e-03) (hash(x)=83803629)
175
+ 1430 train 4.288841 (lr=1.4686e-03) (hash(x)=83840110)
176
+ 1440 train 4.482316 (lr=1.4650e-03) (hash(x)=97711831)
177
+ 1450 train 4.354192 (lr=1.4614e-03) (hash(x)=81110139)
178
+ 1460 train 4.280200 (lr=1.4577e-03) (hash(x)=90942839)
179
+ 1470 train 4.316661 (lr=1.4540e-03) (hash(x)=80721564)
180
+ 1480 train 4.411684 (lr=1.4503e-03) (hash(x)=77852059)
181
+ 1490 train 4.306622 (lr=1.4465e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3502
183
+ 1500 val perplexity 77.4904
184
+ 1500 train 4.336867 (lr=1.4426e-03) (hash(x)=84496142)
185
+ 1510 train 4.271552 (lr=1.4388e-03) (hash(x)=70186729)
186
+ 1520 train 4.293674 (lr=1.4348e-03) (hash(x)=74854227)
187
+ 1530 train 4.200214 (lr=1.4309e-03) (hash(x)=77468161)
188
+ 1540 train 4.220964 (lr=1.4269e-03) (hash(x)=87853059)
189
+ 1550 train 4.153920 (lr=1.4228e-03) (hash(x)=71225436)
190
+ 1560 train 4.255099 (lr=1.4188e-03) (hash(x)=84072783)
191
+ 1570 train 4.304233 (lr=1.4147e-03) (hash(x)=72723098)
192
+ 1580 train 4.322647 (lr=1.4105e-03) (hash(x)=90409866)
193
+ 1590 train 4.391687 (lr=1.4063e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.3189
195
+ 1600 val perplexity 75.1082
196
+ 1600 train 4.381252 (lr=1.4021e-03) (hash(x)=77643862)
197
+ 1610 train 4.265498 (lr=1.3978e-03) (hash(x)=88862575)
198
+ 1620 train 4.299489 (lr=1.3935e-03) (hash(x)=84612581)
199
+ 1630 train 4.320031 (lr=1.3891e-03) (hash(x)=87075989)
200
+ 1640 train 4.200347 (lr=1.3847e-03) (hash(x)=88277361)
201
+ 1650 train 4.313198 (lr=1.3803e-03) (hash(x)=78750236)
202
+ 1660 train 4.243175 (lr=1.3758e-03) (hash(x)=82604581)
203
+ 1670 train 4.080521 (lr=1.3713e-03) (hash(x)=68482265)
204
+ 1680 train 4.201686 (lr=1.3668e-03) (hash(x)=75088835)
205
+ 1690 train 4.149337 (lr=1.3622e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.3221
207
+ 1700 val perplexity 75.3463
208
+ 1700 train 4.312813 (lr=1.3576e-03) (hash(x)=79986754)
209
+ 1710 train 4.296456 (lr=1.3530e-03) (hash(x)=83657930)
210
+ 1720 train 4.247497 (lr=1.3483e-03) (hash(x)=81754135)
211
+ 1730 train 4.332112 (lr=1.3436e-03) (hash(x)=78904427)
212
+ 1740 train 4.223833 (lr=1.3389e-03) (hash(x)=85920177)
213
+ 1750 train 4.279968 (lr=1.3341e-03) (hash(x)=86573211)
214
+ 1760 train 4.232305 (lr=1.3293e-03) (hash(x)=81737128)
215
+ 1770 train 4.220219 (lr=1.3244e-03) (hash(x)=77400968)
216
+ 1780 train 4.239125 (lr=1.3195e-03) (hash(x)=73545497)
217
+ 1790 train 4.487744 (lr=1.3146e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.2718
219
+ 1800 val perplexity 71.6511
220
+ 1800 train 4.337315 (lr=1.3097e-03) (hash(x)=87819781)
221
+ 1810 train 4.198979 (lr=1.3047e-03) (hash(x)=86870770)
222
+ 1820 train 4.203950 (lr=1.2997e-03) (hash(x)=82522211)
223
+ 1830 train 4.288516 (lr=1.2947e-03) (hash(x)=61947437)
224
+ 1840 train 4.168036 (lr=1.2896e-03) (hash(x)=79865406)
225
+ 1850 train 4.131707 (lr=1.2845e-03) (hash(x)=79828721)
