Upload sigma_lambda_v3/history.json with huggingface_hub
Browse files- sigma_lambda_v3/history.json +23 -11
sigma_lambda_v3/history.json
CHANGED
|
@@ -4,76 +4,87 @@
|
|
| 4 |
2,
|
| 5 |
3,
|
| 6 |
4,
|
| 7 |
-
5
|
|
|
|
| 8 |
],
|
| 9 |
"gen_loss": [
|
| 10 |
30.399562595346868,
|
| 11 |
22.764286764869794,
|
| 12 |
15.396741087545925,
|
| 13 |
13.480450591629728,
|
| 14 |
-
13.051396704105676
|
|
|
|
| 15 |
],
|
| 16 |
"ret_loss": [
|
| 17 |
796116674.0579515,
|
| 18 |
5450725.260929077,
|
| 19 |
350043.1876737281,
|
| 20 |
108628.19653386119,
|
| 21 |
-
58865.573026343496
|
|
|
|
| 22 |
],
|
| 23 |
"ppl": [
|
| 24 |
48361.553911128496,
|
| 25 |
10405.292656771971,
|
| 26 |
1254.1789620978543,
|
| 27 |
264.47601817816616,
|
| 28 |
-
127.8872290404574
|
|
|
|
| 29 |
],
|
| 30 |
"tf_acc": [
|
| 31 |
9.734349599431513e-06,
|
| 32 |
0.39589275342567987,
|
| 33 |
0.4354401710649703,
|
| 34 |
0.44354888428129674,
|
| 35 |
-
0.4487827195825911
|
|
|
|
| 36 |
],
|
| 37 |
"fr_acc": [
|
| 38 |
0.0,
|
| 39 |
0.160561711572741,
|
| 40 |
0.23937194341606105,
|
| 41 |
0.23937194341606105,
|
| 42 |
-
0.23937194341606105
|
|
|
|
| 43 |
],
|
| 44 |
"mrr": [
|
| 45 |
0.007931923493742943,
|
| 46 |
0.008492429740726948,
|
| 47 |
0.00886960793286562,
|
| 48 |
0.009372583590447903,
|
| 49 |
-
0.010170594789087772
|
|
|
|
| 50 |
],
|
| 51 |
"r1": [
|
| 52 |
0.0010416667209938169,
|
| 53 |
0.0010416667209938169,
|
| 54 |
0.0010416667209938169,
|
| 55 |
0.0010416667209938169,
|
| 56 |
-
0.0010416667209938169
|
|
|
|
| 57 |
],
|
| 58 |
"step_var": [
|
| 59 |
0.15913955867290497,
|
| 60 |
0.1598370373249054,
|
| 61 |
0.16560138761997223,
|
| 62 |
0.17514024674892426,
|
| 63 |
-
0.18433114886283875
|
|
|
|
| 64 |
],
|
| 65 |
"pos_slope": [
|
| 66 |
0.4995734989643097,
|
| 67 |
0.4980151951313019,
|
| 68 |
0.4959738850593567,
|
| 69 |
0.49392592906951904,
|
| 70 |
-
0.4921555817127228
|
|
|
|
| 71 |
],
|
| 72 |
"grad_norm": [
|
| 73 |
2887832256709451.0,
|
| 74 |
NaN,
|
| 75 |
NaN,
|
| 76 |
NaN,
|
|
|
|
| 77 |
NaN
|
| 78 |
],
|
| 79 |
"w_ret": [
|
|
@@ -81,6 +92,7 @@
|
|
| 81 |
0.05,
|
| 82 |
0.05,
|
| 83 |
0.083,
|
| 84 |
-
0.116
|
|
|
|
| 85 |
]
|
| 86 |
}
|
|
|
|
| 4 |
2,
|
| 5 |
3,
|
| 6 |
4,
|
| 7 |
+
5,
|
| 8 |
+
6
|
| 9 |
],
|
| 10 |
"gen_loss": [
|
| 11 |
30.399562595346868,
|
| 12 |
22.764286764869794,
|
| 13 |
15.396741087545925,
|
| 14 |
13.480450591629728,
|
| 15 |
+
13.051396704105676,
|
| 16 |
+
12.889946385535268
|
| 17 |
],
|
| 18 |
"ret_loss": [
|
| 19 |
796116674.0579515,
|
| 20 |
5450725.260929077,
|
| 21 |
350043.1876737281,
|
| 22 |
108628.19653386119,
|
| 23 |
+
58865.573026343496,
|
| 24 |
+
34838.68130633002
|
| 25 |
],
|
| 26 |
"ppl": [
|
| 27 |
48361.553911128496,
|
| 28 |
10405.292656771971,
|
| 29 |
1254.1789620978543,
|
| 30 |
264.47601817816616,
|
| 31 |
+
127.8872290404574,
|
| 32 |
+
96.4789019458568
|
| 33 |
],
|
| 34 |
"tf_acc": [
|
| 35 |
9.734349599431513e-06,
|
| 36 |
0.39589275342567987,
|
| 37 |
0.4354401710649703,
|
| 38 |
0.44354888428129674,
|
| 39 |
+
0.4487827195825911,
|
| 40 |
+
0.4525466680943713
|
| 41 |
],
|
| 42 |
"fr_acc": [
|
| 43 |
0.0,
|
| 44 |
0.160561711572741,
|
| 45 |
0.23937194341606105,
|
| 46 |
0.23937194341606105,
|
| 47 |
+
0.23937194341606105,
|
| 48 |
+
0.21731311988664928
|
| 49 |
],
|
| 50 |
"mrr": [
|
| 51 |
0.007931923493742943,
|
| 52 |
0.008492429740726948,
|
| 53 |
0.00886960793286562,
|
| 54 |
0.009372583590447903,
|
| 55 |
+
0.010170594789087772,
|
| 56 |
+
0.011823299340903759
|
| 57 |
],
|
| 58 |
"r1": [
|
| 59 |
0.0010416667209938169,
|
| 60 |
0.0010416667209938169,
|
| 61 |
0.0010416667209938169,
|
| 62 |
0.0010416667209938169,
|
| 63 |
+
0.0010416667209938169,
|
| 64 |
+
0.0020833334419876337
|
| 65 |
],
|
| 66 |
"step_var": [
|
| 67 |
0.15913955867290497,
|
| 68 |
0.1598370373249054,
|
| 69 |
0.16560138761997223,
|
| 70 |
0.17514024674892426,
|
| 71 |
+
0.18433114886283875,
|
| 72 |
+
0.19140200316905975
|
| 73 |
],
|
| 74 |
"pos_slope": [
|
| 75 |
0.4995734989643097,
|
| 76 |
0.4980151951313019,
|
| 77 |
0.4959738850593567,
|
| 78 |
0.49392592906951904,
|
| 79 |
+
0.4921555817127228,
|
| 80 |
+
0.49082034826278687
|
| 81 |
],
|
| 82 |
"grad_norm": [
|
| 83 |
2887832256709451.0,
|
| 84 |
NaN,
|
| 85 |
NaN,
|
| 86 |
NaN,
|
| 87 |
+
NaN,
|
| 88 |
NaN
|
| 89 |
],
|
| 90 |
"w_ret": [
|
|
|
|
| 92 |
0.05,
|
| 93 |
0.05,
|
| 94 |
0.083,
|
| 95 |
+
0.116,
|
| 96 |
+
0.14900000000000002
|
| 97 |
]
|
| 98 |
}
|