Commit
·
771390e
1
Parent(s):
78de417
Training in progress, step 500
Browse files- all_results.json +14 -0
- post-training eval_results.json +8 -0
- prediction_output.jsonl +465 -0
- pytorch_model.bin +1 -1
- train_results.json +8 -0
- trainer_state.json +265 -0
- training_args.bin +1 -1
all_results.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_samples": 465,
|
| 4 |
+
"test_f1": 0.43148798131355104,
|
| 5 |
+
"test_loss": 0.707080066204071,
|
| 6 |
+
"test_runtime": 2.722,
|
| 7 |
+
"test_samples_per_second": 170.83,
|
| 8 |
+
"test_steps_per_second": 34.166,
|
| 9 |
+
"train_loss": 0.9243036530092237,
|
| 10 |
+
"train_runtime": 435.0692,
|
| 11 |
+
"train_samples": 4056,
|
| 12 |
+
"train_samples_per_second": 27.968,
|
| 13 |
+
"train_steps_per_second": 5.599
|
| 14 |
+
}
|
post-training eval_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eval_samples": 465,
|
| 3 |
+
"test_f1": 0.43148798131355104,
|
| 4 |
+
"test_loss": 0.707080066204071,
|
| 5 |
+
"test_runtime": 2.722,
|
| 6 |
+
"test_samples_per_second": 170.83,
|
| 7 |
+
"test_steps_per_second": 34.166
|
| 8 |
+
}
|
prediction_output.jsonl
ADDED
|
@@ -0,0 +1,465 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[{"pred": -2.510410785675049, "label": 7.0}]
|
| 2 |
+
[{"pred": -3.2507312297821045, "label": 2.0}]
|
| 3 |
+
[{"pred": -3.24859356880188, "label": 2.0}]
|
| 4 |
+
[{"pred": -3.247575283050537, "label": 2.0}]
|
| 5 |
+
[{"pred": -3.2481486797332764, "label": 2.0}]
|
| 6 |
+
[{"pred": -3.2966833114624023, "label": 2.0}]
|
| 7 |
+
[{"pred": -3.2462878227233887, "label": 2.0}]
|
| 8 |
+
[{"pred": -1.9828402996063232, "label": 3.0}]
|
| 9 |
+
[{"pred": -3.253715753555298, "label": 2.0}]
|
| 10 |
+
[{"pred": -3.2591922283172607, "label": 2.0}]
|
| 11 |
+
[{"pred": -2.6352808475494385, "label": 2.0}]
|
| 12 |
+
[{"pred": -1.7829365730285645, "label": 3.0}]
|
| 13 |
+
[{"pred": -3.1377956867218018, "label": 6.0}]
|
| 14 |
+
[{"pred": -3.2481777667999268, "label": 2.0}]
|
| 15 |
+
[{"pred": -3.2511088848114014, "label": 2.0}]
|
| 16 |
+
[{"pred": -1.9706364870071411, "label": 6.0}]
|
| 17 |
+
[{"pred": -3.134568929672241, "label": 1.0}]
|
| 18 |
+
[{"pred": -1.9673999547958374, "label": 5.0}]
|
| 19 |
+
[{"pred": -1.9584543704986572, "label": 5.0}]
|
| 20 |
+
[{"pred": -1.9373170137405396, "label": 5.0}]
|
| 21 |
+
[{"pred": -3.2483811378479004, "label": 2.0}]
|
| 22 |
+
[{"pred": -3.286836862564087, "label": 2.0}]
|
| 23 |
+
[{"pred": -2.512089490890503, "label": 1.0}]
|
| 24 |
+
[{"pred": -1.5872291326522827, "label": 3.0}]
|
| 25 |
+
[{"pred": -3.0798821449279785, "label": 6.0}]
|
| 26 |
+
[{"pred": -2.494011640548706, "label": 6.0}]
|
| 27 |
+
[{"pred": -3.2540135383605957, "label": 2.0}]
|
| 28 |
+
[{"pred": -3.0015485286712646, "label": 6.0}]
|
| 29 |
+
[{"pred": -1.8807361125946045, "label": 3.0}]
|
| 30 |
+
[{"pred": -2.0784761905670166, "label": 5.0}]
|
| 31 |
+
[{"pred": -3.251410484313965, "label": 2.0}]
|
| 32 |
+
[{"pred": -3.2524852752685547, "label": 2.0}]
|
| 33 |
+
[{"pred": -2.9110841751098633, "label": 6.0}]
|
| 34 |
+
[{"pred": -3.2642126083374023, "label": 2.0}]
|
| 35 |
+
[{"pred": -2.2871689796447754, "label": 7.0}]
|
| 36 |
+
[{"pred": -1.9736731052398682, "label": 7.0}]
|
| 37 |
+
[{"pred": -2.0843722820281982, "label": 7.0}]
|
| 38 |
+
[{"pred": -1.9872080087661743, "label": 5.0}]
|
| 39 |
+
[{"pred": -3.2478902339935303, "label": 2.0}]
|
| 40 |
+
[{"pred": -1.9475884437561035, "label": 5.0}]
|
| 41 |
+
[{"pred": -3.2465603351593018, "label": 2.0}]
|
| 42 |
+
[{"pred": -1.4629011154174805, "label": 3.0}]
|
| 43 |
+
[{"pred": -1.9921798706054688, "label": 5.0}]
|
| 44 |
+
[{"pred": -1.9291678667068481, "label": 2.0}]
|
| 45 |
+
[{"pred": -2.140066623687744, "label": 5.0}]
|
| 46 |
+
[{"pred": -3.2536351680755615, "label": 2.0}]
|
| 47 |
+
[{"pred": -3.2501513957977295, "label": 2.0}]
|
| 48 |
+
[{"pred": -3.246175527572632, "label": 2.0}]
|
| 49 |
+
[{"pred": -3.249424695968628, "label": 2.0}]
|
| 50 |
+
[{"pred": -3.2488503456115723, "label": 2.0}]
|
| 51 |
+
[{"pred": -3.246807098388672, "label": 2.0}]
|
| 52 |
+
[{"pred": -3.2480902671813965, "label": 2.0}]
|
| 53 |
+
[{"pred": -3.246915102005005, "label": 2.0}]
|
| 54 |
+
[{"pred": -3.2511396408081055, "label": 2.0}]
|
| 55 |
+
[{"pred": -1.9705449342727661, "label": 5.0}]
|
| 56 |
+
[{"pred": -1.785231113433838, "label": 3.0}]
|
| 57 |
+
[{"pred": -3.134200096130371, "label": 6.0}]
|
| 58 |
+
[{"pred": -3.2488577365875244, "label": 2.0}]
|
| 59 |
+
[{"pred": -3.247814655303955, "label": 2.0}]
|
| 60 |
+
[{"pred": -3.195293426513672, "label": 2.0}]
|
| 61 |
+
