Upload frozen checkpoint v10_v32_roberta_batch32_44
Browse files
runs/v10_v32_roberta_batch32_44/best.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a79a0a3988b979ddf23c15d2a8e314047516e6401d39fc069b28b46050a2ee25
|
| 3 |
+
size 497760441
|
runs/v10_v32_roberta_batch32_44/config.json
ADDED
|
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"name": "v32_roberta_batch32",
|
| 4 |
+
"backbone": "roberta",
|
| 5 |
+
"seed": 44,
|
| 6 |
+
"lr": 1e-05,
|
| 7 |
+
"head_lr": 0.0003,
|
| 8 |
+
"epochs": 10,
|
| 9 |
+
"batch": 32,
|
| 10 |
+
"accumulation": 1,
|
| 11 |
+
"sanity": false
|
| 12 |
+
},
|
| 13 |
+
"model_config": {
|
| 14 |
+
"backbone": "roberta",
|
| 15 |
+
"hidden": 128,
|
| 16 |
+
"head_seed": 44
|
| 17 |
+
},
|
| 18 |
+
"host": "\u6587\u707f\u5b81\u76845070",
|
| 19 |
+
"device": "NVIDIA GeForce RTX 5070",
|
| 20 |
+
"gpu": "0",
|
| 21 |
+
"trainable_parameters": 124417298,
|
| 22 |
+
"git": "8e558baaae1c0dd65e0effe535f79a75f91038db",
|
| 23 |
+
"dirty": true,
|
| 24 |
+
"source_sha256": {
|
| 25 |
+
"scripts/optimize_v10.py": "7ed96f3727e164d2a349661c2b968577629939e7926d5bc19af66b6a49558d66",
|
| 26 |
+
"scripts/backbone_v10.py": "f88b62676bb6439466561b452dda4d6019e35a24bca2cb14f2ad39f8d7a295cc",
|
| 27 |
+
"scripts/optimize_v3.py": "50ec304d17ad3e9eb032a507e18c64296e5fe30ecbe2ab482bdba58ff856430d",
|
| 28 |
+
"scripts/modeling.py": "d38b477e4ccffff6c1eff32822fcfc8d197ca166155c9b1c3eff582c860795ea",
|
| 29 |
+
"scripts/missing_v6.py": "ba39831015d25916adb36b9f790319b4a9a968d0cd4eff6efed12b79363516fc",
|
| 30 |
+
"pyproject.toml": "13237c5fe333803d6e59b51af45f5280d5f025b6e94ee8ec9daa8db3ab5d6641",
|
| 31 |
+
"uv.lock": "478ac5a17d913e06b42e5ba94c2a679b30f6173d1d5c68334aa622956c0ab481"
|
| 32 |
+
},
|
| 33 |
+
"selection": "mean MAE+1-MacroF1 over clean, legacy40, random20/40/60",
|
| 34 |
+
"preprocessing": "only observed original BERT-token IDs decoded then native-tokenized; no raw_text access; no extra truncation",
|
| 35 |
+
"selection_profiles": {
|
| 36 |
+
"r20": {
|
| 37 |
+
"rate": 0.2,
|
| 38 |
+
"position": "random",
|
| 39 |
+
"seed": 6100,
|
| 40 |
+
"modalities": [
|
| 41 |
+
0,
|
| 42 |
+
1,
|
| 43 |
+
2
|
| 44 |
+
]
|
| 45 |
+
},
|
| 46 |
+
"r40": {
|
| 47 |
+
"rate": 0.4,
|
| 48 |
+
"position": "random",
|
| 49 |
+
"seed": 6101,
|
| 50 |
+
"modalities": [
|
| 51 |
+
0,
|
| 52 |
+
1,
|
| 53 |
+
2
|
| 54 |
+
]
|
| 55 |
+
},
|
| 56 |
+
"r60": {
|
| 57 |
+
"rate": 0.6,
|
| 58 |
+
"position": "random",
|
| 59 |
+
"seed": 6102,
|
| 60 |
+
"modalities": [
|
| 61 |
+
0,
|
| 62 |
+
1,
|
| 63 |
+
2
|
| 64 |
+
]
|