226
+ 1860 train 4.272793 (lr=1.2794e-03) (hash(x)=80869571)
227
+ 1870 train 4.262117 (lr=1.2742e-03) (hash(x)=73780971)
228
+ 1880 train 4.221981 (lr=1.2690e-03) (hash(x)=79249549)
229
+ 1890 train 4.277023 (lr=1.2638e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.2506
231
+ 1900 val perplexity 70.1459
232
+ 1900 train 4.325261 (lr=1.2586e-03) (hash(x)=82456430)
233
+ 1910 train 4.113831 (lr=1.2533e-03) (hash(x)=82222135)
234
+ 1920 train 4.156413 (lr=1.2480e-03) (hash(x)=70033249)
235
+ 1930 train 4.134238 (lr=1.2427e-03) (hash(x)=72887360)
236
+ 1940 train 4.184471 (lr=1.2373e-03) (hash(x)=83251100)
237
+ 1950 train 4.225343 (lr=1.2319e-03) (hash(x)=79660266)
238
+ 1960 train 4.108869 (lr=1.2265e-03) (hash(x)=78919068)
239
+ 1970 train 4.204139 (lr=1.2211e-03) (hash(x)=94690431)
240
+ 1980 train 4.041992 (lr=1.2157e-03) (hash(x)=88153756)
241
+ 1990 train 4.090948 (lr=1.2102e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.2234
243
+ 2000 val perplexity 68.2630
244
+ 2000 train 4.188568 (lr=1.2047e-03) (hash(x)=81308591)
245
+ 2010 train 4.242372 (lr=1.1991e-03) (hash(x)=80205479)
246
+ 2020 train 4.161426 (lr=1.1936e-03) (hash(x)=81008704)
247
+ 2030 train 4.223245 (lr=1.1880e-03) (hash(x)=82914358)
248
+ 2040 train 4.237023 (lr=1.1824e-03) (hash(x)=88064399)
249
+ 2050 train 4.237744 (lr=1.1768e-03) (hash(x)=82889390)
250
+ 2060 train 4.312442 (lr=1.1712e-03) (hash(x)=80913554)
251
+ 2070 train 4.147934 (lr=1.1655e-03) (hash(x)=78121791)
252
+ 2080 train 4.285027 (lr=1.1598e-03) (hash(x)=75356657)
253
+ 2090 train 4.108556 (lr=1.1541e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.2035
255
+ 2100 val perplexity 66.9180
256
+ 2100 train 4.029610 (lr=1.1484e-03) (hash(x)=68928225)
257
+ 2110 train 4.079400 (lr=1.1426e-03) (hash(x)=93734745)
258
+ 2120 train 4.078905 (lr=1.1369e-03) (hash(x)=72878164)
259
+ 2130 train 4.208786 (lr=1.1311e-03) (hash(x)=71508226)
260
+ 2140 train 4.167331 (lr=1.1253e-03) (hash(x)=84168671)
261
+ 2150 train 4.103094 (lr=1.1195e-03) (hash(x)=74673639)
262
+ 2160 train 4.249025 (lr=1.1136e-03) (hash(x)=75470031)
263
+ 2170 train 4.210122 (lr=1.1078e-03) (hash(x)=74307890)
264
+ 2180 train 4.319509 (lr=1.1019e-03) (hash(x)=77214245)
265
+ 2190 train 4.099950 (lr=1.0960e-03) (hash(x)=88628359)
266
+ 2200 val loss 4.1806
267
+ 2200 val perplexity 65.4070
268
+ 2200 train 4.122203 (lr=1.0901e-03) (hash(x)=74779126)
269
+ 2210 train 4.159711 (lr=1.0842e-03) (hash(x)=79817976)
270
+ 2220 train 4.111415 (lr=1.0783e-03) (hash(x)=75258996)
271
+ 2230 train 4.085649 (lr=1.0723e-03) (hash(x)=88164047)
272
+ 2240 train 4.065951 (lr=1.0663e-03) (hash(x)=78580686)
273
+ 2250 train 4.006695 (lr=1.0604e-03) (hash(x)=76510617)
274
+ 2260 train 4.082248 (lr=1.0544e-03) (hash(x)=76043966)
275
+ 2270 train 4.060785 (lr=1.0484e-03) (hash(x)=73062098)
276
+ 2280 train 4.285598 (lr=1.0424e-03) (hash(x)=81885909)
277
