Upload 10 files
Browse files- config.json +50 -0
- pytorch_model.bin +3 -0
- rng_state.pth +3 -0
- scheduler.pt +3 -0
- special_tokens_map.json +7 -0
- tokenizer.json +0 -0
- tokenizer_config.json +17 -0
- trainer_state.json +1780 -0
- training_args.bin +3 -0
- vocab.txt +0 -0
config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "pranaydeeps/Ancient-Greek-BERT",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"BertForTokenClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.1,
|
| 7 |
+
"classifier_dropout": null,
|
| 8 |
+
"gradient_checkpointing": false,
|
| 9 |
+
"hidden_act": "gelu",
|
| 10 |
+
"hidden_dropout_prob": 0.1,
|
| 11 |
+
"hidden_size": 768,
|
| 12 |
+
"id2label": {
|
| 13 |
+
"0": "O",
|
| 14 |
+
"1": "B-PER",
|
| 15 |
+
"2": "I-PER",
|
| 16 |
+
"3": "B-NORP",
|
| 17 |
+
"4": "I-NORP",
|
| 18 |
+
"5": "B-LOC",
|
| 19 |
+
"6": "I-LOC",
|
| 20 |
+
"7": "B-MISC",
|
| 21 |
+
"8": "I-MISC"
|
| 22 |
+
},
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 3072,
|
| 25 |
+
"label2id": {
|
| 26 |
+
"B-LOC": "5",
|
| 27 |
+
"B-MISC": "7",
|
| 28 |
+
"B-NORP": "3",
|
| 29 |
+
"B-PER": "1",
|
| 30 |
+
"I-LOC": "6",
|
| 31 |
+
"I-MISC": "8",
|
| 32 |
+
"I-NORP": "4",
|
| 33 |
+
"I-PER": "2",
|
| 34 |
+
"O": "0"
|
| 35 |
+
},
|
| 36 |
+
"layer_norm_eps": 1e-12,
|
| 37 |
+
"max_position_embeddings": 512,
|
| 38 |
+
"model_max_length": 512,
|
| 39 |
+
"model_type": "bert",
|
| 40 |
+
"num_attention_heads": 12,
|
| 41 |
+
"num_hidden_layers": 12,
|
| 42 |
+
"output_past": true,
|
| 43 |
+
"pad_token_id": 0,
|
| 44 |
+
"position_embedding_type": "absolute",
|
| 45 |
+
"torch_dtype": "float32",
|
| 46 |
+
"transformers_version": "4.28.0",
|
| 47 |
+
"type_vocab_size": 2,
|
| 48 |
+
"use_cache": true,
|
| 49 |
+
"vocab_size": 35000
|
| 50 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7576e106a61a6b0676d480d84d95f161f482dffd8931e8af977d859c3223f378
|
| 3 |
+
size 449423274
|
rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b6219ef5e5ef2db80e10c8ee60ef209beff25772cdd9af05319af889817bf9e
|
| 3 |
+
size 14244
|
scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74971fdf77702f2b62c524375aa1c538b11ad0b4c171649f0fd219c0757c7268
|
| 3 |
+
size 1064
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cls_token": "[CLS]",
|
| 3 |
+
"mask_token": "[MASK]",
|
| 4 |
+
"pad_token": "[PAD]",
|
| 5 |
+
"sep_token": "[SEP]",
|
| 6 |
+
"unk_token": "[UNK]"
|
| 7 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"clean_up_tokenization_spaces": true,
|
| 3 |
+
"cls_token": "[CLS]",
|
| 4 |
+
"do_basic_tokenize": true,
|
| 5 |
+
"do_lower_case": true,
|
| 6 |
+
"mask_token": "[MASK]",
|
| 7 |
+
"model_max_length": 512,
|
| 8 |
+
"never_split": null,
|
| 9 |
+
"pad_token": "[PAD]",
|
| 10 |
+
"padding": true,
|
| 11 |
+
"sep_token": "[SEP]",
|
| 12 |
+
"strip_accents": null,
|
| 13 |
+
"tokenize_chinese_chars": true,
|
| 14 |
+
"tokenizer_class": "BertTokenizer",
|
| 15 |
+
"truncation": true,
|
| 16 |
+
"unk_token": "[UNK]"
|
| 17 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1780 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 10.0,
|
| 5 |
+
"global_step": 21190,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.24,
|
| 12 |
+
"learning_rate": 1.952807928268051e-05,
|
| 13 |
+
"loss": 0.2656,
|
| 14 |
+
"step": 500
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.24,
|
| 18 |
+
"eval_LOC": {
|
| 19 |
+
"f1": 0.6786599504277605,
|
| 20 |
+
"number": 6349,
|
| 21 |
+
"precision": 0.6891848002598246,
|
| 22 |
+
"recall": 0.6684517246810522
|
| 23 |
+
},
|
| 24 |
+
"eval_MISC": {
|
| 25 |
+
"f1": 0.29338446788111217,
|
| 26 |
+
"number": 1602,
|
| 27 |
+
"precision": 0.6322314049586777,
|
| 28 |
+
"recall": 0.19101123595505617
|
| 29 |
+
},
|
| 30 |
+
"eval_NORP": {
|
| 31 |
+
"f1": 0.7726153601779048,
|
| 32 |
+
"number": 5690,
|
| 33 |
+
"precision": 0.7048253876249819,
|
| 34 |
+
"recall": 0.8548330404217926
|
| 35 |
+
},
|
| 36 |
+
"eval_PER": {
|
| 37 |
+
"f1": 0.848227848101266,
|
| 38 |
+
"number": 16183,
|
| 39 |
+
"precision": 0.8693001232405786,
|
| 40 |
+
"recall": 0.8281530000617933
|
| 41 |
+
},
|
| 42 |
+
"eval_loss": 0.1907876580953598,
|
| 43 |
+
"eval_overall_accuracy": 0.9462788899739584,
|
| 44 |
+
"eval_overall_f1": 0.7762656505171475,
|
| 45 |
+
"eval_overall_precision": 0.7878453038674034,
|
| 46 |
+
"eval_overall_recall": 0.7650214592274678,
|
| 47 |
+
"eval_runtime": 36.0943,
|
| 48 |
+
"eval_samples_per_second": 117.359,
|
| 49 |
+
"eval_steps_per_second": 14.684,
|
| 50 |
+
"step": 500
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.47,
|
| 54 |
+
"learning_rate": 1.905615856536102e-05,
|
| 55 |
+
"loss": 0.1899,
|
| 56 |
+
"step": 1000
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.47,
|
| 60 |
+
"eval_LOC": {
|
| 61 |
+
"f1": 0.7273017902813299,
|
| 62 |
+
"number": 6349,
|
| 63 |
+
"precision": 0.738276813240305,
|
| 64 |
+
"recall": 0.7166482910694597
|
| 65 |
+
},
|
| 66 |
+
"eval_MISC": {
|
| 67 |
+
"f1": 0.45315161839863716,
|
| 68 |
+
"number": 1602,
|
| 69 |
+
"precision": 0.7131367292225201,
|
| 70 |
+
"recall": 0.33208489388264667
|
| 71 |
+
},
|
| 72 |
+
"eval_NORP": {
|
| 73 |
+
"f1": 0.8094145636842554,
|
| 74 |
+
"number": 5690,
|
| 75 |
+
"precision": 0.7835170258266162,
|
| 76 |
+
"recall": 0.8370826010544815
|
| 77 |
+
},
|
| 78 |
+
"eval_PER": {
|
| 79 |
+
"f1": 0.8706320857275974,