| 65 |
+
}
|
| 66 |
+
},
|
| 67 |
+
"tuning": "full",
|
| 68 |
+
"class_weights": [
|
| 69 |
+
1.0379084348678589,
|
| 70 |
+
1.1722970008850098,
|
| 71 |
+
0.789794385433197
|
| 72 |
+
],
|
| 73 |
+
"pretrained": {
|
| 74 |
+
"repo": "FacebookAI/roberta-base",
|
| 75 |
+
"revision": "e2da8e2f811d1448a5b465c236feacd80ffbac7b",
|
| 76 |
+
"files": [
|
| 77 |
+
{
|
| 78 |
+
"file": "config.json",
|
| 79 |
+
"bytes": 481,
|
| 80 |
+
"sha256": "ef0185e2aae6e06c5f105a285006952c340e20c7dbf43c86ec82601b13fc45e9",
|
| 81 |
+
"hub_lfs_sha256": null
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"file": "tokenizer.json",
|
| 85 |
+
"bytes": 1355863,
|
| 86 |
+
"sha256": "847bbeab6174d66a88898f729d52fa8d355fafe1bea101cf960dd404581df70e",
|
| 87 |
+
"hub_lfs_sha256": null
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"file": "tokenizer_config.json",
|
| 91 |
+
"bytes": 25,
|
| 92 |
+
"sha256": "994f46754c5bf4014f1aa92d34b1374319c3a6b3f702105cd5b742beaecd18ce",
|
| 93 |
+
"hub_lfs_sha256": null
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"file": "vocab.json",
|
| 97 |
+
"bytes": 898823,
|
| 98 |
+
"sha256": "9e7f63c2d15d666b52e21d250d2e513b87c9b713cfa6987a82ed89e5e6e50655",
|
| 99 |
+
"hub_lfs_sha256": null
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"file": "merges.txt",
|
| 103 |
+
"bytes": 456318,
|
| 104 |
+
"sha256": "1ce1664773c50f3e0cc8842619a93edc4624525b728b188a9e0be33b7726adc5",
|
| 105 |
+
"hub_lfs_sha256": null
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"file": "model.safetensors",
|
| 109 |
+
"bytes": 498818054,
|
| 110 |
+
"sha256": "5bde1d28afb363d0103324efeb5afc8b2b397fe5e04beabb9b1ef355255ade81",
|
| 111 |
+
"hub_lfs_sha256": "5bde1d28afb363d0103324efeb5afc8b2b397fe5e04beabb9b1ef355255ade81"
|
| 112 |
+
}
|
| 113 |
+
],
|
| 114 |
+
"purpose": "general pretrained encoder; no sentiment-finetuned weights or external task data"
|
| 115 |
+
}
|
| 116 |
+
}
|
runs/v10_v32_roberta_batch32_44/history.json
ADDED
|
@@ -0,0 +1,432 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"epoch": 1,
|
| 4 |
+
"score": 1.0634023394747332,
|
| 5 |
+
"metrics": {
|
| 6 |
+
"clean": {
|
| 7 |
+
"accuracy": 0.6497252747252747,
|
| 8 |
+
"macro_f1": 0.593787872869087,
|
| 9 |
+
"weighted_f1": 0.6253826137923172,
|
| 10 |
+
"mae": 0.5807569026947021,
|
| 11 |
+
"pearson": 0.694477791558282
|
| 12 |
+
},
|
| 13 |
+
"missing": {
|
| 14 |
+
"accuracy": 0.6002747252747253,
|
| 15 |
+
"macro_f1": 0.5521917822297767,
|
| 16 |
+
"weighted_f1": 0.5865232610816189,
|
| 17 |
+
"mae": 0.6416024565696716,
|
| 18 |
+