[{"pred": -2.403369903564453, "label": 2.0}]
|
| 62 |
+
[{"pred": -3.24745774269104, "label": 2.0}]
|
| 63 |
+
[{"pred": -3.24977970123291, "label": 2.0}]
|
| 64 |
+
[{"pred": -3.2521111965179443, "label": 2.0}]
|
| 65 |
+
[{"pred": -3.248173952102661, "label": 2.0}]
|
| 66 |
+
[{"pred": -2.3274316787719727, "label": 1.0}]
|
| 67 |
+
[{"pred": -2.9034299850463867, "label": 4.0}]
|
| 68 |
+
[{"pred": -3.137951374053955, "label": 6.0}]
|
| 69 |
+
[{"pred": -2.8546154499053955, "label": 1.0}]
|
| 70 |
+
[{"pred": -3.2485239505767822, "label": 2.0}]
|
| 71 |
+
[{"pred": -1.9351248741149902, "label": 5.0}]
|
| 72 |
+
[{"pred": -1.9447581768035889, "label": 5.0}]
|
| 73 |
+
[{"pred": -3.2462339401245117, "label": 2.0}]
|
| 74 |
+
[{"pred": -2.7758700847625732, "label": 6.0}]
|
| 75 |
+
[{"pred": -1.5794405937194824, "label": 1.0}]
|
| 76 |
+
[{"pred": -3.130115270614624, "label": 6.0}]
|
| 77 |
+
[{"pred": -1.9553866386413574, "label": 5.0}]
|
| 78 |
+
[{"pred": -1.936835527420044, "label": 5.0}]
|
| 79 |
+
[{"pred": -3.2477223873138428, "label": 2.0}]
|
| 80 |
+
[{"pred": -3.2483720779418945, "label": 2.0}]
|
| 81 |
+
[{"pred": -3.2658145427703857, "label": 2.0}]
|
| 82 |
+
[{"pred": -3.2472360134124756, "label": 2.0}]
|
| 83 |
+
[{"pred": -3.2549071311950684, "label": 2.0}]
|
| 84 |
+
[{"pred": -3.25020694732666, "label": 2.0}]
|
| 85 |
+
[{"pred": -3.247680902481079, "label": 2.0}]
|
| 86 |
+
[{"pred": -3.2514231204986572, "label": 2.0}]
|
| 87 |
+
[{"pred": -3.250274896621704, "label": 2.0}]
|
| 88 |
+
[{"pred": -1.5630723237991333, "label": 6.0}]
|
| 89 |
+
[{"pred": -3.2489383220672607, "label": 2.0}]
|
| 90 |
+
[{"pred": -2.192727565765381, "label": 5.0}]
|
| 91 |
+
[{"pred": -3.068551540374756, "label": 1.0}]
|
| 92 |
+
[{"pred": -3.1351382732391357, "label": 1.0}]
|
| 93 |
+
[{"pred": -2.6512224674224854, "label": 1.0}]
|
| 94 |
+
[{"pred": -3.1275999546051025, "label": 6.0}]
|
| 95 |
+
[{"pred": -3.2466089725494385, "label": 2.0}]
|
| 96 |
+
[{"pred": -3.249180316925049, "label": 2.0}]
|
| 97 |
+
[{"pred": -2.2104737758636475, "label": 1.0}]
|
| 98 |
+
[{"pred": -3.249380111694336, "label": 2.0}]
|
| 99 |
+
[{"pred": -3.343925952911377, "label": 1.0}]
|
| 100 |
+
[{"pred": -3.251756191253662, "label": 2.0}]
|
| 101 |
+
[{"pred": -3.253300905227661, "label": 2.0}]
|
| 102 |
+
[{"pred": -2.5220165252685547, "label": 3.0}]
|
| 103 |
+
[{"pred": -1.5412929058074951, "label": 3.0}]
|
| 104 |
+
[{"pred": -2.1729302406311035, "label": 5.0}]
|
| 105 |
+
[{"pred": -3.252185583114624, "label": 2.0}]
|
| 106 |
+
[{"pred": -3.2534749507904053, "label": 2.0}]
|
| 107 |
+
[{"pred": -3.25490403175354, "label": 2.0}]
|
| 108 |
+
[{"pred": -3.249511241912842, "label": 2.0}]
|
| 109 |
+
[{"pred": -3.3549394607543945, "label": 2.0}]
|
| 110 |
+
[{"pred": -2.105039358139038, "label": 2.0}]
|
| 111 |
+
[{"pred": -3.247619152069092, "label": 2.0}]
|
| 112 |
+
[{"pred": -3.2496585845947266, "label": 2.0}]
|
| 113 |
+
[{"pred": -3.253883123397827, "label": 2.0}]
|
| 114 |
+
[{"pred": -3.139411449432373, "label": 6.0}]
|
| 115 |
+
[{"pred": -3.2470102310180664, "label": 2.0}]
|
| 116 |
+
[{"pred": -3.25020170211792, "label": 2.0}]
|
| 117 |
+
[{"pred": -3.24996280670166, "label": 2.0}]
|
| 118 |
+
[{"pred": -3.2462241649627686, "label": 2.0}]
|
| 119 |
+
[{"pred": -3.2490100860595703, "label": 2.0}]
|
| 120 |
+
[{"pred": -1.4974870681762695, "label": 2.0}]
|
| 121 |
+
[{"pred": -1.9251744747161865, "label": 5.0}]
|
| 122 |
+
[{"pred": -1.9782328605651855, "label": 6.0}]
|
| 123 |
+
[{"pred": -1.682198405265808, "label": 2.0}]
|
| 124 |
+
[{"pred": -3.2484428882598877, "label": 2.0}]
|
| 125 |
+
[{"pred": -1.7946439981460571, "label": 2.0}]
|
| 126 |
+
[{"pred": -2.9240119457244873, "label": 3.0}]
|
| 127 |
+
[{"pred": -1.5903632640838623, "label": 3.0}]
|
| 128 |
+
[{"pred": -2.835261106491089, "label": 6.0}]
|
| 129 |
+
[{"pred": -3.249758243560791, "label": 2.0}]
|
| 130 |
+
[{"pred": -3.275826930999756, "label": 2.0}]
|
| 131 |
+
[{"pred": -3.127063512802124, "label": 6.0}]
|
| 132 |
+
[{"pred": -3.2524330615997314, "label": 2.0}]
|
| 133 |
+
[{"pred": -3.039443254470825, "label": 6.0}]
|
| 134 |
+
[{"pred": -3.2478444576263428, "label": 2.0}]
|
| 135 |
+
[{"pred": -1.9380242824554443, "label": 5.0}]
|
| 136 |
+
[{"pred": -1.9762585163116455, "label": 5.0}]
|
| 137 |
+
[{"pred": -2.26621150970459, "label": 5.0}]
|
| 138 |
+
[{"pred": -1.9694421291351318, "label": 5.0}]
|
| 139 |
+
[{"pred": -2.02016282081604, "label": 5.0}]
|
| 140 |
+
[{"pred": -2.190138339996338, "label": 5.0}]
|
| 141 |
+
[{"pred": -3.247195243835449, "label": 2.0}]
|
| 142 |
+