|
| 80 |
+
"number": 16183,
|
| 81 |
+
"precision": 0.8752810392205845,
|
| 82 |
+
"recall": 0.8660322560711858
|
| 83 |
+
},
|
| 84 |
+
"eval_loss": 0.1607111096382141,
|
| 85 |
+
"eval_overall_accuracy": 0.9542287190755209,
|
| 86 |
+
"eval_overall_f1": 0.811233510131919,
|
| 87 |
+
"eval_overall_precision": 0.8227586206896552,
|
| 88 |
+
"eval_overall_recall": 0.8000268240343348,
|
| 89 |
+
"eval_runtime": 33.8998,
|
| 90 |
+
"eval_samples_per_second": 124.956,
|
| 91 |
+
"eval_steps_per_second": 15.634,
|
| 92 |
+
"step": 1000
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.71,
|
| 96 |
+
"learning_rate": 1.858423784804153e-05,
|
| 97 |
+
"loss": 0.1609,
|
| 98 |
+
"step": 1500
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.71,
|
| 102 |
+
"eval_LOC": {
|
| 103 |
+
"f1": 0.7361812892359516,
|
| 104 |
+
"number": 6349,
|
| 105 |
+
"precision": 0.7162222553254879,
|
| 106 |
+
"recall": 0.7572846117498818
|
| 107 |
+
},
|
| 108 |
+
"eval_MISC": {
|
| 109 |
+
"f1": 0.5344885038320558,
|
| 110 |
+
"number": 1602,
|
| 111 |
+
"precision": 0.5732666190135811,
|
| 112 |
+
"recall": 0.5006242197253433
|
| 113 |
+
},
|
| 114 |
+
"eval_NORP": {
|
| 115 |
+
"f1": 0.8192258613356019,
|
| 116 |
+
"number": 5690,
|
| 117 |
+
"precision": 0.7938994229183842,
|
| 118 |
+
"recall": 0.8462214411247804
|
| 119 |
+
},
|
| 120 |
+
"eval_PER": {
|
| 121 |
+
"f1": 0.8834497380892989,
|
| 122 |
+
"number": 16183,
|
| 123 |
+
"precision": 0.8916231355025489,
|
| 124 |
+
"recall": 0.8754248285237595
|
| 125 |
+
},
|
| 126 |
+
"eval_loss": 0.14290471374988556,
|
| 127 |
+
"eval_overall_accuracy": 0.9570109049479166,
|
| 128 |
+
"eval_overall_f1": 0.8212389380530973,
|
| 129 |
+
"eval_overall_precision": 0.8179338787999734,
|
| 130 |
+
"eval_overall_recall": 0.8245708154506438,
|
| 131 |
+
"eval_runtime": 33.8876,
|
| 132 |
+
"eval_samples_per_second": 125.001,
|
| 133 |
+
"eval_steps_per_second": 15.64,
|
| 134 |
+
"step": 1500
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"epoch": 0.94,
|
| 138 |
+
"learning_rate": 1.8112317130722038e-05,
|
| 139 |
+
"loss": 0.1468,
|
| 140 |
+
"step": 2000
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 0.94,
|
| 144 |
+
"eval_LOC": {
|
| 145 |
+
"f1": 0.7594586281144263,
|
| 146 |
+
"number": 6349,
|
| 147 |
+
"precision": 0.7419984973703982,
|
| 148 |
+
"recall": 0.7777602772090093
|
| 149 |
+
},
|
| 150 |
+
"eval_MISC": {
|
| 151 |
+
"f1": 0.5709592641261497,
|
| 152 |
+
"number": 1602,
|
| 153 |
+
"precision": 0.602635228848821,
|
| 154 |
+
"recall": 0.5424469413233458
|
| 155 |
+
},
|
| 156 |
+
"eval_NORP": {
|
| 157 |
+
"f1": 0.8396145241782826,
|
| 158 |
+
"number": 5690,
|
| 159 |
+
"precision": 0.8224881995954147,
|
| 160 |
+
"recall": 0.8574692442882249
|
| 161 |
+
},
|
| 162 |
+
"eval_PER": {
|
| 163 |
+
"f1": 0.8936196585169348,
|
| 164 |
+
"number": 16183,
|
| 165 |
+
"precision": 0.8962580805569369,
|
| 166 |
+
"recall": 0.8909967249582895
|
| 167 |
+
},
|
| 168 |
+
"eval_loss": 0.13329361379146576,
|
| 169 |
+
"eval_overall_accuracy": 0.9607187906901041,
|
| 170 |
+
"eval_overall_f1": 0.8376570294122554,
|
| 171 |
+
"eval_overall_precision": 0.8335823621210612,
|
| 172 |
+
"eval_overall_recall": 0.8417717274678111,
|
| 173 |
+
"eval_runtime": 35.3996,
|
| 174 |
+
"eval_samples_per_second": 119.662,
|
| 175 |
+
"eval_steps_per_second": 14.972,
|
| 176 |
+
"step": 2000
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 1.18,
|
| 180 |
+
"learning_rate": 1.7640396413402548e-05,
|
| 181 |
+
"loss": 0.1227,
|
| 182 |
+
"step": 2500
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 1.18,
|
| 186 |
+
"eval_LOC": {
|
| 187 |
+
"f1": 0.7548371685643762,
|
| 188 |
+
"number": 6349,
|
| 189 |
+
"precision": 0.7632850241545893,
|
| 190 |
+
"recall": 0.746574263663569
|
| 191 |
+
},
|
| 192 |
+
"eval_MISC": {
|
| 193 |
+
"f1": 0.6036542515811666,
|
| 194 |
+
"number": 1602,
|
| 195 |
+
"precision": 0.6905144694533762,
|
| 196 |
+
"recall": 0.5362047440699126
|
| 197 |
+
},
|
| 198 |
+
"eval_NORP": {
|
| 199 |
+
"f1": 0.8488587291795189,
|
| 200 |
+
"number": 5690,
|
| 201 |
+
"precision": 0.8513346296623652,
|
| 202 |
+
"recall": 0.8463971880492092
|
| 203 |
+
},
|
| 204 |
+
"eval_PER": {
|
| 205 |
+
"f1": 0.895450291338345,
|
| 206 |
+
"number": 16183,
|
| 207 |
+
"precision": 0.8754427390791027,
|
| 208 |
+
"recall": 0.9163937465241303
|
| 209 |
+
},
|
| 210 |
+
"eval_loss": 0.1358954906463623,
|
| 211 |
+
"eval_overall_accuracy": 0.9613240559895834,
|
| 212 |
+
"eval_overall_f1": 0.8432567849686847,
|
| 213 |
+
"eval_overall_precision": 0.8400718778077269,
|
| 214 |
+
"eval_overall_recall": 0.8464659334763949,
|
| 215 |
+
"eval_runtime": 33.6007,
|
| 216 |
+
"eval_samples_per_second": 126.069,
|
| 217 |
+
"eval_steps_per_second": 15.773,
|
| 218 |
+
"step": 2500
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 1.42,
|
| 222 |
+
"learning_rate": 1.7168475696083062e-05,
|
| 223 |
+
"loss": 0.1043,
|
| 224 |
+
"step": 3000
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 1.42,
|
| 228 |
+
"eval_LOC": {
|
| 229 |
+
"f1": 0.7524268013574303,
|
| 230 |
+
"number": 6349,
|
| 231 |
+
"precision": 0.7540335336918697,
|
| 232 |
+
"recall": 0.7508269018743109
|
| 233 |
+
},
|
| 234 |
+
"eval_MISC": {
|
| 235 |
+
"f1": 0.6292059219380888,
|
| 236 |
+
"number": 1602,
|
| 237 |
+
"precision": 0.6824817518248175,