+ 2290 train 4.224711 (lr=1.0363e-03) (hash(x)=81481741)
278
+ 2300 val loss 4.1603
279
+ 2300 val perplexity 64.0923
280
+ 2300 train 4.080654 (lr=1.0303e-03) (hash(x)=82104275)
281
+ 2310 train 4.129953 (lr=1.0243e-03) (hash(x)=82137309)
282
+ 2320 train 4.111996 (lr=1.0182e-03) (hash(x)=86741167)
283
+ 2330 train 4.125095 (lr=1.0121e-03) (hash(x)=84796263)
284
+ 2340 train 4.095835 (lr=1.0060e-03) (hash(x)=169966529)
285
+ 2350 train 3.970949 (lr=9.9995e-04) (hash(x)=80006061)
286
+ 2360 train 4.028465 (lr=9.9385e-04) (hash(x)=71090999)
287
+ 2370 train 4.159739 (lr=9.8775e-04) (hash(x)=86641850)
288
+ 2380 train 4.058403 (lr=9.8163e-04) (hash(x)=84573253)
289
+ 2390 train 4.095134 (lr=9.7551e-04) (hash(x)=74483764)
290
+ 2400 val loss 4.1545
291
+ 2400 val perplexity 63.7212
292
+ 2400 train 4.063015 (lr=9.6938e-04) (hash(x)=78327659)
293
+ 2410 train 4.129412 (lr=9.6325e-04) (hash(x)=77273627)
294
+ 2420 train 3.991803 (lr=9.5710e-04) (hash(x)=76938049)
295
+ 2430 train 3.998128 (lr=9.5096e-04) (hash(x)=98449442)
296
+ 2440 train 3.858179 (lr=9.4480e-04) (hash(x)=78157797)
297
+ 2450 train 3.896312 (lr=9.3865e-04) (hash(x)=80637582)
298
+ 2460 train 3.885936 (lr=9.3248e-04) (hash(x)=80023854)
299
+ 2470 train 3.949436 (lr=9.2632e-04) (hash(x)=86678884)
300
+ 2480 train 4.172156 (lr=9.2015e-04) (hash(x)=84086469)
301
+ 2490 train 4.194436 (lr=9.1398e-04) (hash(x)=83131141)
302
+ 2500 val loss 4.1298
303
+ 2500 val perplexity 62.1667
304
+ 2500 train 4.107302 (lr=9.0780e-04) (hash(x)=82583497)
305
+ 2510 train 4.171342 (lr=9.0163e-04) (hash(x)=79164326)
306
+ 2520 train 4.077996 (lr=8.9545e-04) (hash(x)=76456503)
307
+ 2530 train 4.136618 (lr=8.8927e-04) (hash(x)=82001228)
308
+ 2540 train 4.174486 (lr=8.8309e-04) (hash(x)=75615595)
309
+ 2550 train 4.114452 (lr=8.7691e-04) (hash(x)=78397869)
310
+ 2560 train 4.119451 (lr=8.7073e-04) (hash(x)=75547032)
311
+ 2570 train 4.032382 (lr=8.6455e-04) (hash(x)=89201025)
312
+ 2580 train 4.019038 (lr=8.5837e-04) (hash(x)=89856704)
313
+ 2590 train 4.022391 (lr=8.5220e-04) (hash(x)=82175682)
314
+ 2600 val loss 4.1141
315
+ 2600 val perplexity 61.1998
316
+ 2600 train 4.035685 (lr=8.4602e-04) (hash(x)=78312826)
317
+ 2610 train 4.025533 (lr=8.3985e-04) (hash(x)=77066588)
318
+ 2620 train 3.868984 (lr=8.3368e-04) (hash(x)=78666061)
319
+ 2630 train 3.865921 (lr=8.2752e-04) (hash(x)=93762143)
320
+ 2640 train 3.883277 (lr=8.2135e-04) (hash(x)=83191587)
321
+ 2650 train 3.803095 (lr=8.1520e-04) (hash(x)=87169585)
322
+ 2660 train 4.018660 (lr=8.0904e-04) (hash(x)=86426388)
323
+ 2670 train 4.157342 (lr=8.0290e-04) (hash(x)=76692638)
324
+ 2680 train 4.072244 (lr=7.9675e-04) (hash(x)=77446063)
325
+ 2690 train 4.073360 (lr=7.9062e-04) (hash(x)=79809050)
326
+ 2700 val loss 4.1075
327
+ 2700 val perplexity 60.7931
328
+ 2700 train 4.003060 (lr=7.8449e-04) (hash(x)=83116823)
329
+ 2710 train 4.117026 (lr=7.7837e-04) (hash(x)=75622148)