[{"pred": -3.2507717609405518, "label": 2.0}]
|
| 143 |
+
[{"pred": -3.2475128173828125, "label": 2.0}]
|
| 144 |
+
[{"pred": -3.24788498878479, "label": 2.0}]
|
| 145 |
+
[{"pred": -3.247504472732544, "label": 2.0}]
|
| 146 |
+
[{"pred": -1.8066171407699585, "label": 3.0}]
|
| 147 |
+
[{"pred": -1.9345637559890747, "label": 5.0}]
|
| 148 |
+
[{"pred": -1.951025128364563, "label": 5.0}]
|
| 149 |
+
[{"pred": -2.899920701980591, "label": 5.0}]
|
| 150 |
+
[{"pred": -3.3853819370269775, "label": 3.0}]
|
| 151 |
+
[{"pred": -1.9470152854919434, "label": 5.0}]
|
| 152 |
+
[{"pred": -1.9361870288848877, "label": 5.0}]
|
| 153 |
+
[{"pred": -2.3150954246520996, "label": 6.0}]
|
| 154 |
+
[{"pred": -3.250133991241455, "label": 2.0}]
|
| 155 |
+
[{"pred": -3.247997999191284, "label": 2.0}]
|
| 156 |
+
[{"pred": -3.251189947128296, "label": 2.0}]
|
| 157 |
+
[{"pred": -3.249349594116211, "label": 2.0}]
|
| 158 |
+
[{"pred": -3.248683214187622, "label": 2.0}]
|
| 159 |
+
[{"pred": -3.249330520629883, "label": 2.0}]
|
| 160 |
+
[{"pred": -3.2471389770507812, "label": 2.0}]
|
| 161 |
+
[{"pred": -2.0028741359710693, "label": 5.0}]
|
| 162 |
+
[{"pred": -2.300469398498535, "label": 6.0}]
|
| 163 |
+
[{"pred": -1.9600811004638672, "label": 5.0}]
|
| 164 |
+
[{"pred": -3.247410297393799, "label": 2.0}]
|
| 165 |
+
[{"pred": -2.8408918380737305, "label": 6.0}]
|
| 166 |
+
[{"pred": -3.115978956222534, "label": 6.0}]
|
| 167 |
+
[{"pred": -1.626998782157898, "label": 3.0}]
|
| 168 |
+
[{"pred": -3.248053789138794, "label": 2.0}]
|
| 169 |
+
[{"pred": -3.252047061920166, "label": 2.0}]
|
| 170 |
+
[{"pred": -3.259648561477661, "label": 2.0}]
|
| 171 |
+
[{"pred": -3.24615478515625, "label": 2.0}]
|
| 172 |
+
[{"pred": -3.2491531372070312, "label": 2.0}]
|
| 173 |
+
[{"pred": -3.2482364177703857, "label": 2.0}]
|
| 174 |
+
[{"pred": -3.2495648860931396, "label": 2.0}]
|
| 175 |
+
[{"pred": -2.1974339485168457, "label": 5.0}]
|
| 176 |
+
[{"pred": -1.929152488708496, "label": 5.0}]
|
| 177 |
+
[{"pred": -3.2490999698638916, "label": 2.0}]
|
| 178 |
+
[{"pred": -3.147801637649536, "label": 6.0}]
|
| 179 |
+
[{"pred": -3.2500851154327393, "label": 2.0}]
|
| 180 |
+
[{"pred": -3.1389920711517334, "label": 6.0}]
|
| 181 |
+
[{"pred": -2.007840156555176, "label": 5.0}]
|
| 182 |
+
[{"pred": -2.1625771522521973, "label": 5.0}]
|
| 183 |
+
[{"pred": -2.0281076431274414, "label": 5.0}]
|
| 184 |
+
[{"pred": -3.250375509262085, "label": 2.0}]
|
| 185 |
+
[{"pred": -3.2485640048980713, "label": 2.0}]
|
| 186 |
+
[{"pred": -3.2500860691070557, "label": 2.0}]
|
| 187 |
+
[{"pred": -3.2488980293273926, "label": 2.0}]
|
| 188 |
+
[{"pred": -3.255521535873413, "label": 2.0}]
|
| 189 |
+
[{"pred": -3.1220204830169678, "label": 1.0}]
|
| 190 |
+
[{"pred": -2.26766037940979, "label": 5.0}]
|
| 191 |
+
[{"pred": -3.0557210445404053, "label": 1.0}]
|
| 192 |
+
[{"pred": -2.1336829662323, "label": 1.0}]
|
| 193 |
+
[{"pred": -2.2413415908813477, "label": 5.0}]
|
| 194 |
+
[{"pred": -3.253527879714966, "label": 2.0}]
|
| 195 |
+
[{"pred": -3.246445894241333, "label": 2.0}]
|
| 196 |
+
[{"pred": -2.9603288173675537, "label": 6.0}]
|
| 197 |
+
[{"pred": -2.018174648284912, "label": 5.0}]
|
| 198 |
+
[{"pred": -2.042717218399048, "label": 5.0}]
|
| 199 |
+
[{"pred": -3.2508528232574463, "label": 2.0}]
|
| 200 |
+
[{"pred": -3.2504591941833496, "label": 2.0}]
|
| 201 |
+
[{"pred": -3.2611947059631348, "label": 2.0}]
|
| 202 |
+
[{"pred": -2.181732416152954, "label": 1.0}]
|
| 203 |
+
[{"pred": -1.9423891305923462, "label": 5.0}]
|
| 204 |
+
[{"pred": -3.252652645111084, "label": 2.0}]
|
| 205 |
+
[{"pred": -2.1885130405426025, "label": 5.0}]
|
| 206 |
+
[{"pred": -3.2477614879608154, "label": 2.0}]
|
| 207 |
+
[{"pred": -3.2476255893707275, "label": 2.0}]
|
| 208 |
+
[{"pred": -2.9612512588500977, "label": 6.0}]
|
| 209 |
+
[{"pred": -3.2479593753814697, "label": 2.0}]
|
| 210 |
+
[{"pred": -3.137547016143799, "label": 6.0}]
|
| 211 |
+
[{"pred": -2.041658401489258, "label": 6.0}]
|
| 212 |
+
[{"pred": -1.9440947771072388, "label": 5.0}]
|
| 213 |
+
[{"pred": -1.9479633569717407, "label": 5.0}]
|
| 214 |
+
[{"pred": -1.952940821647644, "label": 5.0}]
|
| 215 |
+
[{"pred": -3.248323440551758, "label": 2.0}]
|
| 216 |
+
[{"pred": -3.1365714073181152, "label": 6.0}]
|
| 217 |
+
[{"pred": -3.2473509311676025, "label": 2.0}]
|
| 218 |
+
[{"pred": -3.2474281787872314, "label": 2.0}]
|
| 219 |
+
[{"pred": -3.2499606609344482, "label": 2.0}]
|
| 220 |
+
[{"pred": -3.2498722076416016, "label": 2.0}]
|
| 221 |
+
[{"pred": -3.280186176300049, "label": 2.0}]
|
| 222 |
+
[{"pred": -3.2550947666168213, "label": 2.0}]
