andrew-healey commited on
Commit
34ebca8
·
verified ·
1 Parent(s): 731f8e5

Upload folder using huggingface_hub

Browse files
scaling_sdpa_nh_2/sdpa_nh_128_seed_1339/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "self", "log_dir": "scaling_sdpa_nh_2/sdpa_nh_128_seed_1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 128, "n_embd": 256, "head_dim": 32, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "scaling_sdpa_nh_2", "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, "disable_selection_head_linear_combo_bias": false, "assert_latent_matches_no_head": false, "n_latent_masks": null, "init_latent_masks_to_identity": false, "init_latent_masks_to_inverse": false, "latent_mask_scale": null, "latent_mask_runtime_multiplier": null, "latent_mask_sigmoid": false, "S_layernorm": false, "one_head_per_latent_mask": false, "att_conv": false, "att_conv_init": "none", "att_conv_scale": 1.0, "att_conv_weight_decay": false, "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, "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, "mup_zero_init": true, "l1_loss": false, "debugpy": false, "key": "sdpa_nh_128", "latent_mask_precision": "bfloat16", "att_conv_precision": "bfloat16", "attn_precision": "bfloat16", "profile_kind": "none", "dense_attention_kind": "mha", "head_dim_value": 32, "sdpa_iter_size": null, "stabilize_attn_scores": false, "override_use_sdpa": false, "autocast_precision": "bfloat16", "simulate_micro_bs": null, "simulate_micro_bs_2": null, "c_proj_scale_init": 1.0}
scaling_sdpa_nh_2/sdpa_nh_128_seed_1339/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ec6682852ed60a9d42cb4047300f65a04f87328c0ad7a4516be84d11b28f216
3
+ size 964
scaling_sdpa_nh_2/sdpa_nh_128_seed_1339/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.2632
3
+ 0 val perplexity 77897.9531
4
+ 0 train 11.260119 (lr=4.1958e-06) (hash(x)=93356070)
5
+ 10 train 9.889592 (lr=4.6154e-05) (hash(x)=91253010)
6
+ 20 train 9.196297 (lr=8.8112e-05) (hash(x)=74934453)
7
+ 30 train 8.474638 (lr=1.3007e-04) (hash(x)=79278034)
8
+ 40 train 7.709443 (lr=1.7203e-04) (hash(x)=80107892)
9
+ 50 train 7.424641 (lr=2.1399e-04) (hash(x)=70745428)
10
+ 60 train 7.421083 (lr=2.5594e-04) (hash(x)=80077589)
11
+ 70 train 7.036756 (lr=2.9790e-04) (hash(x)=76213766)
12
+ 80 train 7.037086 (lr=3.3986e-04) (hash(x)=83218328)
13
+ 90 train 6.854251 (lr=3.8182e-04) (hash(x)=74855845)
14
+ 100 val loss 6.7922
15
+ 100 val perplexity 890.8463
16
+ 100 train 6.652158 (lr=4.2378e-04) (hash(x)=82814902)
17
+ 110 train 6.630437 (lr=4.6573e-04) (hash(x)=84907741)
18
+ 120 train 6.540632 (lr=5.0769e-04) (hash(x)=82613223)
19
+ 130 train 6.443177 (lr=5.4965e-04) (hash(x)=83540876)
20
+ 140 train 6.420722 (lr=5.9161e-04) (hash(x)=75095216)
21
+ 150 train 6.520318 (lr=6.3357e-04) (hash(x)=97190944)
22
+ 160 train 6.467161 (lr=6.7552e-04) (hash(x)=82117809)
23
+ 170 train 6.391827 (lr=7.1748e-04) (hash(x)=70514724)
24
+ 180 train 6.332396 (lr=7.5944e-04) (hash(x)=81029624)
25
+ 190 train 6.272781 (lr=8.0140e-04) (hash(x)=78996841)
26
+ 200 val loss 6.1906
27
+ 200 val perplexity 488.1447
28
+ 200 train 6.228766 (lr=8.4336e-04) (hash(x)=79845097)
29
+ 210 train 6.145173 (lr=8.8531e-04) (hash(x)=78997683)
30
+ 220 train 6.099105 (lr=9.2727e-04) (hash(x)=74895865)
31
+ 230 train 6.221406 (lr=9.6923e-04) (hash(x)=80933276)
32
+ 240 train 5.950763 (lr=1.0112e-03) (hash(x)=77664606)
33
+ 250 train 6.018729 (lr=1.0531e-03) (hash(x)=75903930)
34
+ 260 train 5.913026 (lr=1.0951e-03) (hash(x)=82996853)
35
+ 270 train 5.874008 (lr=1.1371e-03) (hash(x)=73269568)
36
+ 280 train 5.835117 (lr=1.1790e-03) (hash(x)=82906427)
37
+ 290 train 5.740913 (lr=1.2210e-03) (hash(x)=71797895)
38
+ 300 val loss 5.8528
39
+ 300 val perplexity 348.1910
40
+ 300 train 5.623203 (lr=1.2629e-03) (hash(x)=85232249)
41
+ 310 train 5.588770 (lr=1.3049e-03) (hash(x)=75252489)
42
+ 320 train 5.649347 (lr=1.3469e-03) (hash(x)=74394644)
43