|
| 238 |
+
"recall": 0.583645443196005
|
| 239 |
+
},
|
| 240 |
+
"eval_NORP": {
|
| 241 |
+
"f1": 0.8393367638650656,
|
| 242 |
+
"number": 5690,
|
| 243 |
+
"precision": 0.7840683656340607,
|
| 244 |
+
"recall": 0.90298769771529
|
| 245 |
+
},
|
| 246 |
+
"eval_PER": {
|
| 247 |
+
"f1": 0.9035958798495035,
|
| 248 |
+
"number": 16183,
|
| 249 |
+
"precision": 0.9019269839315397,
|
| 250 |
+
"recall": 0.9052709633566088
|
| 251 |
+
},
|
| 252 |
+
"eval_loss": 0.1304474025964737,
|
| 253 |
+
"eval_overall_accuracy": 0.96197509765625,
|
| 254 |
+
"eval_overall_f1": 0.8452712561347658,
|
| 255 |
+
"eval_overall_precision": 0.8360666491734453,
|
| 256 |
+
"eval_overall_recall": 0.8546807939914163,
|
| 257 |
+
"eval_runtime": 34.7702,
|
| 258 |
+
"eval_samples_per_second": 121.828,
|
| 259 |
+
"eval_steps_per_second": 15.243,
|
| 260 |
+
"step": 3000
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 1.65,
|
| 264 |
+
"learning_rate": 1.669655497876357e-05,
|
| 265 |
+
"loss": 0.1078,
|
| 266 |
+
"step": 3500
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.65,
|
| 270 |
+
"eval_LOC": {
|
| 271 |
+
"f1": 0.7776814563795412,
|
| 272 |
+
"number": 6349,
|
| 273 |
+
"precision": 0.7782334384858044,
|
| 274 |
+
"recall": 0.7771302567333438
|
| 275 |
+
},
|
| 276 |
+
"eval_MISC": {
|
| 277 |
+
"f1": 0.6092849881396137,
|
| 278 |
+
"number": 1602,
|
| 279 |
+
"precision": 0.6664195700518903,
|
| 280 |
+
"recall": 0.5611735330836455
|
| 281 |
+
},
|
| 282 |
+
"eval_NORP": {
|
| 283 |
+
"f1": 0.8464839881148893,
|
| 284 |
+
"number": 5690,
|
| 285 |
+
"precision": 0.7980080921257392,
|
| 286 |
+
"recall": 0.9012302284710018
|
| 287 |
+
},
|
| 288 |
+
"eval_PER": {
|
| 289 |
+
"f1": 0.9055894496826277,
|
| 290 |
+
"number": 16183,
|
| 291 |
+
"precision": 0.9031405568188802,
|
| 292 |
+
"recall": 0.9080516591484892
|
| 293 |
+
},
|
| 294 |
+
"eval_loss": 0.12490628659725189,
|
| 295 |
+
"eval_overall_accuracy": 0.9640248616536459,
|
| 296 |
+
"eval_overall_f1": 0.8522172396611858,
|
| 297 |
+
"eval_overall_precision": 0.8443362074639637,
|
| 298 |
+
"eval_overall_recall": 0.8602467811158798,
|
| 299 |
+
"eval_runtime": 35.4399,
|
| 300 |
+
"eval_samples_per_second": 119.526,
|
| 301 |
+
"eval_steps_per_second": 14.955,
|
| 302 |
+
"step": 3500
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 1.89,
|
| 306 |
+
"learning_rate": 1.622463426144408e-05,
|
| 307 |
+
"loss": 0.1041,
|
| 308 |
+
"step": 4000
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 1.89,
|
| 312 |
+
"eval_LOC": {
|
| 313 |
+
"f1": 0.7832865606485812,
|
| 314 |
+
"number": 6349,
|
| 315 |
+
"precision": 0.7754283068374749,
|
| 316 |
+
"recall": 0.7913057174358167
|
| 317 |
+
},
|
| 318 |
+
"eval_MISC": {
|
| 319 |
+
"f1": 0.6451395896400942,
|
| 320 |
+
"number": 1602,
|
| 321 |
+
"precision": 0.6994894237782641,
|
| 322 |
+
"recall": 0.5986267166042447
|
| 323 |
+
},
|
| 324 |
+
"eval_NORP": {
|
| 325 |
+
"f1": 0.8637246248896735,
|
| 326 |
+
"number": 5690,
|
| 327 |
+
"precision": 0.8675531914893617,
|
| 328 |
+
"recall": 0.8599297012302285
|
| 329 |
+
},
|
| 330 |
+
"eval_PER": {
|
| 331 |
+
"f1": 0.9092253651321871,
|
| 332 |
+
"number": 16183,
|
| 333 |
+
"precision": 0.9067666400344171,
|
| 334 |
+
"recall": 0.9116974602978434
|
| 335 |
+
},
|
| 336 |
+
"eval_loss": 0.11825834214687347,
|
| 337 |
+
"eval_overall_accuracy": 0.9657440185546875,
|
| 338 |
+
"eval_overall_f1": 0.8602836284299741,
|
| 339 |
+
"eval_overall_precision": 0.8611941803030813,
|
| 340 |
+
"eval_overall_recall": 0.859375,
|
| 341 |
+
"eval_runtime": 35.0652,
|
| 342 |
+
"eval_samples_per_second": 120.804,
|
| 343 |
+
"eval_steps_per_second": 15.115,
|
| 344 |
+
"step": 4000
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 2.12,
|
| 348 |
+
"learning_rate": 1.575271354412459e-05,
|
| 349 |
+
"loss": 0.0799,
|
| 350 |
+
"step": 4500
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 2.12,
|
| 354 |
+
"eval_LOC": {
|
| 355 |
+
"f1": 0.7903951172502409,
|
| 356 |
+
"number": 6349,
|
| 357 |
+
"precision": 0.8063247583155825,
|
| 358 |
+
"recall": 0.7750826901874311
|
| 359 |
+
},
|
| 360 |
+
"eval_MISC": {
|
| 361 |
+
"f1": 0.6561734872332983,
|
| 362 |
+
"number": 1602,
|
| 363 |
+
"precision": 0.7462211614956245,
|
| 364 |
+
"recall": 0.5855181023720349
|
| 365 |
+
},
|
| 366 |
+
"eval_NORP": {
|
| 367 |
+
"f1": 0.8658915416703648,
|
| 368 |
+
"number": 5690,
|
| 369 |
+
"precision": 0.8746637977407208,
|
| 370 |
+
"recall": 0.8572934973637961
|
| 371 |
+
},
|
| 372 |
+
"eval_PER": {
|
| 373 |
+
"f1": 0.912650510125923,
|
| 374 |
+
"number": 16183,
|
| 375 |
+
"precision": 0.9050802139037433,
|
| 376 |
+
"recall": 0.9203485138725823
|
| 377 |
+
},
|
| 378 |
+
"eval_loss": 0.13416212797164917,
|
| 379 |
+
"eval_overall_accuracy": 0.9664764404296875,
|
| 380 |
+
"eval_overall_f1": 0.8656635763378759,
|
| 381 |
+
"eval_overall_precision": 0.8720103425985779,
|
| 382 |
+
"eval_overall_recall": 0.8594085300429185,
|
| 383 |
+
"eval_runtime": 34.7392,
|
| 384 |
+
"eval_samples_per_second": 121.937,
|
| 385 |
+
"eval_steps_per_second": 15.257,
|
| 386 |
+
"step": 4500
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 2.36,
|
| 390 |
+
"learning_rate": 1.52807928268051e-05,
|
| 391 |
+
"loss": 0.0679,
|
| 392 |
+
"step": 5000
|
| 393 |
+
},
|
| 394 |
+
{
|
| 395 |
+
"epoch": 2.36,
|
| 396 |
+
"eval_LOC": {