"pearson": 0.5912174096167946
|
| 19 |
+
},
|
| 20 |
+
"r20": {
|
| 21 |
+
"accuracy": 0.6195054945054945,
|
| 22 |
+
"macro_f1": 0.5691640968558181,
|
| 23 |
+
"weighted_f1": 0.6017090524879563,
|
| 24 |
+
"mae": 0.6125735640525818,
|
| 25 |
+
"pearson": 0.6556717731030359
|
| 26 |
+
},
|
| 27 |
+
"r40": {
|
| 28 |
+
"accuracy": 0.5892857142857143,
|
| 29 |
+
"macro_f1": 0.5448793199037477,
|
| 30 |
+
"weighted_f1": 0.5766750482720572,
|
| 31 |
+
"mae": 0.629589855670929,
|
| 32 |
+
"pearson": 0.6303253597360091
|
| 33 |
+
},
|
| 34 |
+
"r60": {
|
| 35 |
+
"accuracy": 0.5796703296703297,
|
| 36 |
+
"macro_f1": 0.5427072666336876,
|
| 37 |
+
"weighted_f1": 0.5744615722900738,
|
| 38 |
+
"mae": 0.6552192568778992,
|
| 39 |
+
"pearson": 0.5379482427362036
|
| 40 |
+
}
|
| 41 |
+
},
|
| 42 |
+
"loss": 1.2600606189709957,
|
| 43 |
+
"elapsed_s": 17.363717794418335
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"epoch": 2,
|
| 47 |
+
"score": 1.0195845531114398,
|
| 48 |
+
"metrics": {
|
| 49 |
+
"clean": {
|
| 50 |
+
"accuracy": 0.6648351648351648,
|
| 51 |
+
"macro_f1": 0.6141338893614855,
|
| 52 |
+
"weighted_f1": 0.643647954587755,
|
| 53 |
+
"mae": 0.5581523180007935,
|
| 54 |
+
"pearson": 0.7142927957610115
|
| 55 |
+
},
|
| 56 |
+
"missing": {
|
| 57 |
+
"accuracy": 0.6263736263736264,
|
| 58 |
+
"macro_f1": 0.5852487288929139,
|
| 59 |
+
"weighted_f1": 0.6133397892724702,
|
| 60 |
+
"mae": 0.6253404021263123,
|
| 61 |
+
"pearson": 0.5987062623845231
|
| 62 |
+
},
|
| 63 |
+
"r20": {
|
| 64 |
+
"accuracy": 0.6442307692307693,
|
| 65 |
+
"macro_f1": 0.6055947725624069,
|
| 66 |
+
"weighted_f1": 0.6320650919206273,
|
| 67 |
+
"mae": 0.5888070464134216,
|
| 68 |
+
"pearson": 0.6652970031901431
|
| 69 |
+
},
|
| 70 |
+
"r40": {
|
| 71 |
+
"accuracy": 0.6057692307692307,
|
| 72 |
+
"macro_f1": 0.5647179915746117,
|
| 73 |
+
"weighted_f1": 0.5931234370507854,
|
| 74 |
+
"mae": 0.6023602485656738,
|
| 75 |
+
"pearson": 0.6449040619231868
|
| 76 |
+
},
|
| 77 |
+
"r60": {
|
| 78 |
+
"accuracy": 0.5879120879120879,
|
| 79 |
+
"macro_f1": 0.5488406018225004,
|
| 80 |
+
"weighted_f1": 0.5800009413198454,
|
| 81 |
+
"mae": 0.641798734664917,
|
| 82 |
+
"pearson": 0.5442305902714384
|
| 83 |
+
}
|
| 84 |
+
},
|
| 85 |
+
"loss": 1.0023173561720091,
|
| 86 |
+
"elapsed_s": 35.01407837867737
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 3,
|
| 90 |
+
"score": 1.0036799544188149,
|
| 91 |
+
"metrics": {
|
| 92 |
+
"clean": {
|
| 93 |
+
"accuracy": 0.6483516483516484,
|
| 94 |
+
"macro_f1": 0.6280905297400259,
|
| 95 |
+
"weighted_f1": 0.6488750917056947,
