Upload sigma_lambda_v5/history.json with huggingface_hub
Browse files- sigma_lambda_v5/history.json +20 -8
sigma_lambda_v5/history.json
CHANGED
|
@@ -4,20 +4,23 @@
|
|
| 4 |
2,
|
| 5 |
3,
|
| 6 |
4,
|
| 7 |
-
5
|
|
|
|
| 8 |
],
|
| 9 |
"gen_loss": [
|
| 10 |
16.586591267879488,
|
| 11 |
22.012301083468923,
|
| 12 |
50.61838954157162,
|
| 13 |
114.44467188566325,
|
| 14 |
-
205.98175486637822
|
|
|
|
| 15 |
],
|
| 16 |
"ret_loss": [
|
| 17 |
0.0,
|
| 18 |
0.0,
|
| 19 |
0.0,
|
| 20 |
0.0,
|
|
|
|
| 21 |
0.0
|
| 22 |
],
|
| 23 |
"ppl": [
|
|
@@ -25,6 +28,7 @@
|
|
| 25 |
160.04577946159412,
|
| 26 |
15470.654177938472,
|
| 27 |
485165195.4097903,
|
|
|
|
| 28 |
485165195.4097903
|
| 29 |
],
|
| 30 |
"tf_acc": [
|
|
@@ -32,27 +36,31 @@
|
|
| 32 |
0.47643615492864355,
|
| 33 |
0.47478209936121774,
|
| 34 |
0.4664326184178535,
|
| 35 |
-
0.4302312730702071
|
|
|
|
| 36 |
],
|
| 37 |
"fr_acc": [
|
| 38 |
0.12302286153389094,
|
| 39 |
0.11233910085013026,
|
| 40 |
0.24578219982631747,
|
| 41 |
0.2376122651857946,
|
| 42 |
-
0.13397800116550115
|
|
|
|
| 43 |
],
|
| 44 |
"mrr": [
|
| 45 |
0.006059600971639156,
|
| 46 |
0.006043825298547745,
|
| 47 |
0.006044529378414154,
|
| 48 |
0.006051086820662022,
|
| 49 |
-
0.006043311208486557
|
|
|
|
| 50 |
],
|
| 51 |
"r1": [
|
| 52 |
0.0007812500116415322,
|
| 53 |
0.0007812500116415322,
|
| 54 |
0.0007812500116415322,
|
| 55 |
0.0007812500116415322,
|
|
|
|
| 56 |
0.0007812500116415322
|
| 57 |
],
|
| 58 |
"step_var": [
|
|
@@ -60,20 +68,23 @@
|
|
| 60 |
0.18193045258522034,
|
| 61 |
0.20552654564380646,
|
| 62 |
0.23860830068588257,
|
| 63 |
-
0.2775871157646179
|
|
|
|
| 64 |
],
|
| 65 |
"pos_slope": [
|
| 66 |
0.4985616207122803,
|
| 67 |
0.4935553967952728,
|
| 68 |
0.48772382736206055,
|
| 69 |
0.48204389214515686,
|
| 70 |
-
0.47680673003196716
|
|
|
|
| 71 |
],
|
| 72 |
"grad_norm": [
|
| 73 |
NaN,
|
| 74 |
NaN,
|
| 75 |
NaN,
|
| 76 |
NaN,
|
|
|
|
| 77 |
NaN
|
| 78 |
],
|
| 79 |
"w_ret": [
|
|
@@ -81,6 +92,7 @@
|
|
| 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 |
16.586591267879488,
|
| 12 |
22.012301083468923,
|
| 13 |
50.61838954157162,
|
| 14 |
114.44467188566325,
|
| 15 |
+
205.98175486637822,
|
| 16 |
+
333.0307641626024
|
| 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 |
160.04577946159412,
|
| 29 |
15470.654177938472,
|
| 30 |
485165195.4097903,
|
| 31 |
+
485165195.4097903,
|
| 32 |
485165195.4097903
|
| 33 |
],
|
| 34 |
"tf_acc": [
|
|
|
|
| 36 |
0.47643615492864355,
|
| 37 |
0.47478209936121774,
|
| 38 |
0.4664326184178535,
|
| 39 |
+
0.4302312730702071,
|
| 40 |
+
0.4411152534131013
|
| 41 |
],
|
| 42 |
"fr_acc": [
|
| 43 |
0.12302286153389094,
|
| 44 |
0.11233910085013026,
|
| 45 |
0.24578219982631747,
|
| 46 |
0.2376122651857946,
|
| 47 |
+
0.13397800116550115,
|
| 48 |
+
0.1528195558526441
|
| 49 |
],
|
| 50 |
"mrr": [
|
| 51 |
0.006059600971639156,
|
| 52 |
0.006043825298547745,
|
| 53 |
0.006044529378414154,
|
| 54 |
0.006051086820662022,
|
| 55 |
+
0.006043311208486557,
|
| 56 |
+
0.0060408106073737144
|
| 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 |
0.18193045258522034,
|
| 69 |
0.20552654564380646,
|
| 70 |
0.23860830068588257,
|
| 71 |
+
0.2775871157646179,
|
| 72 |
+
0.312656432390213
|
| 73 |
],
|
| 74 |
"pos_slope": [
|
| 75 |
0.4985616207122803,
|
| 76 |
0.4935553967952728,
|
| 77 |
0.48772382736206055,
|
| 78 |
0.48204389214515686,
|
| 79 |
+
0.47680673003196716,
|
| 80 |
+
0.47214341163635254
|
| 81 |
],
|
| 82 |
"grad_norm": [
|
| 83 |
NaN,
|
| 84 |
NaN,
|
| 85 |
NaN,
|
| 86 |
NaN,
|
| 87 |
+
NaN,
|
| 88 |
NaN
|
| 89 |
],
|
| 90 |
"w_ret": [
|
|
|
|
| 92 |
0.0,
|
| 93 |
0.0,
|
| 94 |
0.01,
|
| 95 |
+
0.02,
|
| 96 |
+
0.03
|
| 97 |
]
|
| 98 |
}
|