|
| 397 |
+
"f1": 0.7919011882426517,
|
| 398 |
+
"number": 6349,
|
| 399 |
+
"precision": 0.7861244761756946,
|
| 400 |
+
"recall": 0.7977634273113876
|
| 401 |
+
},
|
| 402 |
+
"eval_MISC": {
|
| 403 |
+
"f1": 0.6594274432379073,
|
| 404 |
+
"number": 1602,
|
| 405 |
+
"precision": 0.697286012526096,
|
| 406 |
+
"recall": 0.6254681647940075
|
| 407 |
+
},
|
| 408 |
+
"eval_NORP": {
|
| 409 |
+
"f1": 0.8735855728429986,
|
| 410 |
+
"number": 5690,
|
| 411 |
+
"precision": 0.8788687299893276,
|
| 412 |
+
"recall": 0.868365553602812
|
| 413 |
+
},
|
| 414 |
+
"eval_PER": {
|
| 415 |
+
"f1": 0.9156303340648623,
|
| 416 |
+
"number": 16183,
|
| 417 |
+
"precision": 0.9034586466165414,
|
| 418 |
+
"recall": 0.9281344620898474
|
| 419 |
+
},
|
| 420 |
+
"eval_loss": 0.1265156865119934,
|
| 421 |
+
"eval_overall_accuracy": 0.9679005940755209,
|
| 422 |
+
"eval_overall_f1": 0.868309119113943,
|
| 423 |
+
"eval_overall_precision": 0.8639426428120954,
|
| 424 |
+
"eval_overall_recall": 0.872719957081545,
|
| 425 |
+
"eval_runtime": 34.8379,
|
| 426 |
+
"eval_samples_per_second": 121.592,
|
| 427 |
+
"eval_steps_per_second": 15.213,
|
| 428 |
+
"step": 5000
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 2.6,
|
| 432 |
+
"learning_rate": 1.4808872109485608e-05,
|
| 433 |
+
"loss": 0.0652,
|
| 434 |
+
"step": 5500
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"epoch": 2.6,
|
| 438 |
+
"eval_LOC": {
|
| 439 |
+
"f1": 0.79075180386376,
|
| 440 |
+
"number": 6349,
|
| 441 |
+
"precision": 0.7792048929663609,
|
| 442 |
+
"recall": 0.802646085997795
|
| 443 |
+
},
|
| 444 |
+
"eval_MISC": {
|
| 445 |
+
"f1": 0.6546880104541,
|
| 446 |
+
"number": 1602,
|
| 447 |
+
"precision": 0.686771761480466,
|
| 448 |
+
"recall": 0.6254681647940075
|
| 449 |
+
},
|
| 450 |
+
"eval_NORP": {
|
| 451 |
+
"f1": 0.8732272570044967,
|
| 452 |
+
"number": 5690,
|
| 453 |
+
"precision": 0.8595505617977528,
|
| 454 |
+
"recall": 0.8873462214411247
|
| 455 |
+
},
|
| 456 |
+
"eval_PER": {
|
| 457 |
+
"f1": 0.9123680851726177,
|
| 458 |
+
"number": 16183,
|
| 459 |
+
"precision": 0.9193224592220828,
|
| 460 |
+
"recall": 0.9055181363158871
|
| 461 |
+
},
|
| 462 |
+
"eval_loss": 0.12422272562980652,
|
| 463 |
+
"eval_overall_accuracy": 0.9671122233072916,
|
| 464 |
+
"eval_overall_f1": 0.8652682059795094,
|
| 465 |
+
"eval_overall_precision": 0.865427833495455,
|
| 466 |
+
"eval_overall_recall": 0.8651086373390557,
|
| 467 |
+
"eval_runtime": 34.0635,
|
| 468 |
+
"eval_samples_per_second": 124.356,
|
| 469 |
+
"eval_steps_per_second": 15.559,
|
| 470 |
+
"step": 5500
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 2.83,
|
| 474 |
+
"learning_rate": 1.4336951392166117e-05,
|
| 475 |
+
"loss": 0.0695,
|
| 476 |
+
"step": 6000
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 2.83,
|
| 480 |
+
"eval_LOC": {
|
| 481 |
+
"f1": 0.7890688921616758,
|
| 482 |
+
"number": 6349,
|
| 483 |
+
"precision": 0.8210454622850332,
|
| 484 |
+
"recall": 0.759489683414711
|
| 485 |
+
},
|
| 486 |
+
"eval_MISC": {
|
| 487 |
+
"f1": 0.6770870337477797,
|
| 488 |
+
"number": 1602,
|
| 489 |
+
"precision": 0.7856553998351196,
|
| 490 |
+
"recall": 0.5948813982521848
|
| 491 |
+
},
|
| 492 |
+
"eval_NORP": {
|
| 493 |
+
"f1": 0.8647034425549563,
|
| 494 |
+
"number": 5690,
|
| 495 |
+
"precision": 0.81885310290652,
|
| 496 |
+
"recall": 0.9159929701230228
|
| 497 |
+
},
|
| 498 |
+
"eval_PER": {
|
| 499 |
+
"f1": 0.9161746617466175,
|
| 500 |
+
"number": 16183,
|
| 501 |
+
"precision": 0.9118565220052641,
|
| 502 |
+
"recall": 0.9205338935920411
|
| 503 |
+
},
|
| 504 |
+
"eval_loss": 0.1276456117630005,
|
| 505 |
+
"eval_overall_accuracy": 0.9684855143229166,
|
| 506 |
+
"eval_overall_f1": 0.8684157552170704,
|
| 507 |
+
"eval_overall_precision": 0.8689405129582383,
|
| 508 |
+
"eval_overall_recall": 0.8678916309012875,
|
| 509 |
+
"eval_runtime": 35.0576,
|
| 510 |
+
"eval_samples_per_second": 120.83,
|
| 511 |
+
"eval_steps_per_second": 15.118,
|
| 512 |
+
"step": 6000
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 3.07,
|
| 516 |
+
"learning_rate": 1.3865030674846627e-05,
|
| 517 |
+
"loss": 0.0573,
|
| 518 |
+
"step": 6500
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"epoch": 3.07,
|
| 522 |
+
"eval_LOC": {
|
| 523 |
+
"f1": 0.7923842629167325,
|
| 524 |
+
"number": 6349,
|
| 525 |
+
"precision": 0.7948961800602314,
|
| 526 |
+
"recall": 0.7898881713655694
|
| 527 |
+
},
|
| 528 |
+
"eval_MISC": {
|
| 529 |
+
"f1": 0.6901140684410647,
|
| 530 |
+
"number": 1602,
|
| 531 |
+
"precision": 0.7007722007722008,
|
| 532 |
+
"recall": 0.6797752808988764
|
| 533 |
+
},
|
| 534 |
+
"eval_NORP": {
|
| 535 |
+
"f1": 0.8771058698350772,
|
| 536 |
+
"number": 5690,
|
| 537 |
+
"precision": 0.8851109520400859,
|
| 538 |
+
"recall": 0.869244288224956
|
| 539 |
+
},
|
| 540 |
+
"eval_PER": {
|
| 541 |
+
"f1": 0.9171229698375871,
|
| 542 |
+
"number": 16183,
|
| 543 |
+
"precision": 0.9182876966918597,
|
| 544 |
+
"recall": 0.9159611938453933
|
| 545 |
+
},
|
| 546 |
+
"eval_loss": 0.1303853839635849,
|
| 547 |
+
"eval_overall_accuracy": 0.967529296875,
|
| 548 |
+
"eval_overall_f1": 0.8708955349479105,
|
| 549 |
+
"eval_overall_precision": 0.8742945966951644,
|
| 550 |
+
"eval_overall_recall": 0.8675228004291845,
|
| 551 |
+
"eval_runtime": 36.8238,
|
| 552 |
+
"eval_samples_per_second": 115.034,
|