+ 330 train 5.763138 (lr=1.3888e-03) (hash(x)=82772910)
44
+ 340 train 5.770804 (lr=1.4308e-03) (hash(x)=81627464)
45
+ 350 train 5.744124 (lr=1.4727e-03) (hash(x)=84678053)
46
+ 360 train 5.690562 (lr=1.5147e-03) (hash(x)=81884128)
47
+ 370 train 5.685913 (lr=1.5566e-03) (hash(x)=76379242)
48
+ 380 train 5.514773 (lr=1.5986e-03) (hash(x)=98182875)
49
+ 390 train 5.477813 (lr=1.6406e-03) (hash(x)=79710436)
50
+ 400 val loss 5.4758
51
+ 400 val perplexity 238.8312
52
+ 400 train 5.457915 (lr=1.6825e-03) (hash(x)=79841071)
53
+ 410 train 5.402215 (lr=1.7245e-03) (hash(x)=75844151)
54
+ 420 train 5.416801 (lr=1.7664e-03) (hash(x)=73125036)
55
+ 430 train 5.431221 (lr=1.8084e-03) (hash(x)=84214858)
56
+ 440 train 5.342627 (lr=1.8503e-03) (hash(x)=80456994)
57
+ 450 train 5.297791 (lr=1.8923e-03) (hash(x)=76735962)
58
+ 460 train 5.227453 (lr=1.9343e-03) (hash(x)=81845446)
59
+ 470 train 5.063970 (lr=1.9762e-03) (hash(x)=76094689)
60
+ 480 train 5.162570 (lr=2.0182e-03) (hash(x)=83806686)
61
+ 490 train 4.981965 (lr=2.0601e-03) (hash(x)=84690227)
62
+ 500 val loss 5.2707
63
+ 500 val perplexity 194.5564
64
+ 500 train 5.064273 (lr=2.1021e-03) (hash(x)=71851938)
65
+ 510 train 5.160795 (lr=2.1441e-03) (hash(x)=77159346)
66
+ 520 train 5.209558 (lr=2.1860e-03) (hash(x)=80755753)
67
+ 530 train 5.344040 (lr=2.2280e-03) (hash(x)=75379679)
68
+ 540 train 5.232060 (lr=2.2699e-03) (hash(x)=82458619)
69
+ 550 train 5.208354 (lr=2.3119e-03) (hash(x)=86366396)
70
+ 560 train 5.212919 (lr=2.3538e-03) (hash(x)=85095044)
71
+ 570 train 5.189861 (lr=2.3958e-03) (hash(x)=78385159)
72
+ 580 train 5.134863 (lr=2.4378e-03) (hash(x)=79342394)
73
+ 590 train 5.125693 (lr=2.4797e-03) (hash(x)=70782192)
74
+ 600 val loss 5.0940
75
+ 600 val perplexity 163.0485
76
+ 600 train 5.123775 (lr=2.5217e-03) (hash(x)=76130353)
77
+ 610 train 5.039973 (lr=2.5636e-03) (hash(x)=74778440)
78
+ 620 train 5.006696 (lr=2.6056e-03) (hash(x)=79129709)
79
+ 630 train 4.923489 (lr=2.6476e-03) (hash(x)=76469962)
80
+ 640 train 4.925449 (lr=2.6895e-03) (hash(x)=78288049)
81
+ 650 train 4.877069 (lr=2.7315e-03) (hash(x)=76641388)
82
+ 660 train 4.800553 (lr=2.7734e-03) (hash(x)=75937906)
83
+ 670 train 4.795075 (lr=2.8154e-03) (hash(x)=74807157)
84
+ 680 train 4.694247 (lr=2.8573e-03) (hash(x)=77490144)
85
+ 690 train 4.736576 (lr=2.8993e-03) (hash(x)=72900124)
86
+ 700 val loss 4.9675
87
+ 700 val perplexity 143.6678
88
+ 700 train 4.705791 (lr=2.9413e-03) (hash(x)=73424218)
89
+ 710 train 5.001739 (lr=2.9832e-03) (hash(x)=86052345)
90
+ 720 train 4.954493 (lr=3.0000e-03) (hash(x)=77462613)
91
+ 730 train 4.970141 (lr=2.9999e-03) (hash(x)=79027471)
92
+ 740 train 4.926847 (lr=2.9997e-03) (hash(x)=78149992)
93
+ 750 train 4.887549 (lr=2.9994e-03) (hash(x)=89147499)
94
+ 760 train 4.904705 (lr=2.9990e-03) (hash(x)=74931538)
95
+ 770 train 4.849627 (lr=2.9985e-03) (hash(x)=89648204)
96
+ 780 train 4.900272 (lr=2.9979e-03) (hash(x)=83708293)
97
+ 790 train 4.903335 (lr=2.9972e-03) (hash(x)=84336512)
98
+ 800 val loss 4.8152
99
+ 800 val perplexity 123.3659
100
+ 800 train 4.777107 (lr=2.9964e-03) (hash(x)=75025285)
101
+ 810 train 4.711834 (lr=2.9955e-03) (hash(x)=79436536)
102
+ 820 train 4.695230 (lr=2.9945e-03) (hash(x)=90467390)
103
+ 830 train 4.731132 (lr=2.9934e-03) (hash(x)=77292020)
104
+ 840 train 4.657682 (lr=2.9922e-03) (hash(x)=75568927)
105
+ 850 train 4.682616 (lr=2.9909e-03) (hash(x)=79671781)
106
+ 860 train 4.757777 (lr=2.9896e-03) (hash(x)=84280943)
107
+ 870 train 4.756111 (lr=2.9881e-03) (hash(x)=80901294)
108
+ 880 train 4.880363 (lr=2.9865e-03) (hash(x)=81437584)
109
+ 890 train 4.711951 (lr=2.9848e-03) (hash(x)=73893778)
110
+ 900 val loss 4.6863
111
+ 900 val perplexity 108.4552
112
+ 900 train 4.665931 (lr=2.9830e-03) (hash(x)=74987794)
113
+ 910 train 4.680180 (lr=2.9811e-03) (hash(x)=93721374)
114
+ 920 train 4.663860 (lr=2.9792e-03) (hash(x)=79149678)
115
+ 930 train 4.622104 (lr=2.9771e-03) (hash(x)=83179387)