330
+ 2720 train 4.045377 (lr=7.7225e-04) (hash(x)=80690512)
331
+ 2730 train 4.112003 (lr=7.6615e-04) (hash(x)=78009984)
332
+ 2740 train 4.066589 (lr=7.6005e-04) (hash(x)=74205488)
333
+ 2750 train 4.068827 (lr=7.5396e-04) (hash(x)=91013332)
334
+ 2760 train 3.993717 (lr=7.4788e-04) (hash(x)=76954961)
335
+ 2770 train 3.999740 (lr=7.4181e-04) (hash(x)=79390317)
336
+ 2780 train 3.868793 (lr=7.3575e-04) (hash(x)=70168783)
337
+ 2790 train 4.032517 (lr=7.2970e-04) (hash(x)=76028417)
338
+ 2800 val loss 4.0883
339
+ 2800 val perplexity 59.6405
340
+ 2800 train 3.859344 (lr=7.2366e-04) (hash(x)=77656050)
341
+ 2810 train 3.902589 (lr=7.1763e-04) (hash(x)=94248216)
342
+ 2820 train 3.975457 (lr=7.1162e-04) (hash(x)=78305078)
343
+ 2830 train 3.878232 (lr=7.0562e-04) (hash(x)=79948848)
344
+ 2840 train 3.749076 (lr=6.9963e-04) (hash(x)=85341024)
345
+ 2850 train 4.055197 (lr=6.9365e-04) (hash(x)=78735170)
346
+ 2860 train 4.304806 (lr=6.8769e-04) (hash(x)=71616419)
347
+ 2870 train 4.077643 (lr=6.8174e-04) (hash(x)=78656517)
348
+ 2880 train 4.089967 (lr=6.7581e-04) (hash(x)=80073987)
349
+ 2890 train 4.020313 (lr=6.6989e-04) (hash(x)=76894809)
350
+ 2900 val loss 4.0621
351
+ 2900 val perplexity 58.0956
352
+ 2900 train 3.995512 (lr=6.6398e-04) (hash(x)=80499838)
353
+ 2910 train 4.049630 (lr=6.5810e-04) (hash(x)=72673354)
354
+ 2920 train 4.029279 (lr=6.5222e-04) (hash(x)=84265768)
355
+ 2930 train 4.061096 (lr=6.4637e-04) (hash(x)=79612060)
356
+ 2940 train 3.915008 (lr=6.4053e-04) (hash(x)=74970087)
357
+ 2950 train 4.068076 (lr=6.3471e-04) (hash(x)=84166818)
358
+ 2960 train 4.056458 (lr=6.2891e-04) (hash(x)=89410221)
359
+ 2970 train 3.998572 (lr=6.2313e-04) (hash(x)=75672566)
360
+ 2980 train 3.853001 (lr=6.1737e-04) (hash(x)=81760314)
361
+ 2990 train 3.828447 (lr=6.1162e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.0621
363
+ 3000 val perplexity 58.0982
364
+ 3000 train 3.820675 (lr=6.0590e-04) (hash(x)=83804735)
365
+ 3010 train 3.753595 (lr=6.0019e-04) (hash(x)=77015303)
366
+ 3020 train 3.821736 (lr=5.9451e-04) (hash(x)=81464523)
367
+ 3030 train 4.018709 (lr=5.8884e-04) (hash(x)=80168230)
368
+ 3040 train 3.954669 (lr=5.8320e-04) (hash(x)=84817006)
369
+ 3050 train 4.067798 (lr=5.7758e-04) (hash(x)=71601811)
370
+ 3060 train 3.986993 (lr=5.7199e-04) (hash(x)=85499733)
371
+ 3070 train 4.043632 (lr=5.6641e-04) (hash(x)=78670408)
372
+ 3080 train 3.935961 (lr=5.6086e-04) (hash(x)=77120468)
373
+ 3090 train 4.041907 (lr=5.5533e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.0429
375
+ 3100 val perplexity 56.9897
376
+ 3100 train 4.060834 (lr=5.4983e-04) (hash(x)=83998606)
377
+ 3110 train 3.958346 (lr=5.4435e-04) (hash(x)=77990218)
378
+ 3120 train 3.867551 (lr=5.3889e-04) (hash(x)=81623970)
379
+ 3130 train 3.846773 (lr=5.3346e-04) (hash(x)=75014781)
380
+ 3140 train 3.898767 (lr=5.2806e-04) (hash(x)=72591250)
381