| 553 |
+
"eval_steps_per_second": 14.393,
|
| 554 |
+
"step": 6500
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 3.3,
|
| 558 |
+
"learning_rate": 1.3393109957527137e-05,
|
| 559 |
+
"loss": 0.0444,
|
| 560 |
+
"step": 7000
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 3.3,
|
| 564 |
+
"eval_LOC": {
|
| 565 |
+
"f1": 0.7984538074990337,
|
| 566 |
+
"number": 6349,
|
| 567 |
+
"precision": 0.7840874582447617,
|
| 568 |
+
"recall": 0.8133564340841077
|
| 569 |
+
},
|
| 570 |
+
"eval_MISC": {
|
| 571 |
+
"f1": 0.6787416587225928,
|
| 572 |
+
"number": 1602,
|
| 573 |
+
"precision": 0.6912621359223301,
|
| 574 |
+
"recall": 0.6666666666666666
|
| 575 |
+
},
|
| 576 |
+
"eval_NORP": {
|
| 577 |
+
"f1": 0.8763066202090593,
|
| 578 |
+
"number": 5690,
|
| 579 |
+
"precision": 0.8687392055267703,
|
| 580 |
+
"recall": 0.8840070298769771
|
| 581 |
+
},
|
| 582 |
+
"eval_PER": {
|
| 583 |
+
"f1": 0.915957745162195,
|
| 584 |
+
"number": 16183,
|
| 585 |
+
"precision": 0.9238747799849133,
|
| 586 |
+
"recall": 0.9081752456281282
|
| 587 |
+
},
|
| 588 |
+
"eval_loss": 0.13717159628868103,
|
| 589 |
+
"eval_overall_accuracy": 0.96856689453125,
|
| 590 |
+
"eval_overall_f1": 0.8703334283271588,
|
| 591 |
+
"eval_overall_precision": 0.87026048476315,
|
| 592 |
+
"eval_overall_recall": 0.8704063841201717,
|
| 593 |
+
"eval_runtime": 33.9103,
|
| 594 |
+
"eval_samples_per_second": 124.918,
|
| 595 |
+
"eval_steps_per_second": 15.629,
|
| 596 |
+
"step": 7000
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 3.54,
|
| 600 |
+
"learning_rate": 1.2921189240207645e-05,
|
| 601 |
+
"loss": 0.0428,
|
| 602 |
+
"step": 7500
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"epoch": 3.54,
|
| 606 |
+
"eval_LOC": {
|
| 607 |
+
"f1": 0.8001958064779311,
|
| 608 |
+
"number": 6349,
|
| 609 |
+
"precision": 0.8300609343263372,
|
| 610 |
+
"recall": 0.7724051031658529
|
| 611 |
+
},
|
| 612 |
+
"eval_MISC": {
|
| 613 |
+
"f1": 0.6906565656565657,
|
| 614 |
+
"number": 1602,
|
| 615 |
+
"precision": 0.698595146871009,
|
| 616 |
+
"recall": 0.682896379525593
|
| 617 |
+
},
|
| 618 |
+
"eval_NORP": {
|
| 619 |
+
"f1": 0.8765986279325824,
|
| 620 |
+
"number": 5690,
|
| 621 |
+
"precision": 0.8460029426189308,
|
| 622 |
+
"recall": 0.9094903339191565
|
| 623 |
+
},
|
| 624 |
+
"eval_PER": {
|
| 625 |
+
"f1": 0.9160423025930381,
|
| 626 |
+
"number": 16183,
|
| 627 |
+
"precision": 0.9141538461538462,
|
| 628 |
+
"recall": 0.9179385775196194
|
| 629 |
+
},
|
| 630 |
+
"eval_loss": 0.13739602267742157,
|
| 631 |
+
"eval_overall_accuracy": 0.9689229329427084,
|
| 632 |
+
"eval_overall_f1": 0.8724712980809519,
|
| 633 |
+
"eval_overall_precision": 0.8722227807379109,
|
| 634 |
+
"eval_overall_recall": 0.872719957081545,
|
| 635 |
+
"eval_runtime": 34.6373,
|
| 636 |
+
"eval_samples_per_second": 122.296,
|
| 637 |
+
"eval_steps_per_second": 15.301,
|
| 638 |
+
"step": 7500
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 3.78,
|
| 642 |
+
"learning_rate": 1.2449268522888157e-05,
|
| 643 |
+
"loss": 0.0406,
|
| 644 |
+
"step": 8000
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 3.78,
|
| 648 |
+
"eval_LOC": {
|
| 649 |
+
"f1": 0.8012557353296305,
|
| 650 |
+
"number": 6349,
|
| 651 |
+
"precision": 0.819394138952914,
|
| 652 |
+
"recall": 0.7839029768467475
|
| 653 |
+
},
|
| 654 |
+
"eval_MISC": {
|
| 655 |
+
"f1": 0.7026332691072574,
|
| 656 |
+
"number": 1602,
|
| 657 |
+
"precision": 0.7235449735449735,
|
| 658 |
+
"recall": 0.682896379525593
|
| 659 |
+
},
|
| 660 |
+
"eval_NORP": {
|
| 661 |
+
"f1": 0.8827633732448963,
|
| 662 |
+
"number": 5690,
|
| 663 |
+
"precision": 0.8656867714140902,
|
| 664 |
+
"recall": 0.9005272407732865
|
| 665 |
+
},
|
| 666 |
+
"eval_PER": {
|
| 667 |
+
"f1": 0.9211724791392853,
|
| 668 |
+
"number": 16183,
|
| 669 |
+
"precision": 0.9113947018265393,
|
| 670 |
+
"recall": 0.931162330841006
|
| 671 |
+
},
|
| 672 |
+
"eval_loss": 0.13875727355480194,
|
| 673 |
+
"eval_overall_accuracy": 0.9702097574869791,
|
| 674 |
+
"eval_overall_f1": 0.8774702236773967,
|
| 675 |
+
"eval_overall_precision": 0.8743300376177636,
|
| 676 |
+
"eval_overall_recall": 0.8806330472103004,
|
| 677 |
+
"eval_runtime": 34.5296,
|
| 678 |
+
"eval_samples_per_second": 122.677,
|
| 679 |
+
"eval_steps_per_second": 15.349,
|
| 680 |
+
"step": 8000
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 4.01,
|
| 684 |
+
"learning_rate": 1.1977347805568667e-05,
|
| 685 |
+
"loss": 0.0449,
|
| 686 |
+
"step": 8500
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 4.01,
|
| 690 |
+
"eval_LOC": {
|
| 691 |
+
"f1": 0.8041580701475957,
|
| 692 |
+
"number": 6349,
|
| 693 |
+
"precision": 0.8103310411002719,
|
| 694 |
+
"recall": 0.7980784375492204
|
| 695 |
+
},
|
| 696 |
+
"eval_MISC": {
|
| 697 |
+
"f1": 0.7114228456913828,
|
| 698 |
+
"number": 1602,
|
| 699 |
+
"precision": 0.7650862068965517,
|
| 700 |
+
"recall": 0.6647940074906367
|
| 701 |
+
},
|
| 702 |
+
"eval_NORP": {
|
| 703 |
+
"f1": 0.8826941390355814,
|
| 704 |
+
"number": 5690,
|
| 705 |
+
"precision": 0.8698174372973895,
|
| 706 |
+
"recall": 0.8959578207381371
|
| 707 |
+
},
|
| 708 |
+
"eval_PER": {
|
| 709 |
+
"f1": 0.9232705174636303,
|
| 710 |
+
"number": 16183,
|
| 711 |
+
"precision": 0.9171900725654003,
|
| 712 |
+
"recall": 0.9294321201260582
|
| 713 |
+
},
|
| 714 |
+