116
+ 940 train 4.617282 (lr=2.9749e-03) (hash(x)=89009978)
117
+ 950 train 4.605812 (lr=2.9726e-03) (hash(x)=76483735)
118
+ 960 train 4.538186 (lr=2.9703e-03) (hash(x)=78638850)
119
+ 970 train 4.566780 (lr=2.9678e-03) (hash(x)=90415206)
120
+ 980 train 4.506215 (lr=2.9652e-03) (hash(x)=90819195)
121
+ 990 train 4.635306 (lr=2.9626e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.5732
123
+ 1000 val perplexity 96.8541
124
+ 1000 train 4.667799 (lr=2.9598e-03) (hash(x)=82436789)
125
+ 1010 train 4.712183 (lr=2.9570e-03) (hash(x)=63022149)
126
+ 1020 train 4.634925 (lr=2.9540e-03) (hash(x)=83501199)
127
+ 1030 train 4.611520 (lr=2.9510e-03) (hash(x)=77093326)
128
+ 1040 train 4.496390 (lr=2.9478e-03) (hash(x)=73437559)
129
+ 1050 train 4.584085 (lr=2.9446e-03) (hash(x)=84550388)
130
+ 1060 train 4.577969 (lr=2.9412e-03) (hash(x)=86184566)
131
+ 1070 train 4.566227 (lr=2.9378e-03) (hash(x)=85644922)
132
+ 1080 train 4.504327 (lr=2.9343e-03) (hash(x)=86179801)
133
+ 1090 train 4.487934 (lr=2.9307e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.5092
135
+ 1100 val perplexity 90.8470
136
+ 1100 train 4.577137 (lr=2.9270e-03) (hash(x)=96780388)
137
+ 1110 train 4.509242 (lr=2.9231e-03) (hash(x)=87223122)
138
+ 1120 train 4.476114 (lr=2.9192e-03) (hash(x)=77292786)
139
+ 1130 train 4.503420 (lr=2.9152e-03) (hash(x)=88761375)
140
+ 1140 train 4.415630 (lr=2.9112e-03) (hash(x)=80956468)
141
+ 1150 train 4.578532 (lr=2.9070e-03) (hash(x)=84725462)
142
+ 1160 train 4.568283 (lr=2.9027e-03) (hash(x)=76098113)
143
+ 1170 train 4.544624 (lr=2.8983e-03) (hash(x)=84387685)
144
+ 1180 train 4.492444 (lr=2.8939e-03) (hash(x)=76120321)
145
+ 1190 train 4.456184 (lr=2.8893e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4219
147
+ 1200 val perplexity 83.2533
148
+ 1200 train 4.350690 (lr=2.8847e-03) (hash(x)=79340644)
149
+ 1210 train 4.442658 (lr=2.8800e-03) (hash(x)=88345025)
150
+ 1220 train 4.350668 (lr=2.8751e-03) (hash(x)=86406230)
151
+ 1230 train 4.346893 (lr=2.8702e-03) (hash(x)=80608975)
152
+ 1240 train 4.402435 (lr=2.8652e-03) (hash(x)=81635225)
153
+ 1250 train 4.235700 (lr=2.8601e-03) (hash(x)=82126376)
154
+ 1260 train 4.416636 (lr=2.8550e-03) (hash(x)=83737972)
155
+ 1270 train 4.321780 (lr=2.8497e-03) (hash(x)=81132345)
156
+ 1280 train 4.404555 (lr=2.8443e-03) (hash(x)=85850404)
157
+ 1290 train 4.421223 (lr=2.8389e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.3778
159
+ 1300 val perplexity 79.6639
160
+ 1300 train 4.391219 (lr=2.8333e-03) (hash(x)=77545187)
161
+ 1310 train 4.438640 (lr=2.8277e-03) (hash(x)=86412685)
162
+ 1320 train 4.339867 (lr=2.8220e-03) (hash(x)=86429640)
163
+ 1330 train 4.479367 (lr=2.8162e-03) (hash(x)=79645281)
164
+ 1340 train 4.422177 (lr=2.8103e-03) (hash(x)=82668541)
165
+ 1350 train 4.289153 (lr=2.8044e-03) (hash(x)=73627248)
166
+ 1360 train 4.386212 (lr=2.7983e-03) (hash(x)=74845133)
167
+ 1370 train 4.298538 (lr=2.7922e-03) (hash(x)=77850497)
168
+ 1380 train 4.316807 (lr=2.7860e-03) (hash(x)=76143954)
169
+ 1390 train 4.331377 (lr=2.7797e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.3547
171
+ 1400 val perplexity 77.8412
172
+ 1400 train 4.307620 (lr=2.7733e-03) (hash(x)=76010938)
173
+ 1410 train 4.352828 (lr=2.7668e-03) (hash(x)=81260300)
174
+ 1420 train 4.474426 (lr=2.7603e-03) (hash(x)=83803629)
175
+ 1430 train 4.292600 (lr=2.7536e-03) (hash(x)=83840110)
176
+ 1440 train 4.487009 (lr=2.7469e-03) (hash(x)=97711831)
177
+ 1450 train 4.357034 (lr=2.7401e-03) (hash(x)=81110139)
178
+ 1460 train 4.279075 (lr=2.7333e-03) (hash(x)=90942839)
179
+ 1470 train 4.325918 (lr=2.7263e-03) (hash(x)=80721564)
180
+ 1480 train 4.414485 (lr=2.7193e-03) (hash(x)=77852059)
181
+ 1490 train 4.308678 (lr=2.7121e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3051
183
+ 1500 val perplexity 74.0739
184
+ 1500 train 4.334611 (lr=2.7050e-03) (hash(x)=84496142)
185
+ 1510 train 4.274515 (lr=2.6977e-03) (hash(x)=70186729)