|
| 223 |
+
[{"pred": -1.929916501045227, "label": 5.0}]
|
| 224 |
+
[{"pred": -3.250938653945923, "label": 2.0}]
|
| 225 |
+
[{"pred": -3.2531137466430664, "label": 2.0}]
|
| 226 |
+
[{"pred": -3.1380059719085693, "label": 6.0}]
|
| 227 |
+
[{"pred": -3.1469132900238037, "label": 6.0}]
|
| 228 |
+
[{"pred": -3.249847412109375, "label": 2.0}]
|
| 229 |
+
[{"pred": -2.036956548690796, "label": 6.0}]
|
| 230 |
+
[{"pred": -3.1337716579437256, "label": 6.0}]
|
| 231 |
+
[{"pred": -3.248302698135376, "label": 2.0}]
|
| 232 |
+
[{"pred": -3.2527787685394287, "label": 2.0}]
|
| 233 |
+
[{"pred": -3.261897325515747, "label": 2.0}]
|
| 234 |
+
[{"pred": -3.2500975131988525, "label": 2.0}]
|
| 235 |
+
[{"pred": -3.251213788986206, "label": 2.0}]
|
| 236 |
+
[{"pred": -3.1308774948120117, "label": 6.0}]
|
| 237 |
+
[{"pred": -3.2498226165771484, "label": 1.0}]
|
| 238 |
+
[{"pred": -1.965856671333313, "label": 5.0}]
|
| 239 |
+
[{"pred": -3.2477619647979736, "label": 2.0}]
|
| 240 |
+
[{"pred": -3.1267306804656982, "label": 6.0}]
|
| 241 |
+
[{"pred": -3.13918137550354, "label": 6.0}]
|
| 242 |
+
[{"pred": -2.059156894683838, "label": 5.0}]
|
| 243 |
+
[{"pred": -2.026672601699829, "label": 5.0}]
|
| 244 |
+
[{"pred": -1.9802623987197876, "label": 5.0}]
|
| 245 |
+
[{"pred": -2.278937816619873, "label": 5.0}]
|
| 246 |
+
[{"pred": -2.1523187160491943, "label": 5.0}]
|
| 247 |
+
[{"pred": -3.2486016750335693, "label": 2.0}]
|
| 248 |
+
[{"pred": -2.2251548767089844, "label": 6.0}]
|
| 249 |
+
[{"pred": -3.2477924823760986, "label": 2.0}]
|
| 250 |
+
[{"pred": -3.24784255027771, "label": 2.0}]
|
| 251 |
+
[{"pred": -3.247274160385132, "label": 2.0}]
|
| 252 |
+
[{"pred": -3.249107599258423, "label": 2.0}]
|
| 253 |
+
[{"pred": -3.305791139602661, "label": 2.0}]
|
| 254 |
+
[{"pred": -3.249842405319214, "label": 2.0}]
|
| 255 |
+
[{"pred": -3.2486002445220947, "label": 2.0}]
|
| 256 |
+
[{"pred": -1.9705396890640259, "label": 3.0}]
|
| 257 |
+
[{"pred": -1.9510258436203003, "label": 5.0}]
|
| 258 |
+
[{"pred": -2.0478053092956543, "label": 5.0}]
|
| 259 |
+
[{"pred": -2.159274101257324, "label": 1.0}]
|
| 260 |
+
[{"pred": -3.263702392578125, "label": 1.0}]
|
| 261 |
+
[{"pred": -3.2565793991088867, "label": 2.0}]
|
| 262 |
+
[{"pred": -2.0912601947784424, "label": 1.0}]
|
| 263 |
+
[{"pred": -3.01347017288208, "label": 6.0}]
|
| 264 |
+
[{"pred": -3.2485098838806152, "label": 2.0}]
|
| 265 |
+
[{"pred": -3.2487339973449707, "label": 2.0}]
|
| 266 |
+
[{"pred": -3.1117453575134277, "label": 6.0}]
|
| 267 |
+
[{"pred": -3.248967170715332, "label": 2.0}]
|
| 268 |
+
[{"pred": -1.9213099479675293, "label": 5.0}]
|
| 269 |
+
[{"pred": -1.9224870204925537, "label": 5.0}]
|
| 270 |
+
[{"pred": -2.1930792331695557, "label": 2.0}]
|
| 271 |
+
[{"pred": -2.510280132293701, "label": 1.0}]
|
| 272 |
+
[{"pred": -3.248014211654663, "label": 2.0}]
|
| 273 |
+
[{"pred": -3.2469801902770996, "label": 2.0}]
|
| 274 |
+
[{"pred": -3.146756410598755, "label": 6.0}]
|
| 275 |
+
[{"pred": -3.12420654296875, "label": 6.0}]
|
| 276 |
+
[{"pred": -3.082371711730957, "label": 1.0}]
|
| 277 |
+
[{"pred": -3.2954421043395996, "label": 2.0}]
|
| 278 |
+
[{"pred": -2.1249806880950928, "label": 6.0}]
|
| 279 |
+
[{"pred": -1.957226276397705, "label": 5.0}]
|
| 280 |
+
[{"pred": -3.2495791912078857, "label": 2.0}]
|
| 281 |
+
[{"pred": -2.007701873779297, "label": 5.0}]
|
| 282 |
+
[{"pred": -2.2248971462249756, "label": 5.0}]
|
| 283 |
+
[{"pred": -3.2481753826141357, "label": 2.0}]
|
| 284 |
+
[{"pred": -3.134796142578125, "label": 6.0}]
|
| 285 |
+
[{"pred": -3.142890214920044, "label": 6.0}]
|
| 286 |
+
[{"pred": -3.308443307876587, "label": 6.0}]
|
| 287 |
+
[{"pred": -3.2465927600860596, "label": 2.0}]
|
| 288 |
+
[{"pred": -3.146916151046753, "label": 6.0}]
|
| 289 |
+
[{"pred": -3.2495830059051514, "label": 2.0}]
|
| 290 |
+
[{"pred": -3.248112916946411, "label": 2.0}]
|
| 291 |
+
[{"pred": -3.151005506515503, "label": 6.0}]
|
| 292 |
+
[{"pred": -3.2692275047302246, "label": 2.0}]
|
| 293 |
+
[{"pred": -3.246873378753662, "label": 2.0}]
|
| 294 |
+
[{"pred": -3.1520209312438965, "label": 6.0}]
|
| 295 |
+
[{"pred": -3.2529990673065186, "label": 2.0}]
|
| 296 |
+
[{"pred": -2.9999866485595703, "label": 1.0}]
|
| 297 |
+
[{"pred": -2.1421968936920166, "label": 2.0}]
|
| 298 |
+
[{"pred": -3.25823712348938, "label": 2.0}]
|
| 299 |
+
[{"pred": -3.260989189147949, "label": 2.0}]
|
| 300 |
+
[{"pred": -3.2480430603027344, "label": 2.0}]
|
| 301 |
+
[{"pred": -3.1267504692077637, "label": 6.0}]
|
| 302 |
+
[{"pred": -2.4134018421173096, "label": 6.0}]
|
| 303 |