"eval_loss": 0.13972027599811554,
|
| 715 |
+
"eval_overall_accuracy": 0.9705251057942709,
|
| 716 |
+
"eval_overall_f1": 0.8796731905774414,
|
| 717 |
+
"eval_overall_precision": 0.8784818592208661,
|
| 718 |
+
"eval_overall_recall": 0.8808677575107297,
|
| 719 |
+
"eval_runtime": 35.3655,
|
| 720 |
+
"eval_samples_per_second": 119.778,
|
| 721 |
+
"eval_steps_per_second": 14.986,
|
| 722 |
+
"step": 8500
|
| 723 |
+
},
|
| 724 |
+
{
|
| 725 |
+
"epoch": 4.25,
|
| 726 |
+
"learning_rate": 1.1505427088249175e-05,
|
| 727 |
+
"loss": 0.0269,
|
| 728 |
+
"step": 9000
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 4.25,
|
| 732 |
+
"eval_LOC": {
|
| 733 |
+
"f1": 0.8022795630837423,
|
| 734 |
+
"number": 6349,
|
| 735 |
+
"precision": 0.8063643595863166,
|
| 736 |
+
"recall": 0.7982359426681367
|
| 737 |
+
},
|
| 738 |
+
"eval_MISC": {
|
| 739 |
+
"f1": 0.716551724137931,
|
| 740 |
+
"number": 1602,
|
| 741 |
+
"precision": 0.800462249614792,
|
| 742 |
+
"recall": 0.6485642946317104
|
| 743 |
+
},
|
| 744 |
+
"eval_NORP": {
|
| 745 |
+
"f1": 0.879294097489934,
|
| 746 |
+
"number": 5690,
|
| 747 |
+
"precision": 0.8577636637138559,
|
| 748 |
+
"recall": 0.901933216168717
|
| 749 |
+
},
|
| 750 |
+
"eval_PER": {
|
| 751 |
+
"f1": 0.9192701977056382,
|
| 752 |
+
"number": 16183,
|
| 753 |
+
"precision": 0.9079129753510516,
|
| 754 |
+
"recall": 0.9309151578817277
|
| 755 |
+
},
|
| 756 |
+
"eval_loss": 0.1521347165107727,
|
| 757 |
+
"eval_overall_accuracy": 0.9700419108072916,
|
| 758 |
+
"eval_overall_f1": 0.8770484970741711,
|
| 759 |
+
"eval_overall_precision": 0.8721774594648364,
|
| 760 |
+
"eval_overall_recall": 0.8819742489270386,
|
| 761 |
+
"eval_runtime": 35.7932,
|
| 762 |
+
"eval_samples_per_second": 118.347,
|
| 763 |
+
"eval_steps_per_second": 14.807,
|
| 764 |
+
"step": 9000
|
| 765 |
+
},
|
| 766 |
+
{
|
| 767 |
+
"epoch": 4.48,
|
| 768 |
+
"learning_rate": 1.1033506370929685e-05,
|
| 769 |
+
"loss": 0.0261,
|
| 770 |
+
"step": 9500
|
| 771 |
+
},
|
| 772 |
+
{
|
| 773 |
+
"epoch": 4.48,
|
| 774 |
+
"eval_LOC": {
|
| 775 |
+
"f1": 0.8101327290385238,
|
| 776 |
+
"number": 6349,
|
| 777 |
+
"precision": 0.8331946062926585,
|
| 778 |
+
"recall": 0.7883131201764058
|
| 779 |
+
},
|
| 780 |
+
"eval_MISC": {
|
| 781 |
+
"f1": 0.707196029776675,
|
| 782 |
+
"number": 1602,
|
| 783 |
+
"precision": 0.7028360049321825,
|
| 784 |
+
"recall": 0.7116104868913857
|
| 785 |
+
},
|
| 786 |
+
"eval_NORP": {
|
| 787 |
+
"f1": 0.8803396586084394,
|
| 788 |
+
"number": 5690,
|
| 789 |
+
"precision": 0.8682276533925825,
|
| 790 |
+
"recall": 0.8927943760984183
|
| 791 |
+
},
|
| 792 |
+
"eval_PER": {
|
| 793 |
+
"f1": 0.9222290263319044,
|
| 794 |
+
"number": 16183,
|
| 795 |
+
"precision": 0.9140013351945135,
|
| 796 |
+
"recall": 0.9306061916826299
|
| 797 |
+
},
|
| 798 |
+
"eval_loss": 0.15231968462467194,
|
| 799 |
+
"eval_overall_accuracy": 0.9703928629557291,
|
| 800 |
+
"eval_overall_f1": 0.8793763904919624,
|
| 801 |
+
"eval_overall_precision": 0.8774243081750509,
|
| 802 |
+
"eval_overall_recall": 0.881337178111588,
|
| 803 |
+
"eval_runtime": 35.4408,
|
| 804 |
+
"eval_samples_per_second": 119.523,
|
| 805 |
+
"eval_steps_per_second": 14.955,
|
| 806 |
+
"step": 9500
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 4.72,
|
| 810 |
+
"learning_rate": 1.0561585653610193e-05,
|
| 811 |
+
"loss": 0.0285,
|
| 812 |
+
"step": 10000
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"epoch": 4.72,
|
| 816 |
+
"eval_LOC": {
|
| 817 |
+
"f1": 0.8007642338555598,
|
| 818 |
+
"number": 6349,
|
| 819 |
+
"precision": 0.7777612826603325,
|
| 820 |
+
"recall": 0.8251693180028351
|
| 821 |
+
},
|
| 822 |
+
"eval_MISC": {
|
| 823 |
+
"f1": 0.7110076488194212,
|
| 824 |
+
"number": 1602,
|
| 825 |
+
"precision": 0.7608540925266903,
|
| 826 |
+
"recall": 0.66729088639201
|
| 827 |
+
},
|
| 828 |
+
"eval_NORP": {
|
| 829 |
+
"f1": 0.8860268172194778,
|
| 830 |
+
"number": 5690,
|
| 831 |
+
"precision": 0.8894792773645058,
|
| 832 |
+
"recall": 0.8826010544815466
|
| 833 |
+
},
|
| 834 |
+
"eval_PER": {
|
| 835 |
+
"f1": 0.9219896875947831,
|
| 836 |
+
"number": 16183,
|
| 837 |
+
"precision": 0.9054029904092452,
|
| 838 |
+
"recall": 0.9391954520175493
|
| 839 |
+
},
|
| 840 |
+
"eval_loss": 0.14308173954486847,
|
| 841 |
+
"eval_overall_accuracy": 0.9701334635416666,
|
| 842 |
+
"eval_overall_f1": 0.8784727971124872,
|
| 843 |
+
"eval_overall_precision": 0.8676980440897495,
|
| 844 |
+
"eval_overall_recall": 0.889518508583691,
|
| 845 |
+
"eval_runtime": 33.8885,
|
| 846 |
+
"eval_samples_per_second": 124.998,
|
| 847 |
+
"eval_steps_per_second": 15.64,
|
| 848 |
+
"step": 10000
|
| 849 |
+
},
|
| 850 |
+
{
|
| 851 |
+
"epoch": 4.96,
|
| 852 |
+
"learning_rate": 1.0089664936290703e-05,
|
| 853 |
+
"loss": 0.0289,
|
| 854 |
+
"step": 10500
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 4.96,
|
| 858 |
+
"eval_LOC": {
|
| 859 |
+
"f1": 0.8048566179407302,
|
| 860 |
+
"number": 6349,
|
| 861 |
+
"precision": 0.8165316045380875,
|
| 862 |
+
"recall": 0.7935107891006458
|
| 863 |
+
},
|
| 864 |
+
"eval_MISC": {
|
| 865 |
+
"f1": 0.718897108271688,
|
| 866 |
+
"number": 1602,
|
| 867 |
+
"precision": 0.7791545189504373,
|
| 868 |
+
"recall": 0.66729088639201
|
| 869 |
+
},
|
| 870 |
+
"eval_NORP": {
|
| 871 |
+
"f1": 0.8847718812224049,
|
| 872 |
+
"number": 5690,
|
| 873 |
+