186
+ 1520 train 4.301567 (lr=2.6903e-03) (hash(x)=74854227)
187
+ 1530 train 4.203382 (lr=2.6829e-03) (hash(x)=77468161)
188
+ 1540 train 4.225653 (lr=2.6754e-03) (hash(x)=87853059)
189
+ 1550 train 4.153411 (lr=2.6678e-03) (hash(x)=71225436)
190
+ 1560 train 4.260757 (lr=2.6602e-03) (hash(x)=84072783)
191
+ 1570 train 4.309518 (lr=2.6525e-03) (hash(x)=72723098)
192
+ 1580 train 4.319462 (lr=2.6447e-03) (hash(x)=90409866)
193
+ 1590 train 4.394246 (lr=2.6368e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.2807
195
+ 1600 val perplexity 72.2910
196
+ 1600 train 4.381636 (lr=2.6289e-03) (hash(x)=77643862)
197
+ 1610 train 4.275207 (lr=2.6208e-03) (hash(x)=88862575)
198
+ 1620 train 4.307343 (lr=2.6128e-03) (hash(x)=84612581)
199
+ 1630 train 4.316410 (lr=2.6046e-03) (hash(x)=87075989)
200
+ 1640 train 4.205328 (lr=2.5964e-03) (hash(x)=88277361)
201
+ 1650 train 4.318647 (lr=2.5881e-03) (hash(x)=78750236)
202
+ 1660 train 4.246531 (lr=2.5797e-03) (hash(x)=82604581)
203
+ 1670 train 4.080985 (lr=2.5713e-03) (hash(x)=68482265)
204
+ 1680 train 4.202220 (lr=2.5628e-03) (hash(x)=75088835)
205
+ 1690 train 4.143069 (lr=2.5542e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.2782
207
+ 1700 val perplexity 72.1075
208
+ 1700 train 4.314699 (lr=2.5455e-03) (hash(x)=79986754)
209
+ 1710 train 4.308959 (lr=2.5368e-03) (hash(x)=83657930)
210
+ 1720 train 4.244353 (lr=2.5281e-03) (hash(x)=81754135)
211
+ 1730 train 4.326231 (lr=2.5192e-03) (hash(x)=78904427)
212
+ 1740 train 4.222290 (lr=2.5103e-03) (hash(x)=85920177)
213
+ 1750 train 4.271566 (lr=2.5014e-03) (hash(x)=86573211)
214
+ 1760 train 4.235229 (lr=2.4924e-03) (hash(x)=81737128)
215
+ 1770 train 4.220489 (lr=2.4833e-03) (hash(x)=77400968)
216
+ 1780 train 4.238130 (lr=2.4741e-03) (hash(x)=73545497)
217
+ 1790 train 4.491808 (lr=2.4649e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.2279
219
+ 1800 val perplexity 68.5714
220
+ 1800 train 4.334773 (lr=2.4556e-03) (hash(x)=87819781)
221
+ 1810 train 4.195046 (lr=2.4463e-03) (hash(x)=86870770)
222
+ 1820 train 4.203197 (lr=2.4369e-03) (hash(x)=82522211)
223
+ 1830 train 4.284912 (lr=2.4275e-03) (hash(x)=61947437)
224
+ 1840 train 4.167178 (lr=2.4180e-03) (hash(x)=79865406)
225
+ 1850 train 4.129744 (lr=2.4084e-03) (hash(x)=79828721)
226
+ 1860 train 4.267334 (lr=2.3988e-03) (hash(x)=80869571)
227
+ 1870 train 4.256584 (lr=2.3891e-03) (hash(x)=73780971)
228
+ 1880 train 4.220387 (lr=2.3794e-03) (hash(x)=79249549)
229
+ 1890 train 4.270810 (lr=2.3696e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.2037
231
+ 1900 val perplexity 66.9330
232
+ 1900 train 4.341752 (lr=2.3598e-03) (hash(x)=82456430)
233
+ 1910 train 4.110814 (lr=2.3499e-03) (hash(x)=82222135)
234
+ 1920 train 4.147085 (lr=2.3400e-03) (hash(x)=70033249)
235
+ 1930 train 4.134484 (lr=2.3300e-03) (hash(x)=72887360)
236
+ 1940 train 4.179896 (lr=2.3200e-03) (hash(x)=83251100)
237
+ 1950 train 4.215040 (lr=2.3099e-03) (hash(x)=79660266)
238
+ 1960 train 4.100243 (lr=2.2998e-03) (hash(x)=78919068)
239
+ 1970 train 4.199206 (lr=2.2896e-03) (hash(x)=94690431)
240
+ 1980 train 4.031239 (lr=2.2793e-03) (hash(x)=88153756)
241
+ 1990 train 4.087008 (lr=2.2691e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.1788
243
+ 2000 val perplexity 65.2891
244
+ 2000 train 4.193192 (lr=2.2588e-03) (hash(x)=81308591)
245
+ 2010 train 4.237267 (lr=2.2484e-03) (hash(x)=80205479)
246
+ 2020 train 4.151566 (lr=2.2380e-03) (hash(x)=81008704)
247
+ 2030 train 4.219313 (lr=2.2275e-03) (hash(x)=82914358)
248
+ 2040 train 4.231478 (lr=2.2170e-03) (hash(x)=88064399)
249
+ 2050 train 4.237382 (lr=2.2065e-03) (hash(x)=82889390)
250
+ 2060 train 4.302837 (lr=2.1959e-03) (hash(x)=80913554)
251
+ 2070 train 4.141653 (lr=2.1853e-03) (hash(x)=78121791)
252
+ 2080 train 4.276073 (lr=2.1746e-03) (hash(x)=75356657)
253
+ 2090 train 4.102517 (lr=2.1639e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.1520
255