+ 3150 train 3.913369 (lr=5.2268e-04) (hash(x)=81421847)
382
+ 3160 train 3.963468 (lr=5.1733e-04) (hash(x)=75112631)
383
+ 3170 train 4.057982 (lr=5.1200e-04) (hash(x)=87518033)
384
+ 3180 train 3.912035 (lr=5.0670e-04) (hash(x)=78648348)
385
+ 3190 train 4.025384 (lr=5.0143e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0248
387
+ 3200 val perplexity 55.9687
388
+ 3200 train 4.102901 (lr=4.9619e-04) (hash(x)=83380714)
389
+ 3210 train 4.278379 (lr=4.9097e-04) (hash(x)=67804991)
390
+ 3220 train 4.008860 (lr=4.8579e-04) (hash(x)=76968804)
391
+ 3230 train 3.941929 (lr=4.8063e-04) (hash(x)=75798670)
392
+ 3240 train 4.020394 (lr=4.7550e-04) (hash(x)=83192811)
393
+ 3250 train 3.956072 (lr=4.7041e-04) (hash(x)=81659789)
394
+ 3260 train 3.995971 (lr=4.6534e-04) (hash(x)=80622502)
395
+ 3270 train 3.966916 (lr=4.6030e-04) (hash(x)=85436511)
396
+ 3280 train 3.912412 (lr=4.5530e-04) (hash(x)=80589180)
397
+ 3290 train 4.049465 (lr=4.5032e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.0193
399
+ 3300 val perplexity 55.6630
400
+ 3300 train 3.915854 (lr=4.4538e-04) (hash(x)=74635692)
401
+ 3310 train 3.903398 (lr=4.4047e-04) (hash(x)=82045455)
402
+ 3320 train 3.772467 (lr=4.3559e-04) (hash(x)=73249173)
403
+ 3330 train 3.731083 (lr=4.3074e-04) (hash(x)=71492338)
404
+ 3340 train 4.064515 (lr=4.2593e-04) (hash(x)=76374071)
405
+ 3350 train 4.037301 (lr=4.2115e-04) (hash(x)=80474064)
406
+ 3360 train 3.948670 (lr=4.1640e-04) (hash(x)=84390892)
407
+ 3370 train 3.996154 (lr=4.1169e-04) (hash(x)=83399949)
408
+ 3380 train 4.046316 (lr=4.0702e-04) (hash(x)=77648059)
409
+ 3390 train 4.014450 (lr=4.0237e-04) (hash(x)=80152701)
410
+ 3400 val loss 4.0004
411
+ 3400 val perplexity 54.6225
412
+ 3400 train 4.063615 (lr=3.9777e-04) (hash(x)=81952545)
413
+ 3410 train 4.024412 (lr=3.9320e-04) (hash(x)=80908993)
414
+ 3420 train 4.064878 (lr=3.8866e-04) (hash(x)=79610037)
415
+ 3430 train 3.942470 (lr=3.8416e-04) (hash(x)=87624382)
416
+ 3440 train 3.990324 (lr=3.7970e-04) (hash(x)=82336381)
417
+ 3450 train 3.967084 (lr=3.7527e-04) (hash(x)=87687835)
418
+ 3460 train 3.972509 (lr=3.7088e-04) (hash(x)=76376135)
419
+ 3470 train 3.882522 (lr=3.6653e-04) (hash(x)=79331391)
420
+ 3480 train 3.792093 (lr=3.6222e-04) (hash(x)=107002681)
421
+ 3490 train 3.872307 (lr=3.5794e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.9972
423
+ 3500 val perplexity 54.4467
424
+ 3500 train 3.834981 (lr=3.5371e-04) (hash(x)=88237229)
425
+ 3510 train 3.900682 (lr=3.4951e-04) (hash(x)=81011739)
426
+ 3520 train 3.950568 (lr=3.4535e-04) (hash(x)=64643427)
427
+ 3530 train 4.228482 (lr=3.4123e-04) (hash(x)=78029539)
428
+ 3540 train 4.060553 (lr=3.3715e-04) (hash(x)=83188968)
429
+ 3550 train 4.075885 (lr=3.3311e-04) (hash(x)=86104185)
430
+ 3560 train 4.003749 (lr=3.2911e-04) (hash(x)=81993629)
431
+ 3570 train 3.977735 (lr=3.2515e-04) (hash(x)=72719368)
432