|
| 96 |
+
"mae": 0.5458493828773499,
|
| 97 |
+
"pearson": 0.7344467335818319
|
| 98 |
+
},
|
| 99 |
+
"missing": {
|
| 100 |
+
"accuracy": 0.5947802197802198,
|
| 101 |
+
"macro_f1": 0.578204415841273,
|
| 102 |
+
"weighted_f1": 0.6025707202718181,
|
| 103 |
+
"mae": 0.6197399497032166,
|
| 104 |
+
"pearson": 0.6058591162922138
|
| 105 |
+
},
|
| 106 |
+
"r20": {
|
| 107 |
+
"accuracy": 0.6401098901098901,
|
| 108 |
+
"macro_f1": 0.6213345792815094,
|
| 109 |
+
"weighted_f1": 0.6438881159440506,
|
| 110 |
+
"mae": 0.572092592716217,
|
| 111 |
+
"pearson": 0.6852942435888384
|
| 112 |
+
},
|
| 113 |
+
"r40": {
|
| 114 |
+
"accuracy": 0.603021978021978,
|
| 115 |
+
"macro_f1": 0.5856968496310827,
|
| 116 |
+
"weighted_f1": 0.6094342642752549,
|
| 117 |
+
"mae": 0.6139847636222839,
|
| 118 |
+
"pearson": 0.6333871850012605
|
| 119 |
+
},
|
| 120 |
+
"r60": {
|
| 121 |
+
"accuracy": 0.5673076923076923,
|
| 122 |
+
"macro_f1": 0.5539290419725202,
|
| 123 |
+
"weighted_f1": 0.5767697752558001,
|
| 124 |
+
"mae": 0.6339884996414185,
|
| 125 |
+
"pearson": 0.5510782784552561
|
| 126 |
+
}
|
| 127 |
+
},
|
| 128 |
+
"loss": 0.9174913574602003,
|
| 129 |
+
"elapsed_s": 54.600980043411255
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 4,
|
| 133 |
+
"score": 0.9849849863957261,
|
| 134 |
+
"metrics": {
|
| 135 |
+
"clean": {
|
| 136 |
+
"accuracy": 0.6538461538461539,
|
| 137 |
+
"macro_f1": 0.6166896864750208,
|
| 138 |
+
"weighted_f1": 0.6418664658911872,
|
| 139 |
+
"mae": 0.5257620215415955,
|
| 140 |
+
"pearson": 0.7391486913605819
|
| 141 |
+
},
|
| 142 |
+
"missing": {
|
| 143 |
+
"accuracy": 0.6236263736263736,
|
| 144 |
+
"macro_f1": 0.5893461203983694,
|
| 145 |
+
"weighted_f1": 0.6164726779969862,
|
| 146 |
+
"mae": 0.6012635827064514,
|
| 147 |
+
"pearson": 0.6182027121003307
|
| 148 |
+
},
|
| 149 |
+
"r20": {
|
| 150 |
+
"accuracy": 0.6510989010989011,
|
| 151 |
+
"macro_f1": 0.6179959429185493,
|
| 152 |
+
"weighted_f1": 0.6432933707886153,
|
| 153 |
+
"mae": 0.5604286789894104,
|
| 154 |
+
"pearson": 0.6898894357337805
|
| 155 |
+
},
|
| 156 |
+
"r40": {
|
| 157 |
+
"accuracy": 0.6195054945054945,
|
| 158 |
+
"macro_f1": 0.5859000101317924,
|
| 159 |
+
"weighted_f1": 0.6125845956963478,
|
| 160 |
+
"mae": 0.5918108820915222,
|
| 161 |
+
"pearson": 0.6596492917114534
|
| 162 |
+
},
|
| 163 |
+
"r60": {
|
| 164 |
+
"accuracy": 0.6002747252747253,
|
| 165 |
+
"macro_f1": 0.5760033101385015,
|
| 166 |
+
"weighted_f1": 0.6010604999296739,
|
| 167 |
+
"mae": 0.6315948367118835,
|
| 168 |
+
"pearson": 0.5469825802423437
|
| 169 |
+