+ 2100 val perplexity 63.5582
256
+ 2100 train 4.017426 (lr=2.1532e-03) (hash(x)=68928225)
257
+ 2110 train 4.071888 (lr=2.1424e-03) (hash(x)=93734745)
258
+ 2120 train 4.067456 (lr=2.1316e-03) (hash(x)=72878164)
259
+ 2130 train 4.198497 (lr=2.1208e-03) (hash(x)=71508226)
260
+ 2140 train 4.158771 (lr=2.1099e-03) (hash(x)=84168671)
261
+ 2150 train 4.085671 (lr=2.0990e-03) (hash(x)=74673639)
262
+ 2160 train 4.241895 (lr=2.0881e-03) (hash(x)=75470031)
263
+ 2170 train 4.193984 (lr=2.0771e-03) (hash(x)=74307890)
264
+ 2180 train 4.311902 (lr=2.0661e-03) (hash(x)=77214245)
265
+ 2190 train 4.092965 (lr=2.0550e-03) (hash(x)=88628359)
266
+ 2200 val loss 4.1269
267
+ 2200 val perplexity 61.9879
268
+ 2200 train 4.110831 (lr=2.0440e-03) (hash(x)=74779126)
269
+ 2210 train 4.147754 (lr=2.0329e-03) (hash(x)=79817976)
270
+ 2220 train 4.101039 (lr=2.0217e-03) (hash(x)=75258996)
271
+ 2230 train 4.072467 (lr=2.0106e-03) (hash(x)=88164047)
272
+ 2240 train 4.058809 (lr=1.9994e-03) (hash(x)=78580686)
273
+ 2250 train 3.991111 (lr=1.9882e-03) (hash(x)=76510617)
274
+ 2260 train 4.065361 (lr=1.9770e-03) (hash(x)=76043966)
275
+ 2270 train 4.048841 (lr=1.9657e-03) (hash(x)=73062098)
276
+ 2280 train 4.276148 (lr=1.9544e-03) (hash(x)=81885909)
277
+ 2290 train 4.210995 (lr=1.9431e-03) (hash(x)=81481741)
278
+ 2300 val loss 4.1026
279
+ 2300 val perplexity 60.4988
280
+ 2300 train 4.067750 (lr=1.9318e-03) (hash(x)=82104275)
281
+ 2310 train 4.113710 (lr=1.9205e-03) (hash(x)=82137309)
282
+ 2320 train 4.100666 (lr=1.9091e-03) (hash(x)=86741167)
283
+ 2330 train 4.109334 (lr=1.8977e-03) (hash(x)=84796263)
284
+ 2340 train 4.081412 (lr=1.8863e-03) (hash(x)=169966529)
285
+ 2350 train 3.954939 (lr=1.8749e-03) (hash(x)=80006061)
286
+ 2360 train 4.011212 (lr=1.8635e-03) (hash(x)=71090999)
287
+ 2370 train 4.140413 (lr=1.8520e-03) (hash(x)=86641850)
288
+ 2380 train 4.038366 (lr=1.8406e-03) (hash(x)=84573253)
289
+ 2390 train 4.075659 (lr=1.8291e-03) (hash(x)=74483764)
290
+ 2400 val loss 4.0958
291
+ 2400 val perplexity 60.0878
292
+ 2400 train 4.050037 (lr=1.8176e-03) (hash(x)=78327659)
293
+ 2410 train 4.116478 (lr=1.8061e-03) (hash(x)=77273627)
294
+ 2420 train 3.980122 (lr=1.7946e-03) (hash(x)=76938049)
295
+ 2430 train 3.980247 (lr=1.7830e-03) (hash(x)=98449442)
296
+ 2440 train 3.838728 (lr=1.7715e-03) (hash(x)=78157797)
297
+ 2450 train 3.878011 (lr=1.7600e-03) (hash(x)=80637582)
298
+ 2460 train 3.865471 (lr=1.7484e-03) (hash(x)=80023854)
299
+ 2470 train 3.928109 (lr=1.7368e-03) (hash(x)=86678884)
300
+ 2480 train 4.161680 (lr=1.7253e-03) (hash(x)=84086469)
301
+ 2490 train 4.173364 (lr=1.7137e-03) (hash(x)=83131141)
302
+ 2500 val loss 4.0696
303
+ 2500 val perplexity 58.5359
304
+ 2500 train 4.091081 (lr=1.7021e-03) (hash(x)=82583497)
305
+ 2510 train 4.155626 (lr=1.6906e-03) (hash(x)=79164326)
306
+ 2520 train 4.059134 (lr=1.6790e-03) (hash(x)=76456503)
307
+ 2530 train 4.114577 (lr=1.6674e-03) (hash(x)=82001228)
308
+ 2540 train 4.151861 (lr=1.6558e-03) (hash(x)=75615595)
309
+ 2550 train 4.098694 (lr=1.6442e-03) (hash(x)=78397869)
310
+ 2560 train 4.099042 (lr=1.6326e-03) (hash(x)=75547032)
311
+ 2570 train 4.007168 (lr=1.6210e-03) (hash(x)=89201025)
312
+ 2580 train 3.991453 (lr=1.6094e-03) (hash(x)=89856704)
313
+ 2590 train 4.008800 (lr=1.5979e-03) (hash(x)=82175682)
314
+ 2600 val loss 4.0498
315
+ 2600 val perplexity 57.3877
316
+ 2600 train 4.016345 (lr=1.5863e-03) (hash(x)=78312826)
317
+ 2610 train 4.003236 (lr=1.5747e-03) (hash(x)=77066588)
318
+ 2620 train 3.841129 (lr=1.5632e-03) (hash(x)=78666061)
319
+ 2630 train 3.838729 (lr=1.5516e-03) (hash(x)=93762143)
320
+ 2640 train 3.852389 (lr=1.5400e-03) (hash(x)=83191587)
321
+ 2650 train 3.773806 (lr=1.5285e-03) (hash(x)=87169585)
322
+ 2660 train 4.002090 (lr=1.5170e-03) (hash(x)=86426388)
323
+ 2670 train 4.135112 (lr=1.5054e-03) (hash(x)=76692638)
324
+ 2680 train 4.047455 (lr=1.4939e-03) (hash(x)=77446063)
325