+
[{"pred": -1.643166422843933, "label": 3.0}]
|
| 304 |
+
[{"pred": -3.0922815799713135, "label": 6.0}]
|
| 305 |
+
[{"pred": -3.2940807342529297, "label": 1.0}]
|
| 306 |
+
[{"pred": -2.550189256668091, "label": 1.0}]
|
| 307 |
+
[{"pred": -2.1299567222595215, "label": 1.0}]
|
| 308 |
+
[{"pred": -2.490447759628296, "label": 1.0}]
|
| 309 |
+
[{"pred": -1.9428058862686157, "label": 5.0}]
|
| 310 |
+
[{"pred": -3.249814748764038, "label": 2.0}]
|
| 311 |
+
[{"pred": -3.1222150325775146, "label": 6.0}]
|
| 312 |
+
[{"pred": -3.2495522499084473, "label": 2.0}]
|
| 313 |
+
[{"pred": -3.2489893436431885, "label": 2.0}]
|
| 314 |
+
[{"pred": -3.2501721382141113, "label": 2.0}]
|
| 315 |
+
[{"pred": -3.2659971714019775, "label": 2.0}]
|
| 316 |
+
[{"pred": -2.3733012676239014, "label": 5.0}]
|
| 317 |
+
[{"pred": -3.141282796859741, "label": 6.0}]
|
| 318 |
+
[{"pred": -1.567196249961853, "label": 3.0}]
|
| 319 |
+
[{"pred": -3.2500994205474854, "label": 2.0}]
|
| 320 |
+
[{"pred": -1.928815245628357, "label": 6.0}]
|
| 321 |
+
[{"pred": -3.2483513355255127, "label": 2.0}]
|
| 322 |
+
[{"pred": -3.252784490585327, "label": 2.0}]
|
| 323 |
+
[{"pred": -3.2485344409942627, "label": 2.0}]
|
| 324 |
+
[{"pred": -3.256577253341675, "label": 2.0}]
|
| 325 |
+
[{"pred": -3.2486727237701416, "label": 2.0}]
|
| 326 |
+
[{"pred": -3.125859260559082, "label": 6.0}]
|
| 327 |
+
[{"pred": -1.5720689296722412, "label": 6.0}]
|
| 328 |
+
[{"pred": -1.5077656507492065, "label": 3.0}]
|
| 329 |
+
[{"pred": -3.257927179336548, "label": 2.0}]
|
| 330 |
+
[{"pred": -3.247222661972046, "label": 2.0}]
|
| 331 |
+
[{"pred": -3.2471203804016113, "label": 2.0}]
|
| 332 |
+
[{"pred": -3.2475616931915283, "label": 2.0}]
|
| 333 |
+
[{"pred": -3.248579502105713, "label": 2.0}]
|
| 334 |
+
[{"pred": -3.2495508193969727, "label": 2.0}]
|
| 335 |
+
[{"pred": -1.9948246479034424, "label": 2.0}]
|
| 336 |
+
[{"pred": -3.247645139694214, "label": 2.0}]
|
| 337 |
+
[{"pred": -3.250845193862915, "label": 2.0}]
|
| 338 |
+
[{"pred": -1.9166219234466553, "label": 5.0}]
|
| 339 |
+
[{"pred": -2.655898332595825, "label": 6.0}]
|
| 340 |
+
[{"pred": -1.9234628677368164, "label": 5.0}]
|
| 341 |
+
[{"pred": -3.127408504486084, "label": 6.0}]
|
| 342 |
+
[{"pred": -3.2998859882354736, "label": 2.0}]
|
| 343 |
+
[{"pred": -2.001486301422119, "label": 5.0}]
|
| 344 |
+
[{"pred": -1.4905157089233398, "label": 3.0}]
|
| 345 |
+
[{"pred": -1.9807993173599243, "label": 5.0}]
|
| 346 |
+
[{"pred": -3.250840187072754, "label": 2.0}]
|
| 347 |
+
[{"pred": -3.2522671222686768, "label": 2.0}]
|
| 348 |
+
[{"pred": -3.247701406478882, "label": 2.0}]
|
| 349 |
+
[{"pred": -3.2514350414276123, "label": 2.0}]
|
| 350 |
+
[{"pred": -2.071831464767456, "label": 6.0}]
|
| 351 |
+
[{"pred": -3.247842311859131, "label": 2.0}]
|
| 352 |
+
[{"pred": -2.3604042530059814, "label": 2.0}]
|
| 353 |
+
[{"pred": -3.252725124359131, "label": 2.0}]
|
| 354 |
+
[{"pred": -3.2494561672210693, "label": 2.0}]
|
| 355 |
+
[{"pred": -3.2502870559692383, "label": 2.0}]
|
| 356 |
+
[{"pred": -3.2481398582458496, "label": 2.0}]
|
| 357 |
+
[{"pred": -2.1085283756256104, "label": 1.0}]
|
| 358 |
+
[{"pred": -3.2482149600982666, "label": 2.0}]
|
| 359 |
+
[{"pred": -3.10147762298584, "label": 1.0}]
|
| 360 |
+
[{"pred": -3.2903101444244385, "label": 2.0}]
|
| 361 |
+
[{"pred": -3.2806971073150635, "label": 2.0}]
|
| 362 |
+
[{"pred": -3.246124744415283, "label": 2.0}]
|
| 363 |
+
[{"pred": -3.251708984375, "label": 2.0}]
|
| 364 |
+
[{"pred": -1.9767757654190063, "label": 5.0}]
|
| 365 |
+
[{"pred": -1.9813804626464844, "label": 5.0}]
|
| 366 |
+
[{"pred": -1.9311048984527588, "label": 5.0}]
|
| 367 |
+
[{"pred": -1.9166024923324585, "label": 5.0}]
|
| 368 |
+
[{"pred": -3.256704330444336, "label": 3.0}]
|
| 369 |
+
[{"pred": -3.2482004165649414, "label": 2.0}]
|
| 370 |
+
[{"pred": -1.7488489151000977, "label": 3.0}]
|
| 371 |
+
[{"pred": -3.131485939025879, "label": 6.0}]
|
| 372 |
+
[{"pred": -3.2568063735961914, "label": 2.0}]
|
| 373 |
+
[{"pred": -3.2773356437683105, "label": 2.0}]
|
| 374 |
+
[{"pred": -3.2846364974975586, "label": 1.0}]
|
| 375 |
+
[{"pred": -3.303809881210327, "label": 3.0}]
|
| 376 |
+
[{"pred": -3.123743772506714, "label": 6.0}]
|
| 377 |
+
[{"pred": -1.92914879322052, "label": 5.0}]
|
| 378 |
+
[{"pred": -3.248990774154663, "label": 2.0}]
|
| 379 |
+
[{"pred": -2.9005420207977295, "label": 2.0}]
|
| 380 |
+
[{"pred": -3.1083006858825684, "label": 6.0}]
|
| 381 |
+
[{"pred": -3.1410179138183594, "label": 6.0}]
|
| 382 |
+
[{"pred": -3.2529895305633545, "label": 2.0}]
|
| 383 |
+