}
|
| 170 |
+
},
|
| 171 |
+
"loss": 0.8609425308548402,
|
| 172 |
+
"elapsed_s": 73.39882183074951
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 5,
|
| 176 |
+
"score": 0.994751073731966,
|
| 177 |
+
"metrics": {
|
| 178 |
+
"clean": {
|
| 179 |
+
"accuracy": 0.6483516483516484,
|
| 180 |
+
"macro_f1": 0.6283666532369235,
|
| 181 |
+
"weighted_f1": 0.6501975655314041,
|
| 182 |
+
"mae": 0.5258708000183105,
|
| 183 |
+
"pearson": 0.7361649640287015
|
| 184 |
+
},
|
| 185 |
+
"missing": {
|
| 186 |
+
"accuracy": 0.5837912087912088,
|
| 187 |
+
"macro_f1": 0.5732878763023601,
|
| 188 |
+
"weighted_f1": 0.5933669593263703,
|
| 189 |
+
"mae": 0.604027271270752,
|
| 190 |
+
"pearson": 0.606806675548597
|
| 191 |
+
},
|
| 192 |
+
"r20": {
|
| 193 |
+
"accuracy": 0.6208791208791209,
|
| 194 |
+
"macro_f1": 0.601307488424307,
|
| 195 |
+
"weighted_f1": 0.6256986099866915,
|
| 196 |
+
"mae": 0.5621582269668579,
|
| 197 |
+
"pearson": 0.6807421637898636
|
| 198 |
+
},
|
| 199 |
+
"r40": {
|
| 200 |
+
"accuracy": 0.5975274725274725,
|
| 201 |
+
"macro_f1": 0.581905495276016,
|
| 202 |
+
"weighted_f1": 0.6057364737819678,
|
| 203 |
+
"mae": 0.5923528075218201,
|
| 204 |
+
"pearson": 0.6483605281318566
|
| 205 |
+
},
|
| 206 |
+
"r60": {
|
| 207 |
+
"accuracy": 0.5673076923076923,
|
| 208 |
+
"macro_f1": 0.5565728812116778,
|
| 209 |
+
"weighted_f1": 0.5782114344155215,
|
| 210 |
+
"mae": 0.630786657333374,
|
| 211 |
+
"pearson": 0.5394695986271718
|
| 212 |
+
}
|
| 213 |
+
},
|
| 214 |
+
"loss": 0.8139529072235678,
|
| 215 |
+
"elapsed_s": 92.33366656303406
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"epoch": 6,
|
| 219 |
+
"score": 0.9917021353398947,
|
| 220 |
+
"metrics": {
|
| 221 |
+
"clean": {
|
| 222 |
+
"accuracy": 0.6648351648351648,
|
| 223 |
+
"macro_f1": 0.6249562745771222,
|
| 224 |
+
"weighted_f1": 0.6520550518635511,
|
| 225 |
+
"mae": 0.5414300560951233,
|
| 226 |
+
"pearson": 0.7362543599482444
|
| 227 |
+
},
|
| 228 |
+
"missing": {
|
| 229 |
+
"accuracy": 0.6208791208791209,
|
| 230 |
+
"macro_f1": 0.5937816450043117,
|
| 231 |
+
"weighted_f1": 0.6165313448326253,
|
| 232 |
+
"mae": 0.6102417707443237,
|
| 233 |
+
"pearson": 0.608655408912323
|
| 234 |
+
},
|
| 235 |
+
"r20": {
|
| 236 |
+
"accuracy": 0.6552197802197802,
|
| 237 |
+
"macro_f1": 0.6235966183574879,
|
| 238 |
+
"weighted_f1": 0.6494078874024526,
|
| 239 |
+
"mae": 0.5633271932601929,
|
| 240 |
+
"pearson": 0.6877312254619227
|
| 241 |
+
},
|
| 242 |
+
"r40": {
|
| 243 |
+
"accuracy": 0.6195054945054945,
|
| 244 |
+
"macro_f1": 0.5916382659873443,
|