+ 2690 train 4.046873 (lr=1.4824e-03) (hash(x)=79809050)
326
+ 2700 val loss 4.0435
327
+ 2700 val perplexity 57.0279
328
+ 2700 train 3.976405 (lr=1.4709e-03) (hash(x)=83116823)
329
+ 2710 train 4.085155 (lr=1.4594e-03) (hash(x)=75622148)
330
+ 2720 train 4.019269 (lr=1.4480e-03) (hash(x)=80690512)
331
+ 2730 train 4.089248 (lr=1.4365e-03) (hash(x)=78009984)
332
+ 2740 train 4.042257 (lr=1.4251e-03) (hash(x)=74205488)
333
+ 2750 train 4.049032 (lr=1.4137e-03) (hash(x)=91013332)
334
+ 2760 train 3.968406 (lr=1.4023e-03) (hash(x)=76954961)
335
+ 2770 train 3.975915 (lr=1.3909e-03) (hash(x)=79390317)
336
+ 2780 train 3.845468 (lr=1.3795e-03) (hash(x)=70168783)
337
+ 2790 train 4.010633 (lr=1.3682e-03) (hash(x)=76028417)
338
+ 2800 val loss 4.0193
339
+ 2800 val perplexity 55.6606
340
+ 2800 train 3.830142 (lr=1.3569e-03) (hash(x)=77656050)
341
+ 2810 train 3.880674 (lr=1.3456e-03) (hash(x)=94248216)
342
+ 2820 train 3.945477 (lr=1.3343e-03) (hash(x)=78305078)
343
+ 2830 train 3.853434 (lr=1.3230e-03) (hash(x)=79948848)
344
+ 2840 train 3.718903 (lr=1.3118e-03) (hash(x)=85341024)
345
+ 2850 train 4.028629 (lr=1.3006e-03) (hash(x)=78735170)
346
+ 2860 train 4.286232 (lr=1.2894e-03) (hash(x)=71616419)
347
+ 2870 train 4.045057 (lr=1.2783e-03) (hash(x)=78656517)
348
+ 2880 train 4.060170 (lr=1.2671e-03) (hash(x)=80073987)
349
+ 2890 train 3.990415 (lr=1.2560e-03) (hash(x)=76894809)
350
+ 2900 val loss 3.9943
351
+ 2900 val perplexity 54.2895
352
+ 2900 train 3.971694 (lr=1.2450e-03) (hash(x)=80499838)
353
+ 2910 train 4.021563 (lr=1.2339e-03) (hash(x)=72673354)
354
+ 2920 train 4.005082 (lr=1.2229e-03) (hash(x)=84265768)
355
+ 2930 train 4.034717 (lr=1.2119e-03) (hash(x)=79612060)
356
+ 2940 train 3.880700 (lr=1.2010e-03) (hash(x)=74970087)
357
+ 2950 train 4.041933 (lr=1.1901e-03) (hash(x)=84166818)
358
+ 2960 train 4.028996 (lr=1.1792e-03) (hash(x)=89410221)
359
+ 2970 train 3.964778 (lr=1.1684e-03) (hash(x)=75672566)
360
+ 2980 train 3.823784 (lr=1.1576e-03) (hash(x)=81760314)
361
+ 2990 train 3.797642 (lr=1.1468e-03) (hash(x)=80605200)
362
+ 3000 val loss 3.9923
363
+ 3000 val perplexity 54.1809
364
+ 3000 train 3.788697 (lr=1.1361e-03) (hash(x)=83804735)
365
+ 3010 train 3.714454 (lr=1.1254e-03) (hash(x)=77015303)
366
+ 3020 train 3.783775 (lr=1.1147e-03) (hash(x)=81464523)
367
+ 3030 train 3.981263 (lr=1.1041e-03) (hash(x)=80168230)
368
+ 3040 train 3.923797 (lr=1.0935e-03) (hash(x)=84817006)
369
+ 3050 train 4.033627 (lr=1.0830e-03) (hash(x)=71601811)
370
+ 3060 train 3.954431 (lr=1.0725e-03) (hash(x)=85499733)
371
+ 3070 train 4.013331 (lr=1.0620e-03) (hash(x)=78670408)
372
+ 3080 train 3.902878 (lr=1.0516e-03) (hash(x)=77120468)
373
+ 3090 train 4.010466 (lr=1.0412e-03) (hash(x)=77927426)
374
+ 3100 val loss 3.9656
375
+ 3100 val perplexity 52.7497
376
+ 3100 train 4.023462 (lr=1.0309e-03) (hash(x)=83998606)
377
+ 3110 train 3.920274 (lr=1.0207e-03) (hash(x)=77990218)
378
+ 3120 train 3.838694 (lr=1.0104e-03) (hash(x)=81623970)
379
+ 3130 train 3.810275 (lr=1.0002e-03) (hash(x)=75014781)
380
+ 3140 train 3.862657 (lr=9.9011e-04) (hash(x)=72591250)
381
+ 3150 train 3.879128 (lr=9.8002e-04) (hash(x)=81421847)
382
+ 3160 train 3.928387 (lr=9.6999e-04) (hash(x)=75112631)
383
+ 3170 train 4.023858 (lr=9.6000e-04) (hash(x)=87518033)
384
+ 3180 train 3.880911 (lr=9.5007e-04) (hash(x)=78648348)
385
+ 3190 train 3.990559 (lr=9.4019e-04) (hash(x)=78270029)
386
+ 3200 val loss 3.9473
387
+ 3200 val perplexity 51.7967
388
+ 3200 train 4.060269 (lr=9.3036e-04) (hash(x)=83380714)
389
+ 3210 train 4.247459 (lr=9.2058e-04) (hash(x)=67804991)
390
+ 3220 train 3.976124 (lr=9.1085e-04) (hash(x)=76968804)
391
+ 3230 train 3.904225 (lr=9.0118e-04) (hash(x)=75798670)
392
+ 3240 train 3.981728 (lr=8.9157e-04) (hash(x)=83192811)
393
+ 3250 train 3.920382 (lr=8.8201e-04) (hash(x)=81659789)
394
+ 3260 train 3.961310 (lr=8.7251e-04) (hash(x)=80622502)
395
+ 3270 train 3.936048 (lr=8.6307e-04) (hash(x)=85436511)