[{"pred": -2.2781615257263184, "label": 1.0}]
|
| 384 |
+
[{"pred": -2.853590250015259, "label": 1.0}]
|
| 385 |
+
[{"pred": -1.926788330078125, "label": 5.0}]
|
| 386 |
+
[{"pred": -2.2261645793914795, "label": 1.0}]
|
| 387 |
+
[{"pred": -3.2489113807678223, "label": 2.0}]
|
| 388 |
+
[{"pred": -2.2386341094970703, "label": 1.0}]
|
| 389 |
+
[{"pred": -3.141995668411255, "label": 6.0}]
|
| 390 |
+
[{"pred": -2.5372889041900635, "label": 5.0}]
|
| 391 |
+
[{"pred": -1.9538708925247192, "label": 5.0}]
|
| 392 |
+
[{"pred": -3.248356342315674, "label": 2.0}]
|
| 393 |
+
[{"pred": -3.1415443420410156, "label": 6.0}]
|
| 394 |
+
[{"pred": -1.7529325485229492, "label": 6.0}]
|
| 395 |
+
[{"pred": -3.1263415813446045, "label": 6.0}]
|
| 396 |
+
[{"pred": -3.159956693649292, "label": 6.0}]
|
| 397 |
+
[{"pred": -2.330596446990967, "label": 5.0}]
|
| 398 |
+
[{"pred": -2.8585636615753174, "label": 6.0}]
|
| 399 |
+
[{"pred": -3.2500030994415283, "label": 2.0}]
|
| 400 |
+
[{"pred": -1.923240065574646, "label": 5.0}]
|
| 401 |
+
[{"pred": -3.280019998550415, "label": 5.0}]
|
| 402 |
+
[{"pred": -3.084291696548462, "label": 5.0}]
|
| 403 |
+
[{"pred": -3.1203529834747314, "label": 5.0}]
|
| 404 |
+
[{"pred": -3.2479312419891357, "label": 2.0}]
|
| 405 |
+
[{"pred": -3.2823798656463623, "label": 2.0}]
|
| 406 |
+
[{"pred": -3.019171953201294, "label": 6.0}]
|
| 407 |
+
[{"pred": -3.2457242012023926, "label": 2.0}]
|
| 408 |
+
[{"pred": -3.2500083446502686, "label": 2.0}]
|
| 409 |
+
[{"pred": -3.248155117034912, "label": 2.0}]
|
| 410 |
+
[{"pred": -3.1289079189300537, "label": 6.0}]
|
| 411 |
+
[{"pred": -3.2471864223480225, "label": 2.0}]
|
| 412 |
+
[{"pred": -3.2449839115142822, "label": 2.0}]
|
| 413 |
+
[{"pred": -3.03255033493042, "label": 1.0}]
|
| 414 |
+
[{"pred": -2.1193833351135254, "label": 6.0}]
|
| 415 |
+
[{"pred": -2.3545727729797363, "label": 5.0}]
|
| 416 |
+
[{"pred": -3.1231086254119873, "label": 6.0}]
|
| 417 |
+
[{"pred": -3.24906587600708, "label": 2.0}]
|
| 418 |
+
[{"pred": -1.934016227722168, "label": 5.0}]
|
| 419 |
+
[{"pred": -3.250458240509033, "label": 2.0}]
|
| 420 |
+
[{"pred": -3.2498672008514404, "label": 2.0}]
|
| 421 |
+
[{"pred": -3.3921172618865967, "label": 2.0}]
|
| 422 |
+
[{"pred": -3.2601516246795654, "label": 2.0}]
|
| 423 |
+
[{"pred": -3.1343483924865723, "label": 6.0}]
|
| 424 |
+
[{"pred": -2.5072219371795654, "label": 1.0}]
|
| 425 |
+
[{"pred": -3.2471389770507812, "label": 2.0}]
|
| 426 |
+
[{"pred": -3.1292591094970703, "label": 6.0}]
|
| 427 |
+
[{"pred": -2.005147695541382, "label": 5.0}]
|
| 428 |
+
[{"pred": -3.2468056678771973, "label": 2.0}]
|
| 429 |
+
[{"pred": -3.248399019241333, "label": 2.0}]
|
| 430 |
+
[{"pred": -1.9832806587219238, "label": 5.0}]
|
| 431 |
+
[{"pred": -1.747692584991455, "label": 6.0}]
|
| 432 |
+
[{"pred": -3.247807741165161, "label": 2.0}]
|
| 433 |
+
[{"pred": -2.0735886096954346, "label": 5.0}]
|
| 434 |
+
[{"pred": -3.1346797943115234, "label": 1.0}]
|
| 435 |
+
[{"pred": -3.2464802265167236, "label": 2.0}]
|
| 436 |
+
[{"pred": -3.2824432849884033, "label": 2.0}]
|
| 437 |
+
[{"pred": -3.24961519241333, "label": 2.0}]
|
| 438 |
+
[{"pred": -3.2528555393218994, "label": 2.0}]
|
| 439 |
+
[{"pred": -3.422175645828247, "label": 2.0}]
|
| 440 |
+
[{"pred": -3.126441240310669, "label": 5.0}]
|
| 441 |
+
[{"pred": -1.993464469909668, "label": 5.0}]
|
| 442 |
+
[{"pred": -3.364798069000244, "label": 6.0}]
|
| 443 |
+
[{"pred": -3.2479796409606934, "label": 2.0}]
|
| 444 |
+
[{"pred": -3.247882604598999, "label": 2.0}]
|
| 445 |
+
[{"pred": -3.227517604827881, "label": 2.0}]
|
| 446 |
+
[{"pred": -2.187645196914673, "label": 5.0}]
|
| 447 |
+
[{"pred": -3.248873233795166, "label": 2.0}]
|
| 448 |
+
[{"pred": -3.252424955368042, "label": 2.0}]
|
| 449 |
+
[{"pred": -3.2584445476531982, "label": 2.0}]
|
| 450 |
+
[{"pred": -3.1489944458007812, "label": 6.0}]
|
| 451 |
+
[{"pred": -3.2492735385894775, "label": 1.0}]
|
| 452 |
+
[{"pred": -3.261714220046997, "label": 2.0}]
|
| 453 |
+
[{"pred": -3.2480428218841553, "label": 2.0}]
|
| 454 |
+
[{"pred": -3.2717626094818115, "label": 1.0}]
|
| 455 |
+
[{"pred": -3.1274783611297607, "label": 6.0}]
|
| 456 |
+
[{"pred": -1.6254037618637085, "label": 2.0}]
|
| 457 |
+
[{"pred": -2.3063008785247803, "label": 5.0}]
|
| 458 |
+
[{"pred": -3.249427080154419, "label": 2.0}]
|
| 459 |
+
[{"pred": -1.936509609222412, "label": 5.0}]
|
| 460 |
+
[{"pred": -3.2561233043670654, "label": 7.0}]
|
| 461 |
+
[{"pred": -3.2576708793640137, "label": 2.0}]
|
| 462 |
+
[{"pred": -3.2466981410980225, "label": 2.0}]
|
| 463 |
+