+ 3580 train 3.984967 (lr=3.2123e-04) (hash(x)=90091487)
433
+ 3590 train 3.871514 (lr=3.1736e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9788
435
+ 3600 val perplexity 53.4507
436
+ 3600 train 3.982993 (lr=3.1352e-04) (hash(x)=73103504)
437
+ 3610 train 3.985508 (lr=3.0973e-04) (hash(x)=84429400)
438
+ 3620 train 3.961816 (lr=3.0598e-04) (hash(x)=77021795)
439
+ 3630 train 3.918333 (lr=3.0227e-04) (hash(x)=92830605)
440
+ 3640 train 4.111105 (lr=2.9860e-04) (hash(x)=78313175)
441
+ 3650 train 3.910505 (lr=2.9498e-04) (hash(x)=85395549)
442
+ 3660 train 3.704443 (lr=2.9139e-04) (hash(x)=78114459)
443
+ 3670 train 3.828346 (lr=2.8786e-04) (hash(x)=74968316)
444
+ 3680 train 3.759316 (lr=2.8436e-04) (hash(x)=73358737)
445
+ 3690 train 3.865646 (lr=2.8091e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9812
447
+ 3700 val perplexity 53.5832
448
+ 3700 train 3.866280 (lr=2.7751e-04) (hash(x)=74039273)
449
+ 3710 train 4.041371 (lr=2.7415e-04) (hash(x)=79299680)
450
+ 3720 train 3.954784 (lr=2.7083e-04) (hash(x)=94701498)
451
+ 3730 train 4.037948 (lr=2.6756e-04) (hash(x)=75352071)
452
+ 3740 train 3.911220 (lr=2.6433e-04) (hash(x)=77909487)
453
+ 3750 train 4.018800 (lr=2.6115e-04) (hash(x)=73986730)
454
+ 3760 train 3.979075 (lr=2.5801e-04) (hash(x)=79325763)
455
+ 3770 train 4.025459 (lr=2.5492e-04) (hash(x)=72457818)
456
+ 3780 train 4.156508 (lr=2.5188e-04) (hash(x)=71775590)
457
+ 3790 train 4.045463 (lr=2.4888e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9651
459
+ 3800 val perplexity 52.7235
460
+ 3800 train 4.018735 (lr=2.4593e-04) (hash(x)=79965893)
461
+ 3810 train 3.951230 (lr=2.4302e-04) (hash(x)=72598235)
462
+ 3820 train 3.938205 (lr=2.4017e-04) (hash(x)=83113889)
463
+ 3830 train 3.997554 (lr=2.3736e-04) (hash(x)=74434590)
464
+ 3840 train 3.915114 (lr=2.3459e-04) (hash(x)=82860348)
465
+ 3850 train 3.848972 (lr=2.3188e-04) (hash(x)=78067565)
466
+ 3860 train 3.917250 (lr=2.2921e-04) (hash(x)=82592498)
467
+ 3870 train 3.850919 (lr=2.2659e-04) (hash(x)=81820733)
468
+ 3880 train 3.864441 (lr=2.2402e-04) (hash(x)=87709040)
469
+ 3890 train 3.908478 (lr=2.2150e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9595
471
+ 3900 val perplexity 52.4303
472
+ 3900 train 4.034097 (lr=2.1902e-04) (hash(x)=76597431)
473
+ 3910 train 3.917224 (lr=2.1659e-04) (hash(x)=90490716)
474
+ 3920 train 4.022794 (lr=2.1422e-04) (hash(x)=81970659)
475
+ 3930 train 4.008457 (lr=2.1189e-04) (hash(x)=81496334)
476
+ 3940 train 3.983382 (lr=2.0961e-04) (hash(x)=75717605)
477
+ 3950 train 3.928293 (lr=2.0738e-04) (hash(x)=82749357)
478
+ 3960 train 3.960322 (lr=2.0520e-04) (hash(x)=89355157)
479
+ 3970 train 3.914793 (lr=2.0307e-04) (hash(x)=78980403)
480
+ 3980 train 3.887836 (lr=2.0099e-04) (hash(x)=76627217)
481
+ 3990 train 3.889910 (lr=1.9896e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9537
483
+ 4000 val perplexity 52.1260
484
+ 4000 train 3.905736 (lr=1.9698e-04) (hash(x)=83018142)