"precision": 0.8718648694761986,
|
| 874 |
+
"recall": 0.8980667838312829
|
| 875 |
+
},
|
| 876 |
+
"eval_PER": {
|
| 877 |
+
"f1": 0.9236515297706123,
|
| 878 |
+
"number": 16183,
|
| 879 |
+
"precision": 0.9305154897780008,
|
| 880 |
+
"recall": 0.9168880924426868
|
| 881 |
+
},
|
| 882 |
+
"eval_loss": 0.1427399069070816,
|
| 883 |
+
"eval_overall_accuracy": 0.970794677734375,
|
| 884 |
+
"eval_overall_f1": 0.8806381288763456,
|
| 885 |
+
"eval_overall_precision": 0.8877644894204232,
|
| 886 |
+
"eval_overall_recall": 0.8736252682403434,
|
| 887 |
+
"eval_runtime": 35.6412,
|
| 888 |
+
"eval_samples_per_second": 118.851,
|
| 889 |
+
"eval_steps_per_second": 14.87,
|
| 890 |
+
"step": 10500
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 5.19,
|
| 894 |
+
"learning_rate": 9.617744218971213e-06,
|
| 895 |
+
"loss": 0.0238,
|
| 896 |
+
"step": 11000
|
| 897 |
+
},
|
| 898 |
+
{
|
| 899 |
+
"epoch": 5.19,
|
| 900 |
+
"eval_LOC": {
|
| 901 |
+
"f1": 0.8100460975997457,
|
| 902 |
+
"number": 6349,
|
| 903 |
+
"precision": 0.8175838280121932,
|
| 904 |
+
"recall": 0.802646085997795
|
| 905 |
+
},
|
| 906 |
+
"eval_MISC": {
|
| 907 |
+
"f1": 0.7157343801112201,
|
| 908 |
+
"number": 1602,
|
| 909 |
+
"precision": 0.7518900343642612,
|
| 910 |
+
"recall": 0.682896379525593
|
| 911 |
+
},
|
| 912 |
+
"eval_NORP": {
|
| 913 |
+
"f1": 0.8887525203822214,
|
| 914 |
+
"number": 5690,
|
| 915 |
+
"precision": 0.8866538394262725,
|
| 916 |
+
"recall": 0.8908611599297013
|
| 917 |
+
},
|
| 918 |
+
"eval_PER": {
|
| 919 |
+
"f1": 0.9244720420521962,
|
| 920 |
+
"number": 16183,
|
| 921 |
+
"precision": 0.919806704183998,
|
| 922 |
+
"recall": 0.9291849471667799
|
| 923 |
+
},
|
| 924 |
+
"eval_loss": 0.15089917182922363,
|
| 925 |
+
"eval_overall_accuracy": 0.9715423583984375,
|
| 926 |
+
"eval_overall_f1": 0.8827567685516223,
|
| 927 |
+
"eval_overall_precision": 0.8838100359627601,
|
| 928 |
+
"eval_overall_recall": 0.881706008583691,
|
| 929 |
+
"eval_runtime": 34.4219,
|
| 930 |
+
"eval_samples_per_second": 123.061,
|
| 931 |
+
"eval_steps_per_second": 15.397,
|
| 932 |
+
"step": 11000
|
| 933 |
+
},
|
| 934 |
+
{
|
| 935 |
+
"epoch": 5.43,
|
| 936 |
+
"learning_rate": 9.145823501651723e-06,
|
| 937 |
+
"loss": 0.0195,
|
| 938 |
+
"step": 11500
|
| 939 |
+
},
|
| 940 |
+
{
|
| 941 |
+
"epoch": 5.43,
|
| 942 |
+
"eval_LOC": {
|
| 943 |
+
"f1": 0.8136754861332485,
|
| 944 |
+
"number": 6349,
|
| 945 |
+
"precision": 0.8235199225681562,
|
| 946 |
+
"recall": 0.8040636320680422
|
| 947 |
+
},
|
| 948 |
+
"eval_MISC": {
|
| 949 |
+
"f1": 0.7222403126017584,
|
| 950 |
+
"number": 1602,
|
| 951 |
+
"precision": 0.7549353301565691,
|
| 952 |
+
"recall": 0.6922596754057428
|
| 953 |
+
},
|
| 954 |
+
"eval_NORP": {
|
| 955 |
+
"f1": 0.888985557682269,
|
| 956 |
+
"number": 5690,
|
| 957 |
+
"precision": 0.8802549965541007,
|
| 958 |
+
"recall": 0.8978910369068541
|
| 959 |
+
},
|
| 960 |
+
"eval_PER": {
|
| 961 |
+
"f1": 0.9241790677653425,
|
| 962 |
+
"number": 16183,
|
| 963 |
+
"precision": 0.9197111559880057,
|
| 964 |
+
"recall": 0.9286906012482234
|
| 965 |
+
},
|
| 966 |
+
"eval_loss": 0.16119226813316345,
|
| 967 |
+
"eval_overall_accuracy": 0.9717254638671875,
|
| 968 |
+
"eval_overall_f1": 0.8837466673373913,
|
| 969 |
+
"eval_overall_precision": 0.883909703820481,
|
| 970 |
+
"eval_overall_recall": 0.8835836909871244,
|
| 971 |
+
"eval_runtime": 34.047,
|
| 972 |
+
"eval_samples_per_second": 124.416,
|
| 973 |
+
"eval_steps_per_second": 15.567,
|
| 974 |
+
"step": 11500
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 5.66,
|
| 978 |
+
"learning_rate": 8.673902784332233e-06,
|
| 979 |
+
"loss": 0.0218,
|
| 980 |
+
"step": 12000
|
| 981 |
+
},
|
| 982 |
+
{
|
| 983 |
+
"epoch": 5.66,
|
| 984 |
+
"eval_LOC": {
|
| 985 |
+
"f1": 0.7971036820212603,
|
| 986 |
+
"number": 6349,
|
| 987 |
+
"precision": 0.7800391979496457,
|
| 988 |
+
"recall": 0.8149314852732714
|
| 989 |
+
},
|
| 990 |
+
"eval_MISC": {
|
| 991 |
+
"f1": 0.7200520833333334,
|
| 992 |
+
"number": 1602,
|
| 993 |
+
"precision": 0.7523809523809524,
|
| 994 |
+
"recall": 0.6903870162297129
|
| 995 |
+
},
|
| 996 |
+
"eval_NORP": {
|
| 997 |
+
"f1": 0.8812952705581594,
|
| 998 |
+
"number": 5690,
|
| 999 |
+
"precision": 0.8554177005789909,
|
| 1000 |
+
"recall": 0.9087873462214411
|
| 1001 |
+
},
|
| 1002 |
+
"eval_PER": {
|
| 1003 |
+
"f1": 0.9216090609707305,
|
| 1004 |
+
"number": 16183,
|
| 1005 |
+
"precision": 0.9180770174147658,
|
| 1006 |
+
"recall": 0.9251683865785083
|
| 1007 |
+
},
|
| 1008 |
+
"eval_loss": 0.16442249715328217,
|
| 1009 |
+
"eval_overall_accuracy": 0.9700113932291666,
|
| 1010 |
+
"eval_overall_f1": 0.8766755142667552,
|
| 1011 |
+
"eval_overall_precision": 0.8675794588915157,
|
| 1012 |
+
"eval_overall_recall": 0.8859643240343348,
|
| 1013 |
+
"eval_runtime": 35.3444,
|
| 1014 |
+
"eval_samples_per_second": 119.849,
|
| 1015 |
+
"eval_steps_per_second": 14.995,
|
| 1016 |
+
"step": 12000
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 5.9,
|
| 1020 |
+
"learning_rate": 8.201982067012742e-06,
|
| 1021 |
+
"loss": 0.0183,
|
| 1022 |
+
"step": 12500
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 5.9,
|
| 1026 |
+
"eval_LOC": {
|
| 1027 |
+
"f1": 0.8106127001832523,
|
| 1028 |
+
"number": 6349,
|
| 1029 |
+
"precision": 0.8202192841019026,
|
| 1030 |
+