[{"pred": -3.126117706298828, "label": 6.0}]
|
| 464 |
+
[{"pred": -1.9359198808670044, "label": 5.0}]
|
| 465 |
+
[{"pred": -1.9201380014419556, "label": 5.0}]
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 501046977
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b68840d05165626d2a80bd09abb444ebd675a7aaf9ed25a8a3a40ff07fc6898
|
| 3 |
size 501046977
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"train_loss": 0.9243036530092237,
|
| 4 |
+
"train_runtime": 435.0692,
|
| 5 |
+
"train_samples": 4056,
|
| 6 |
+
"train_samples_per_second": 27.968,
|
| 7 |
+
"train_steps_per_second": 5.599
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,265 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"global_step": 2436,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.12,
|
| 12 |
+
"eval_f1": 0.0956468424279583,
|
| 13 |
+
"eval_loss": 1.3489933013916016,
|
| 14 |
+
"eval_runtime": 12.0925,
|
| 15 |
+
"eval_samples_per_second": 38.454,
|
| 16 |
+
"eval_steps_per_second": 7.691,
|
| 17 |
+
"step": 100
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.25,
|
| 21 |
+
"eval_f1": 0.0956468424279583,
|
| 22 |
+
"eval_loss": 1.4751325845718384,
|
| 23 |
+
"eval_runtime": 2.7372,
|
| 24 |
+
"eval_samples_per_second": 169.883,
|
| 25 |
+
"eval_steps_per_second": 33.977,
|
| 26 |
+
"step": 200
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.37,
|
| 30 |
+
"eval_f1": 0.24267804768973908,
|
| 31 |
+
"eval_loss": 0.9687431454658508,
|
| 32 |
+
"eval_runtime": 2.7453,
|
| 33 |
+
"eval_samples_per_second": 169.381,
|
| 34 |
+
"eval_steps_per_second": 33.876,
|
| 35 |
+
"step": 300
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"epoch": 0.49,
|
| 39 |
+
"eval_f1": 0.189060606060606,
|
| 40 |
+
"eval_loss": 1.0624897480010986,
|
| 41 |
+
"eval_runtime": 2.7393,
|
| 42 |
+
"eval_samples_per_second": 169.753,
|
| 43 |
+
"eval_steps_per_second": 33.951,
|
| 44 |
+
"step": 400
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.62,
|
| 48 |
+
"learning_rate": 3.9737274220032846e-05,
|
| 49 |
+
"loss": 1.2336,
|
| 50 |
+
"step": 500
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.62,
|
| 54 |
+
"eval_f1": 0.1949172949812336,
|
| 55 |
+
"eval_loss": 1.0953644514083862,
|
| 56 |
+
"eval_runtime": 2.7252,
|
| 57 |
+
"eval_samples_per_second": 170.632,
|
| 58 |
+
"eval_steps_per_second": 34.126,
|
| 59 |
+
"step": 500
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 0.74,
|
| 63 |
+
"eval_f1": 0.3080140543853414,
|
| 64 |
+
"eval_loss": 0.9969261288642883,
|
| 65 |
+
"eval_runtime": 2.7252,
|
| 66 |
+
"eval_samples_per_second": 170.632,
|
| 67 |
+
"eval_steps_per_second": 34.126,
|
| 68 |
+
"step": 600
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.86,
|
| 72 |
+
"eval_f1": 0.3174607190495667,
|
| 73 |
+
"eval_loss": 0.917083203792572,
|
| 74 |
+
"eval_runtime": 2.7253,
|
| 75 |
+
"eval_samples_per_second": 170.626,
|
| 76 |
+
"eval_steps_per_second": 34.125,
|
| 77 |
+
"step": 700
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"epoch": 0.99,
|
| 81 |
+
"eval_f1": 0.3136389264934501,
|
| 82 |
+
"eval_loss": 0.9600306153297424,
|
| 83 |
+
"eval_runtime": 2.7547,
|
| 84 |
+
"eval_samples_per_second": 168.801,
|
| 85 |
+
"eval_steps_per_second": 33.76,
|
| 86 |
+
"step": 800
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 1.11,
|
| 90 |
+
"eval_f1": 0.3161058727288384,
|
| 91 |
+
"eval_loss": 0.9637317061424255,
|
| 92 |
+
"eval_runtime": 2.7379,
|
| 93 |
+
"eval_samples_per_second": 169.835,
|
| 94 |
+
"eval_steps_per_second": 33.967,
|
| 95 |
+
"step": 900
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.23,
|
| 99 |
+
"learning_rate": 2.947454844006568e-05,
|
| 100 |
+
"loss": 1.0269,
|
| 101 |
+
"step": 1000
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"epoch": 1.23,
|
| 105 |
+
"eval_f1": 0.32568520729341943,
|
| 106 |
+
"eval_loss": 0.9592134356498718,
|
| 107 |
+
"eval_runtime": 2.7586,
|
| 108 |
+
"eval_samples_per_second": 168.567,
|
| 109 |
+
"eval_steps_per_second": 33.713,
|
| 110 |
+
"step": 1000
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 1.35,
|
| 114 |
+
"eval_f1": 0.33416337186541306,
|
| 115 |
+
"eval_loss": 0.9117490649223328,
|
| 116 |
+
"eval_runtime": 2.7292,
|
| 117 |
+
"eval_samples_per_second": 170.381,
|
| 118 |
+
"eval_steps_per_second": 34.076,
|
| 119 |
+
"step": 1100
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"epoch": 1.48,
|
| 123 |
+
"eval_f1": 0.32049717164329167,
|
| 124 |
+
"eval_loss": 0.8890902996063232,
|
| 125 |
+
"eval_runtime": 2.7191,
|
| 126 |
+
"eval_samples_per_second": 171.014,
|
| 127 |
+
"eval_steps_per_second": 34.203,
|
| 128 |
+