| 245 |
+
"weighted_f1": 0.6163134131086929,
|
| 246 |
+
"mae": 0.6062328815460205,
|
| 247 |
+
"pearson": 0.6360492630699497
|
| 248 |
+
},
|
| 249 |
+
"r60": {
|
| 250 |
+
"accuracy": 0.5934065934065934,
|
| 251 |
+
"macro_f1": 0.5678244751161418,
|
| 252 |
+
"weighted_f1": 0.5940878034628034,
|
| 253 |
+
"mae": 0.6390760540962219,
|
| 254 |
+
"pearson": 0.5330283836283981
|
| 255 |
+
}
|
| 256 |
+
},
|
| 257 |
+
"loss": 0.7783940011095778,
|
| 258 |
+
"elapsed_s": 110.75146150588989
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 7,
|
| 262 |
+
"score": 0.9766074775902893,
|
| 263 |
+
"metrics": {
|
| 264 |
+
"clean": {
|
| 265 |
+
"accuracy": 0.6648351648351648,
|
| 266 |
+
"macro_f1": 0.639050620976238,
|
| 267 |
+
"weighted_f1": 0.6601333794838934,
|
| 268 |
+
"mae": 0.5273420810699463,
|
| 269 |
+
"pearson": 0.7429371741074097
|
| 270 |
+
},
|
| 271 |
+
"missing": {
|
| 272 |
+
"accuracy": 0.6098901098901099,
|
| 273 |
+
"macro_f1": 0.58870874692351,
|
| 274 |
+
"weighted_f1": 0.6097805124642018,
|
| 275 |
+
"mae": 0.6085813641548157,
|
| 276 |
+
"pearson": 0.6127199461215123
|
| 277 |
+
},
|
| 278 |
+
"r20": {
|
| 279 |
+
"accuracy": 0.6538461538461539,
|
| 280 |
+
"macro_f1": 0.632804468569537,
|
| 281 |
+
"weighted_f1": 0.6537393490090454,
|
| 282 |
+
"mae": 0.553382933139801,
|
| 283 |
+
"pearson": 0.7007892086439809
|
| 284 |
+
},
|
| 285 |
+
"r40": {
|
| 286 |
+
"accuracy": 0.6181318681318682,
|
| 287 |
+
"macro_f1": 0.5993663577774052,
|
| 288 |
+
"weighted_f1": 0.6191286208752647,
|
| 289 |
+
"mae": 0.59934401512146,
|
| 290 |
+
"pearson": 0.644591262667922
|
| 291 |
+
},
|
| 292 |
+
"r60": {
|
| 293 |
+
"accuracy": 0.6016483516483516,
|
| 294 |
+
"macro_f1": 0.5833735938348825,
|
| 295 |
+
"weighted_f1": 0.6055280305468724,
|
| 296 |
+
"mae": 0.6376907825469971,
|
| 297 |
+
"pearson": 0.547209690592358
|
| 298 |
+
}
|
| 299 |
+
},
|
| 300 |
+
"loss": 0.7567573281649117,
|
| 301 |
+
"elapsed_s": 128.27667355537415
|
| 302 |
+
},
|
| 303 |
+
{
|
| 304 |
+
"epoch": 8,
|
| 305 |
+
"score": 0.9835097878741467,
|
| 306 |
+
"metrics": {
|
| 307 |
+
"clean": {
|
| 308 |
+
"accuracy": 0.6634615384615384,
|
| 309 |
+
"macro_f1": 0.6416795894498838,
|
| 310 |
+
"weighted_f1": 0.6622107309429609,
|
| 311 |
+
"mae": 0.5367588400840759,
|
| 312 |
+
"pearson": 0.7426415904105956
|
| 313 |
+
},
|
| 314 |
+
"missing": {
|
| 315 |
+
"accuracy": 0.6126373626373627,
|
| 316 |
+
"macro_f1": 0.5926248474904288,
|
| 317 |
+
"weighted_f1": 0.6143804447477298,
|
| 318 |
+
"mae": 0.6092569828033447,
|
| 319 |
+
"pearson": 0.611368513445317
|
| 320 |