485
+ 4010 train 3.920435 (lr=1.9505e-04) (hash(x)=81272436)
486
+ 4020 train 3.847416 (lr=1.9317e-04) (hash(x)=85497482)
487
+ 4030 train 3.921211 (lr=1.9134e-04) (hash(x)=85241734)
488
+ 4040 train 3.883780 (lr=1.8956e-04) (hash(x)=77925307)
489
+ 4050 train 3.884185 (lr=1.8783e-04) (hash(x)=84826179)
490
+ 4060 train 3.943526 (lr=1.8616e-04) (hash(x)=83606764)
491
+ 4070 train 3.853800 (lr=1.8453e-04) (hash(x)=80567590)
492
+ 4080 train 3.908240 (lr=1.8296e-04) (hash(x)=76860998)
493
+ 4090 train 3.942564 (lr=1.8144e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9429
495
+ 4100 val perplexity 51.5678
496
+ 4100 train 3.987089 (lr=1.7997e-04) (hash(x)=82832041)
497
+ 4110 train 3.869711 (lr=1.7855e-04) (hash(x)=79143262)
498
+ 4120 train 3.892467 (lr=1.7718e-04) (hash(x)=77038149)
499
+ 4130 train 4.075899 (lr=1.7586e-04) (hash(x)=86339074)
500
+ 4140 train 4.137580 (lr=1.7460e-04) (hash(x)=76686216)
501
+ 4150 train 3.920357 (lr=1.7339e-04) (hash(x)=70522682)
502
+ 4160 train 3.882373 (lr=1.7223e-04) (hash(x)=90958555)
503
+ 4170 train 3.909845 (lr=1.7112e-04) (hash(x)=91463532)
504
+ 4180 train 3.897070 (lr=1.7006e-04) (hash(x)=81959329)
505
+ 4190 train 3.851868 (lr=1.6906e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9458
507
+ 4200 val perplexity 51.7182
508
+ 4200 train 3.835744 (lr=1.6811e-04) (hash(x)=78361715)
509
+ 4210 train 3.984055 (lr=1.6721e-04) (hash(x)=87364889)
510
+ 4220 train 3.837393 (lr=1.6636e-04) (hash(x)=70465156)
511
+ 4230 train 3.992865 (lr=1.6557e-04) (hash(x)=84524081)
512
+ 4240 train 3.944539 (lr=1.6483e-04) (hash(x)=77824868)
513
+ 4250 train 3.958789 (lr=1.6414e-04) (hash(x)=81710711)
514
+ 4260 train 3.942909 (lr=1.6350e-04) (hash(x)=76362728)
515
+ 4270 train 3.932465 (lr=1.6292e-04) (hash(x)=83115208)
516
+ 4280 train 3.844678 (lr=1.6239e-04) (hash(x)=87218314)
517
+ 4290 train 4.006399 (lr=1.6192e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9395
519
+ 4300 val perplexity 51.3953
520
+ 4300 train 3.881313 (lr=1.6149e-04) (hash(x)=77379615)
521
+ 4310 train 3.856133 (lr=1.6112e-04) (hash(x)=78669579)
522
+ 4320 train 3.971976 (lr=1.6080e-04) (hash(x)=83066608)
523
+ 4330 train 3.935461 (lr=1.6054e-04) (hash(x)=83037340)
524
+ 4340 train 3.977983 (lr=1.6032e-04) (hash(x)=82849771)
525
+ 4350 train 3.912361 (lr=1.6017e-04) (hash(x)=76693985)
526
+ 4360 train 3.995797 (lr=1.6006e-04) (hash(x)=77745394)
527
+ 4370 train 3.963867 (lr=1.6001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9343
529
+ 4374 val perplexity 51.1240
12_head_baseline_lr_16e-4_head_dim_22_fixed/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f63abc4cc474f1070d3d30f62b5cb0d442573d4ea28d8cb63ff73f0ee107ca0
3
+ size 96858242
12_head_baseline_lr_16e-4_head_dim_22_fixed/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d52b386ae71183787605993e3f9b497e05daed176469e15c2626426e34a886c
3
+ size 187435910