"step": 1200
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 1.6,
|
| 132 |
+
"eval_f1": 0.33748089562391775,
|
| 133 |
+
"eval_loss": 0.8135806918144226,
|
| 134 |
+
"eval_runtime": 2.7319,
|
| 135 |
+
"eval_samples_per_second": 170.209,
|
| 136 |
+
"eval_steps_per_second": 34.042,
|
| 137 |
+
"step": 1300
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.72,
|
| 141 |
+
"eval_f1": 0.3299982086985783,
|
| 142 |
+
"eval_loss": 0.9675911664962769,
|
| 143 |
+
"eval_runtime": 2.7207,
|
| 144 |
+
"eval_samples_per_second": 170.911,
|
| 145 |
+
"eval_steps_per_second": 34.182,
|
| 146 |
+
"step": 1400
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"epoch": 1.85,
|
| 150 |
+
"learning_rate": 1.921182266009852e-05,
|
| 151 |
+
"loss": 0.8592,
|
| 152 |
+
"step": 1500
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 1.85,
|
| 156 |
+
"eval_f1": 0.3315730877046458,
|
| 157 |
+
"eval_loss": 0.8777754306793213,
|
| 158 |
+
"eval_runtime": 2.7307,
|
| 159 |
+
"eval_samples_per_second": 170.287,
|
| 160 |
+
"eval_steps_per_second": 34.057,
|
| 161 |
+
"step": 1500
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"epoch": 1.97,
|
| 165 |
+
"eval_f1": 0.33789886027350413,
|
| 166 |
+
"eval_loss": 0.840674102306366,
|
| 167 |
+
"eval_runtime": 2.7102,
|
| 168 |
+
"eval_samples_per_second": 171.575,
|
| 169 |
+
"eval_steps_per_second": 34.315,
|
| 170 |
+
"step": 1600
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 2.09,
|
| 174 |
+
"eval_f1": 0.33685244804747255,
|
| 175 |
+
"eval_loss": 0.8408811092376709,
|
| 176 |
+
"eval_runtime": 2.7395,
|
| 177 |
+
"eval_samples_per_second": 169.74,
|
| 178 |
+
"eval_steps_per_second": 33.948,
|
| 179 |
+
"step": 1700
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 2.22,
|
| 183 |
+
"eval_f1": 0.3342829560124293,
|
| 184 |
+
"eval_loss": 0.8817650675773621,
|
| 185 |
+
"eval_runtime": 2.7283,
|
| 186 |
+
"eval_samples_per_second": 170.435,
|
| 187 |
+
"eval_steps_per_second": 34.087,
|
| 188 |
+
"step": 1800
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.34,
|
| 192 |
+
"eval_f1": 0.3385578564861433,
|
| 193 |
+
"eval_loss": 0.9258896708488464,
|
| 194 |
+
"eval_runtime": 2.7323,
|
| 195 |
+
"eval_samples_per_second": 170.186,
|
| 196 |
+
"eval_steps_per_second": 34.037,
|
| 197 |
+
"step": 1900
|
| 198 |
+
},
|
| 199 |
+
{
|
| 200 |
+
"epoch": 2.46,
|
| 201 |
+
"learning_rate": 8.949096880131364e-06,
|
| 202 |
+
"loss": 0.7521,
|
| 203 |
+
"step": 2000
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"epoch": 2.46,
|
| 207 |
+
"eval_f1": 0.33802107795500247,
|
| 208 |
+
"eval_loss": 0.941909670829773,
|
| 209 |
+
"eval_runtime": 2.7353,
|
| 210 |
+
"eval_samples_per_second": 169.998,
|
| 211 |
+
"eval_steps_per_second": 34.0,
|
| 212 |
+
"step": 2000
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 2.59,
|
| 216 |
+
"eval_f1": 0.3473817078391378,
|
| 217 |
+
"eval_loss": 0.8049965500831604,
|
| 218 |
+
"eval_runtime": 2.7367,
|
| 219 |
+
"eval_samples_per_second": 169.913,
|
| 220 |
+
"eval_steps_per_second": 33.983,
|
| 221 |
+
"step": 2100
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 2.71,
|
| 225 |
+
"eval_f1": 0.4052810578970328,
|
| 226 |
+
"eval_loss": 0.7772806286811829,
|
| 227 |
+
"eval_runtime": 2.7323,
|
| 228 |
+
"eval_samples_per_second": 170.185,
|
| 229 |
+
"eval_steps_per_second": 34.037,
|
| 230 |
+
"step": 2200
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 2.83,
|
| 234 |
+
"eval_f1": 0.43365497482920284,
|
| 235 |
+
"eval_loss": 0.7114064693450928,
|
| 236 |
+
"eval_runtime": 2.7465,
|
| 237 |
+
"eval_samples_per_second": 169.308,
|
| 238 |
+
"eval_steps_per_second": 33.862,
|
| 239 |
+
"step": 2300
|
| 240 |
+
},
|
| 241 |
+
{
|
| 242 |
+
"epoch": 2.96,
|
| 243 |
+
"eval_f1": 0.43148798131355104,
|
| 244 |
+
"eval_loss": 0.7161591649055481,
|
| 245 |
+
"eval_runtime": 2.7392,
|
| 246 |
+
"eval_samples_per_second": 169.755,
|
| 247 |
+
"eval_steps_per_second": 33.951,
|
| 248 |
+
"step": 2400
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 3.0,
|
| 252 |
+
"step": 2436,
|
| 253 |
+
"total_flos": 1184978465021280.0,
|
| 254 |
+
"train_loss": 0.9243036530092237,
|
| 255 |
+
"train_runtime": 435.0692,
|
| 256 |
+
"train_samples_per_second": 27.968,
|
| 257 |
+
"train_steps_per_second": 5.599
|
| 258 |
+
}
|
| 259 |
+
],
|
| 260 |
+
"max_steps": 2436,
|
| 261 |
+
"num_train_epochs": 3,
|
| 262 |
+
"total_flos": 1184978465021280.0,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3899
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0319295226b2fc649a5ead1ea9658b8b4d4cb74efca0f997398c9b62bb39867
|
| 3 |
size 3899
|