+
},
|
| 321 |
+
"r20": {
|
| 322 |
+
"accuracy": 0.6483516483516484,
|
| 323 |
+
"macro_f1": 0.6274787307864269,
|
| 324 |
+
"weighted_f1": 0.6493604233284738,
|
| 325 |
+
"mae": 0.5631369352340698,
|
| 326 |
+
"pearson": 0.6979943986158078
|
| 327 |
+
},
|
| 328 |
+
"r40": {
|
| 329 |
+
"accuracy": 0.614010989010989,
|
| 330 |
+
"macro_f1": 0.5946268066219106,
|
| 331 |
+
"weighted_f1": 0.6166250042627032,
|
| 332 |
+
"mae": 0.603191077709198,
|
| 333 |
+
"pearson": 0.6439100633725229
|
| 334 |
+
},
|
| 335 |
+
"r60": {
|
| 336 |
+
"accuracy": 0.5892857142857143,
|
| 337 |
+
"macro_f1": 0.5720976310613479,
|
| 338 |
+
"weighted_f1": 0.5956883614309197,
|
| 339 |
+
"mae": 0.6337127089500427,
|
| 340 |
+
"pearson": 0.5424241862620975
|
| 341 |
+
}
|
| 342 |
+
},
|
| 343 |
+
"loss": 0.7283847855630322,
|
| 344 |
+
"elapsed_s": 146.49096965789795
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 9,
|
| 348 |
+
"score": 0.9805468632162617,
|
| 349 |
+
"metrics": {
|
| 350 |
+
"clean": {
|
| 351 |
+
"accuracy": 0.6593406593406593,
|
| 352 |
+
"macro_f1": 0.6453392169261846,
|
| 353 |
+
"weighted_f1": 0.6624258140790239,
|
| 354 |
+
"mae": 0.5289399027824402,
|
| 355 |
+
"pearson": 0.7390382982435463
|
| 356 |
+
},
|
| 357 |
+
"missing": {
|
| 358 |
+
"accuracy": 0.6016483516483516,
|
| 359 |
+
"macro_f1": 0.5893009377664109,
|
| 360 |
+
"weighted_f1": 0.6079532147307083,
|
| 361 |
+
"mae": 0.6070418357849121,
|
| 362 |
+
"pearson": 0.6079682954609357
|
| 363 |
+
},
|
| 364 |
+
"r20": {
|
| 365 |
+
"accuracy": 0.635989010989011,
|
| 366 |
+
"macro_f1": 0.6226160260876291,
|
| 367 |
+
"weighted_f1": 0.6409916359146436,
|
| 368 |
+
"mae": 0.5552302598953247,
|
| 369 |
+
"pearson": 0.6955526377603347
|
| 370 |
+
},
|
| 371 |
+
"r40": {
|
| 372 |
+
"accuracy": 0.603021978021978,
|
| 373 |
+
"macro_f1": 0.5877265113446019,
|
| 374 |
+
"weighted_f1": 0.6075361824105542,
|
| 375 |
+
"mae": 0.6007192730903625,
|
| 376 |
+
"pearson": 0.6404599784128497
|
| 377 |
+
},
|
| 378 |
+
"r60": {
|
| 379 |
+
"accuracy": 0.5892857142857143,
|
| 380 |
+
"macro_f1": 0.5793528942139093,
|
| 381 |
+
"weighted_f1": 0.5985736219547432,
|
| 382 |
+
"mae": 0.6351386308670044,
|
| 383 |
+
"pearson": 0.538765965484167
|
| 384 |
+
}
|
| 385 |
+
},
|
| 386 |
+
"loss": 0.704179537073474,
|
| 387 |
+
"elapsed_s": 163.5631959438324
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 10,
|
| 391 |
+
"score": 0.9865081086464688,
|
| 392 |
+
"metrics": {
|
| 393 |
+
"clean": {
|
| 394 |
+
"accuracy": 0.6634615384615384,
|
| 395 |
+
"macro_f1": 0.6447903183776292,
|
| 396 |
