Upload sigma_lambda_v6/history.json with huggingface_hub
Browse files- sigma_lambda_v6/history.json +22 -10
sigma_lambda_v6/history.json
CHANGED
|
@@ -4,20 +4,23 @@
|
|
| 4 |
2,
|
| 5 |
3,
|
| 6 |
4,
|
| 7 |
-
5
|
|
|
|
| 8 |
],
|
| 9 |
"gen_loss": [
|
| 10 |
16.730865429819772,
|
| 11 |
10.815642579259727,
|
| 12 |
11.253040295459758,
|
| 13 |
12.599300429961511,
|
| 14 |
-
15.0517748842187
|
|
|
|
| 15 |
],
|
| 16 |
"ret_loss": [
|
| 17 |
0.0,
|
| 18 |
0.0,
|
| 19 |
0.0,
|
| 20 |
0.0,
|
|
|
|
| 21 |
0.0
|
| 22 |
],
|
| 23 |
"ppl": [
|
|
@@ -25,34 +28,39 @@
|
|
| 25 |
38.05776687300114,
|
| 26 |
32.30925492774381,
|
| 27 |
35.023953072291675,
|
| 28 |
-
45.514587863029114
|
|
|
|
| 29 |
],
|
| 30 |
"tf_acc": [
|
| 31 |
0.4522865752576864,
|
| 32 |
0.484411723533269,
|
| 33 |
0.49136870335305427,
|
| 34 |
0.4924278146573637,
|
| 35 |
-
0.4915731578831036
|
|
|
|
| 36 |
],
|
| 37 |
"fr_acc": [
|
| 38 |
0.12302286153389094,
|
| 39 |
0.10218952820055761,
|
| 40 |
0.23302846051007814,
|
| 41 |
0.25665842531651356,
|
| 42 |
-
0.20978342531651356
|
|
|
|
| 43 |
],
|
| 44 |
"mrr": [
|
| 45 |
0.006079697050154209,
|
| 46 |
0.00607482623308897,
|
| 47 |
0.0060824002139270306,
|
| 48 |
0.006060858257114887,
|
| 49 |
-
0.006081360392272472
|
|
|
|
| 50 |
],
|
| 51 |
"r1": [
|
| 52 |
0.0007812500116415322,
|
| 53 |
0.0007812500116415322,
|
| 54 |
0.0007812500116415322,
|
| 55 |
0.0007812500116415322,
|
|
|
|
| 56 |
0.0007812500116415322
|
| 57 |
],
|
| 58 |
"step_var": [
|
|
@@ -60,27 +68,31 @@
|
|
| 60 |
0.1678009033203125,
|
| 61 |
0.17375488579273224,
|
| 62 |
0.17968906462192535,
|
| 63 |
-
0.1862482726573944
|
|
|
|
| 64 |
],
|
| 65 |
"pos_slope": [
|
| 66 |
0.49914172291755676,
|
| 67 |
0.4964430332183838,
|
| 68 |
0.4936324656009674,
|
| 69 |
0.49108341336250305,
|
| 70 |
-
0.48867347836494446
|
|
|
|
| 71 |
],
|
| 72 |
"grad_norm": [
|
| 73 |
NaN,
|
| 74 |
NaN,
|
| 75 |
NaN,
|
| 76 |
NaN,
|
| 77 |
-
Infinity
|
|
|
|
| 78 |
],
|
| 79 |
"w_ret": [
|
| 80 |
0.0,
|
| 81 |
0.0,
|
| 82 |
0.0,
|
| 83 |
0.01,
|
| 84 |
-
0.02
|
|
|
|
| 85 |
]
|
| 86 |
}
|
|
|
|
| 4 |
2,
|
| 5 |
3,
|
| 6 |
4,
|
| 7 |
+
5,
|
| 8 |
+
6
|
| 9 |
],
|
| 10 |
"gen_loss": [
|
| 11 |
16.730865429819772,
|
| 12 |
10.815642579259727,
|
| 13 |
11.253040295459758,
|
| 14 |
12.599300429961511,
|
| 15 |
+
15.0517748842187,
|
| 16 |
+
17.54592911206563
|
| 17 |
],
|
| 18 |
"ret_loss": [
|
| 19 |
0.0,
|
| 20 |
0.0,
|
| 21 |
0.0,
|
| 22 |
0.0,
|
| 23 |
+
0.0,
|
| 24 |
0.0
|
| 25 |
],
|
| 26 |
"ppl": [
|
|
|
|
| 28 |
38.05776687300114,
|
| 29 |
32.30925492774381,
|
| 30 |
35.023953072291675,
|
| 31 |
+
45.514587863029114,
|
| 32 |
+
57.41471868476501
|
| 33 |
],
|
| 34 |
"tf_acc": [
|
| 35 |
0.4522865752576864,
|
| 36 |
0.484411723533269,
|
| 37 |
0.49136870335305427,
|
| 38 |
0.4924278146573637,
|
| 39 |
+
0.4915731578831036,
|
| 40 |
+
0.49396767039719436
|
| 41 |
],
|
| 42 |
"fr_acc": [
|
| 43 |
0.12302286153389094,
|
| 44 |
0.10218952820055761,
|
| 45 |
0.23302846051007814,
|
| 46 |
0.25665842531651356,
|
| 47 |
+
0.20978342531651356,
|
| 48 |
+
0.1863969874080168
|
| 49 |
],
|
| 50 |
"mrr": [
|
| 51 |
0.006079697050154209,
|
| 52 |
0.00607482623308897,
|
| 53 |
0.0060824002139270306,
|
| 54 |
0.006060858257114887,
|
| 55 |
+
0.006081360392272472,
|
| 56 |
+
0.00604696711525321
|
| 57 |
],
|
| 58 |
"r1": [
|
| 59 |
0.0007812500116415322,
|
| 60 |
0.0007812500116415322,
|
| 61 |
0.0007812500116415322,
|
| 62 |
0.0007812500116415322,
|
| 63 |
+
0.0007812500116415322,
|
| 64 |
0.0007812500116415322
|
| 65 |
],
|
| 66 |
"step_var": [
|
|
|
|
| 68 |
0.1678009033203125,
|
| 69 |
0.17375488579273224,
|
| 70 |
0.17968906462192535,
|
| 71 |
+
0.1862482726573944,
|
| 72 |
+
0.19229243695735931
|
| 73 |
],
|
| 74 |
"pos_slope": [
|
| 75 |
0.49914172291755676,
|
| 76 |
0.4964430332183838,
|
| 77 |
0.4936324656009674,
|
| 78 |
0.49108341336250305,
|
| 79 |
+
0.48867347836494446,
|
| 80 |
+
0.48643457889556885
|
| 81 |
],
|
| 82 |
"grad_norm": [
|
| 83 |
NaN,
|
| 84 |
NaN,
|
| 85 |
NaN,
|
| 86 |
NaN,
|
| 87 |
+
Infinity,
|
| 88 |
+
24505.10658412117
|
| 89 |
],
|
| 90 |
"w_ret": [
|
| 91 |
0.0,
|
| 92 |
0.0,
|
| 93 |
0.0,
|
| 94 |
0.01,
|
| 95 |
+
0.02,
|
| 96 |
+
0.03
|
| 97 |
]
|
| 98 |
}
|