396
+ 3280 train 3.871230 (lr=8.5368e-04) (hash(x)=80589180)
397
+ 3290 train 4.010092 (lr=8.4435e-04) (hash(x)=85747193)
398
+ 3300 val loss 3.9384
399
+ 3300 val perplexity 51.3358
400
+ 3300 train 3.874314 (lr=8.3508e-04) (hash(x)=74635692)
401
+ 3310 train 3.864569 (lr=8.2588e-04) (hash(x)=82045455)
402
+ 3320 train 3.737459 (lr=8.1673e-04) (hash(x)=73249173)
403
+ 3330 train 3.696985 (lr=8.0764e-04) (hash(x)=71492338)
404
+ 3340 train 4.029667 (lr=7.9862e-04) (hash(x)=76374071)
405
+ 3350 train 3.996060 (lr=7.8966e-04) (hash(x)=80474064)
406
+ 3360 train 3.909472 (lr=7.8076e-04) (hash(x)=84390892)
407
+ 3370 train 3.951520 (lr=7.7192e-04) (hash(x)=83399949)
408
+ 3380 train 4.004305 (lr=7.6315e-04) (hash(x)=77648059)
409
+ 3390 train 3.976869 (lr=7.5445e-04) (hash(x)=80152701)
410
+ 3400 val loss 3.9147
411
+ 3400 val perplexity 50.1346
412
+ 3400 train 4.024065 (lr=7.4581e-04) (hash(x)=81952545)
413
+ 3410 train 3.987144 (lr=7.3724e-04) (hash(x)=80908993)
414
+ 3420 train 4.028808 (lr=7.2874e-04) (hash(x)=79610037)
415
+ 3430 train 3.872138 (lr=7.2030e-04) (hash(x)=87624382)
416
+ 3440 train 3.949775 (lr=7.1193e-04) (hash(x)=82336381)
417
+ 3450 train 3.927718 (lr=7.0363e-04) (hash(x)=87687835)
418
+ 3460 train 3.927873 (lr=6.9541e-04) (hash(x)=76376135)
419
+ 3470 train 3.840427 (lr=6.8725e-04) (hash(x)=79331391)
420
+ 3480 train 3.743590 (lr=6.7916e-04) (hash(x)=107002681)
421
+ 3490 train 3.828921 (lr=6.7114e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.9109
423
+ 3500 val perplexity 49.9436
424
+ 3500 train 3.786550 (lr=6.6320e-04) (hash(x)=88237229)
425
+ 3510 train 3.853591 (lr=6.5533e-04) (hash(x)=81011739)
426
+ 3520 train 3.901562 (lr=6.4753e-04) (hash(x)=64643427)
427
+ 3530 train 4.183866 (lr=6.3981e-04) (hash(x)=78029539)
428
+ 3540 train 4.012329 (lr=6.3216e-04) (hash(x)=83188968)
429
+ 3550 train 4.033752 (lr=6.2458e-04) (hash(x)=86104185)
430
+ 3560 train 3.959700 (lr=6.1708e-04) (hash(x)=81993629)
431
+ 3570 train 3.934817 (lr=6.0966e-04) (hash(x)=72719368)
432
+ 3580 train 3.933805 (lr=6.0231e-04) (hash(x)=90091487)
433
+ 3590 train 3.824438 (lr=5.9504e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.8892
435
+ 3600 val perplexity 48.8725
436
+ 3600 train 3.933207 (lr=5.8785e-04) (hash(x)=73103504)
437
+ 3610 train 3.939070 (lr=5.8074e-04) (hash(x)=84429400)
438
+ 3620 train 3.915471 (lr=5.7370e-04) (hash(x)=77021795)
439
+ 3630 train 3.867811 (lr=5.6675e-04) (hash(x)=92830605)
440
+ 3640 train 4.068911 (lr=5.5987e-04) (hash(x)=78313175)
441
+ 3650 train 3.864909 (lr=5.5308e-04) (hash(x)=85395549)
442
+ 3660 train 3.655353 (lr=5.4636e-04) (hash(x)=78114459)
443
+ 3670 train 3.783789 (lr=5.3973e-04) (hash(x)=74968316)
444
+ 3680 train 3.713686 (lr=5.3318e-04) (hash(x)=73358737)
445
+ 3690 train 3.810288 (lr=5.2671e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.8917
447
+ 3700 val perplexity 48.9943
448
+ 3700 train 3.818909 (lr=5.2033e-04) (hash(x)=74039273)
449
+ 3710 train 3.997716 (lr=5.1402e-04) (hash(x)=79299680)
450
+ 3720 train 3.904476 (lr=5.0780e-04) (hash(x)=94701498)
451
+ 3730 train 3.983106 (lr=5.0167e-04) (hash(x)=75352071)
452
+ 3740 train 3.863127 (lr=4.9562e-04) (hash(x)=77909487)
453
+ 3750 train 3.975137 (lr=4.8965e-04) (hash(x)=73986730)
454
+ 3760 train 3.928871 (lr=4.8377e-04) (hash(x)=79325763)
455
+ 3770 train 3.975177 (lr=4.7798e-04) (hash(x)=72457818)
456
+ 3780 train 4.095606 (lr=4.7227e-04) (hash(x)=71775590)
457
+ 3790 train 3.990363 (lr=4.6665e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.8732
459
+ 3800 val perplexity 48.0983
460
+ 3800 train 3.970338 (lr=4.6112e-04) (hash(x)=79965893)
461
+ 3810 train 3.905279 (lr=4.5567e-04) (hash(x)=72598235)
462
+ 3820 train 3.890053 (lr=4.5031e-04) (hash(x)=83113889)
463
+ 3830 train 3.945420 (lr=4.4504e-04) (hash(x)=74434590)
464
+ 3840 train 3.864025 (lr=4.3986e-04) (hash(x)=82860348)
465
+ 3850 train 3.803088 (lr=4.3477e-04) (hash(x)=78067565)