+
"weighted_f1": 0.6638476205988397,
|
| 397 |
+
"mae": 0.5317210555076599,
|
| 398 |
+
"pearson": 0.7395357801623986
|
| 399 |
+
},
|
| 400 |
+
"missing": {
|
| 401 |
+
"accuracy": 0.5989010989010989,
|
| 402 |
+
"macro_f1": 0.5826914496129646,
|
| 403 |
+
"weighted_f1": 0.6029404948140431,
|
| 404 |
+
"mae": 0.6076988577842712,
|
| 405 |
+
"pearson": 0.6087833555516328
|
| 406 |
+
},
|
| 407 |
+
"r20": {
|
| 408 |
+
"accuracy": 0.6414835164835165,
|
| 409 |
+
"macro_f1": 0.6232086198290845,
|
| 410 |
+
"weighted_f1": 0.6435387807431379,
|
| 411 |
+
"mae": 0.5565186142921448,
|
| 412 |
+
"pearson": 0.6966352308700348
|
| 413 |
+
},
|
| 414 |
+
"r40": {
|
| 415 |
+
"accuracy": 0.603021978021978,
|
| 416 |
+
"macro_f1": 0.5845217347908709,
|
| 417 |
+
"weighted_f1": 0.606201965862265,
|
| 418 |
+
"mae": 0.6016542911529541,
|
| 419 |
+
"pearson": 0.6426370460805673
|
| 420 |
+
},
|
| 421 |
+
"r60": {
|
| 422 |
+
"accuracy": 0.5810439560439561,
|
| 423 |
+
"macro_f1": 0.5640543753253199,
|
| 424 |
+
"weighted_f1": 0.586878362658513,
|
| 425 |
+
"mae": 0.6342142224311829,
|
| 426 |
+
"pearson": 0.5419232127251675
|
| 427 |
+
}
|
| 428 |
+
},
|
| 429 |
+
"loss": 0.6877186657669389,
|
| 430 |
+
"elapsed_s": 180.80053687095642
|
| 431 |
+
}
|
| 432 |
+
]
|
runs/v10_v32_roberta_batch32_44/validation.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"clean": {
|
| 3 |
+
"accuracy": 0.6648351648351648,
|
| 4 |
+
"macro_f1": 0.639050620976238,
|
| 5 |
+
"weighted_f1": 0.6601333794838934,
|
| 6 |
+
"mae": 0.5273420810699463,
|
| 7 |
+
"pearson": 0.7429371741074097
|
| 8 |
+
},
|
| 9 |
+
"missing": {
|
| 10 |
+
"accuracy": 0.6098901098901099,
|
| 11 |
+
"macro_f1": 0.58870874692351,
|
| 12 |
+
"weighted_f1": 0.6097805124642018,
|
| 13 |
+
"mae": 0.6085813641548157,
|
| 14 |
+
"pearson": 0.6127199461215123
|
| 15 |
+
},
|
| 16 |
+
"r20": {
|
| 17 |
+
"accuracy": 0.6538461538461539,
|
| 18 |
+
"macro_f1": 0.632804468569537,
|
| 19 |
+
"weighted_f1": 0.6537393490090454,
|
| 20 |
+
"mae": 0.553382933139801,
|
| 21 |
+
"pearson": 0.7007892086439809
|
| 22 |
+
},
|
| 23 |
+
"r40": {
|
| 24 |
+
"accuracy": 0.6181318681318682,
|
| 25 |
+
"macro_f1": 0.5993663577774052,
|
| 26 |
+
"weighted_f1": 0.6191286208752647,
|
| 27 |
+
"mae": 0.59934401512146,
|
| 28 |
+
"pearson": 0.644591262667922
|
| 29 |
+
},
|
| 30 |
+
"r60": {
|
| 31 |
+
"accuracy": 0.6016483516483516,
|
| 32 |
+
"macro_f1": 0.5833735938348825,
|
| 33 |
+
"weighted_f1": 0.6055280305468724,
|
| 34 |
+
"mae": 0.6376907825469971,
|
| 35 |
+
"pearson": 0.547209690592358
|
| 36 |
+
}
|
| 37 |
+
}
|