466
+ 3860 train 3.864648 (lr=4.2977e-04) (hash(x)=82592498)
467
+ 3870 train 3.795835 (lr=4.2486e-04) (hash(x)=81820733)
468
+ 3880 train 3.811324 (lr=4.2004e-04) (hash(x)=87709040)
469
+ 3890 train 3.860178 (lr=4.1530e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.8635
471
+ 3900 val perplexity 47.6341
472
+ 3900 train 3.982210 (lr=4.1066e-04) (hash(x)=76597431)
473
+ 3910 train 3.865528 (lr=4.0611e-04) (hash(x)=90490716)
474
+ 3920 train 3.969979 (lr=4.0166e-04) (hash(x)=81970659)
475
+ 3930 train 3.954739 (lr=3.9729e-04) (hash(x)=81496334)
476
+ 3940 train 3.931343 (lr=3.9302e-04) (hash(x)=75717605)
477
+ 3950 train 3.880025 (lr=3.8884e-04) (hash(x)=82749357)
478
+ 3960 train 3.905216 (lr=3.8475e-04) (hash(x)=89355157)
479
+ 3970 train 3.859373 (lr=3.8076e-04) (hash(x)=78980403)
480
+ 3980 train 3.835056 (lr=3.7685e-04) (hash(x)=76627217)
481
+ 3990 train 3.837644 (lr=3.7305e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.8575
483
+ 4000 val perplexity 47.3463
484
+ 4000 train 3.855161 (lr=3.6933e-04) (hash(x)=83018142)
485
+ 4010 train 3.860563 (lr=3.6572e-04) (hash(x)=81272436)
486
+ 4020 train 3.792856 (lr=3.6219e-04) (hash(x)=85497482)
487
+ 4030 train 3.856960 (lr=3.5876e-04) (hash(x)=85241734)
488
+ 4040 train 3.829192 (lr=3.5543e-04) (hash(x)=77925307)
489
+ 4050 train 3.828469 (lr=3.5219e-04) (hash(x)=84826179)
490
+ 4060 train 3.883486 (lr=3.4905e-04) (hash(x)=83606764)
491
+ 4070 train 3.802079 (lr=3.4600e-04) (hash(x)=80567590)
492
+ 4080 train 3.851406 (lr=3.4305e-04) (hash(x)=76860998)
493
+ 4090 train 3.889248 (lr=3.4019e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.8442
495
+ 4100 val perplexity 46.7201
496
+ 4100 train 3.931875 (lr=3.3744e-04) (hash(x)=82832041)
497
+ 4110 train 3.811171 (lr=3.3477e-04) (hash(x)=79143262)
498
+ 4120 train 3.837020 (lr=3.3221e-04) (hash(x)=77038149)
499
+ 4130 train 4.021562 (lr=3.2974e-04) (hash(x)=86339074)
500
+ 4140 train 4.085483 (lr=3.2737e-04) (hash(x)=76686216)
501
+ 4150 train 3.858521 (lr=3.2510e-04) (hash(x)=70522682)
502
+ 4160 train 3.820054 (lr=3.2292e-04) (hash(x)=90958555)
503
+ 4170 train 3.842227 (lr=3.2085e-04) (hash(x)=91463532)
504
+ 4180 train 3.839122 (lr=3.1887e-04) (hash(x)=81959329)
505
+ 4190 train 3.798010 (lr=3.1699e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.8484
507
+ 4200 val perplexity 46.9163
508
+ 4200 train 3.780458 (lr=3.1520e-04) (hash(x)=78361715)
509
+ 4210 train 3.928388 (lr=3.1352e-04) (hash(x)=87364889)
510
+ 4220 train 3.782473 (lr=3.1193e-04) (hash(x)=70465156)
511
+ 4230 train 3.936930 (lr=3.1044e-04) (hash(x)=84524081)
512
+ 4240 train 3.890481 (lr=3.0905e-04) (hash(x)=77824868)
513
+ 4250 train 3.904539 (lr=3.0776e-04) (hash(x)=81710711)
514
+ 4260 train 3.881053 (lr=3.0657e-04) (hash(x)=76362728)
515
+ 4270 train 3.880073 (lr=3.0548e-04) (hash(x)=83115208)
516
+ 4280 train 3.792924 (lr=3.0449e-04) (hash(x)=87218314)
517
+ 4290 train 3.945635 (lr=3.0359e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.8397
519
+ 4300 val perplexity 46.5093
520
+ 4300 train 3.822798 (lr=3.0280e-04) (hash(x)=77379615)
521
+ 4310 train 3.796120 (lr=3.0210e-04) (hash(x)=78669579)
522
+ 4320 train 3.911337 (lr=3.0150e-04) (hash(x)=83066608)
523
+ 4330 train 3.877347 (lr=3.0101e-04) (hash(x)=83037340)
524
+ 4340 train 3.912841 (lr=3.0061e-04) (hash(x)=82849771)
525
+ 4350 train 3.851682 (lr=3.0031e-04) (hash(x)=76693985)
526
+ 4360 train 3.939415 (lr=3.0011e-04) (hash(x)=77745394)
527
+ 4370 train 3.902475 (lr=3.0001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.8330
529
+ 4374 val perplexity 46.2017
scaling_sdpa_nh_2/sdpa_nh_128_seed_1339/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2e87b1470681a24dbecbdab7e8fd30bf2ca9cf323f26af2e470b6d3717a7de0
3
+ size 278988778
scaling_sdpa_nh_2/sdpa_nh_128_seed_1339/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f51ae57541bd69dfb13cfa7f8b31462d71c610353cc7a94abe24cc55452120d6
3
+ size 557999750