"recall": 0.8012285399275476
|
| 1031 |
+
},
|
| 1032 |
+
"eval_MISC": {
|
| 1033 |
+
"f1": 0.7068697544295928,
|
| 1034 |
+
"number": 1602,
|
| 1035 |
+
"precision": 0.7040247678018576,
|
| 1036 |
+
"recall": 0.7097378277153558
|
| 1037 |
+
},
|
| 1038 |
+
"eval_NORP": {
|
| 1039 |
+
"f1": 0.8847426002883555,
|
| 1040 |
+
"number": 5690,
|
| 1041 |
+
"precision": 0.8549418128175709,
|
| 1042 |
+
"recall": 0.9166959578207381
|
| 1043 |
+
},
|
| 1044 |
+
"eval_PER": {
|
| 1045 |
+
"f1": 0.9248411591507826,
|
| 1046 |
+
"number": 16183,
|
| 1047 |
+
"precision": 0.9277453053102848,
|
| 1048 |
+
"recall": 0.921955138107891
|
| 1049 |
+
},
|
| 1050 |
+
"eval_loss": 0.16248470544815063,
|
| 1051 |
+
"eval_overall_accuracy": 0.971343994140625,
|
| 1052 |
+
"eval_overall_f1": 0.8812516715699384,
|
| 1053 |
+
"eval_overall_precision": 0.8786666666666667,
|
| 1054 |
+
"eval_overall_recall": 0.8838519313304721,
|
| 1055 |
+
"eval_runtime": 34.2026,
|
| 1056 |
+
"eval_samples_per_second": 123.85,
|
| 1057 |
+
"eval_steps_per_second": 15.496,
|
| 1058 |
+
"step": 12500
|
| 1059 |
+
},
|
| 1060 |
+
{
|
| 1061 |
+
"epoch": 6.13,
|
| 1062 |
+
"learning_rate": 7.730061349693252e-06,
|
| 1063 |
+
"loss": 0.016,
|
| 1064 |
+
"step": 13000
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 6.13,
|
| 1068 |
+
"eval_LOC": {
|
| 1069 |
+
"f1": 0.8088959491660047,
|
| 1070 |
+
"number": 6349,
|
| 1071 |
+
"precision": 0.8158948886396411,
|
| 1072 |
+
"recall": 0.8020160655221295
|
| 1073 |
+
},
|
| 1074 |
+
"eval_MISC": {
|
| 1075 |
+
"f1": 0.7173844702923609,
|
| 1076 |
+
"number": 1602,
|
| 1077 |
+
"precision": 0.7226092463584547,
|
| 1078 |
+
"recall": 0.712234706616729
|
| 1079 |
+
},
|
| 1080 |
+
"eval_NORP": {
|
| 1081 |
+
"f1": 0.8844587716321684,
|
| 1082 |
+
"number": 5690,
|
| 1083 |
+
"precision": 0.8548704493276484,
|
| 1084 |
+
"recall": 0.9161687170474516
|
| 1085 |
+
},
|
| 1086 |
+
"eval_PER": {
|
| 1087 |
+
"f1": 0.9247578842811025,
|
| 1088 |
+
"number": 16183,
|
| 1089 |
+
"precision": 0.9290837647352336,
|
| 1090 |
+
"recall": 0.9204721003522215
|
| 1091 |
+
},
|
| 1092 |
+
"eval_loss": 0.17019298672676086,
|
| 1093 |
+
"eval_overall_accuracy": 0.9712270100911459,
|
| 1094 |
+
"eval_overall_f1": 0.8813718109577583,
|
| 1095 |
+
"eval_overall_precision": 0.8795031885412841,
|
| 1096 |
+
"eval_overall_recall": 0.8832483905579399,
|
| 1097 |
+
"eval_runtime": 33.7485,
|
| 1098 |
+
"eval_samples_per_second": 125.517,
|
| 1099 |
+
"eval_steps_per_second": 15.704,
|
| 1100 |
+
"step": 13000
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 6.37,
|
| 1104 |
+
"learning_rate": 7.258140632373761e-06,
|
| 1105 |
+
"loss": 0.0135,
|
| 1106 |
+
"step": 13500
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 6.37,
|
| 1110 |
+
"eval_LOC": {
|
| 1111 |
+
"f1": 0.8114028274040702,
|
| 1112 |
+
"number": 6349,
|
| 1113 |
+
"precision": 0.8004597701149425,
|
| 1114 |
+
"recall": 0.8226492361001733
|
| 1115 |
+
},
|
| 1116 |
+
"eval_MISC": {
|
| 1117 |
+
"f1": 0.7318982387475539,
|
| 1118 |
+
"number": 1602,
|
| 1119 |
+
"precision": 0.7663934426229508,
|
| 1120 |
+
"recall": 0.700374531835206
|
| 1121 |
+
},
|
| 1122 |
+
"eval_NORP": {
|
| 1123 |
+
"f1": 0.883330448329583,
|
| 1124 |
+
"number": 5690,
|
| 1125 |
+
"precision": 0.870225102319236,
|
| 1126 |
+
"recall": 0.8968365553602812
|
| 1127 |
+
},
|
| 1128 |
+
"eval_PER": {
|
| 1129 |
+
"f1": 0.9247437955303124,
|
| 1130 |
+
"number": 16183,
|
| 1131 |
+
"precision": 0.923888237833837,
|
| 1132 |
+
"recall": 0.9256009392572453
|
| 1133 |
+
},
|
| 1134 |
+
"eval_loss": 0.16678640246391296,
|
| 1135 |
+
"eval_overall_accuracy": 0.9714457194010416,
|
| 1136 |
+
"eval_overall_f1": 0.8825179495742194,
|
| 1137 |
+
"eval_overall_precision": 0.8789662741967671,
|
| 1138 |
+
"eval_overall_recall": 0.8860984442060086,
|
| 1139 |
+
"eval_runtime": 33.9369,
|
| 1140 |
+
"eval_samples_per_second": 124.82,
|
| 1141 |
+
"eval_steps_per_second": 15.617,
|
| 1142 |
+
"step": 13500
|
| 1143 |
+
},
|
| 1144 |
+
{
|
| 1145 |
+
"epoch": 6.61,
|
| 1146 |
+
"learning_rate": 6.786219915054272e-06,
|
| 1147 |
+
"loss": 0.0127,
|
| 1148 |
+
"step": 14000
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 6.61,
|
| 1152 |
+
"eval_LOC": {
|
| 1153 |
+
"f1": 0.8155262740117478,
|
| 1154 |
+
"number": 6349,
|
| 1155 |
+
"precision": 0.8220515282445191,
|
| 1156 |
+
"recall": 0.8091037958733659
|
| 1157 |
+
},
|
| 1158 |
+
"eval_MISC": {
|
| 1159 |
+
"f1": 0.7328699106256207,
|
| 1160 |
+
"number": 1602,
|
| 1161 |
+
"precision": 0.7801268498942917,
|
| 1162 |
+
"recall": 0.6910112359550562
|
| 1163 |
+
},
|
| 1164 |
+
"eval_NORP": {
|
| 1165 |
+
"f1": 0.885006518904824,
|
| 1166 |
+
"number": 5690,
|
| 1167 |
+
"precision": 0.8754944110060189,
|
| 1168 |
+
"recall": 0.8947275922671353
|
| 1169 |
+
},
|
| 1170 |
+
"eval_PER": {
|
| 1171 |
+
"f1": 0.9258804761320408,
|
| 1172 |
+
"number": 16183,
|
| 1173 |
+
"precision": 0.9193931639554012,
|
| 1174 |
+
"recall": 0.9324599888772168
|
| 1175 |
+
},
|
| 1176 |
+
"eval_loss": 0.1748281568288803,
|
| 1177 |
+
"eval_overall_accuracy": 0.9720052083333334,
|
| 1178 |
+
"eval_overall_f1": 0.8849631614199598,
|
| 1179 |
+
"eval_overall_precision": 0.8838975113727588,
|
| 1180 |
+
"eval_overall_recall": 0.8860313841201717,
|
| 1181 |
+
"eval_runtime": 34.1866,
|
| 1182 |
+
"eval_samples_per_second": 123.908,
|
| 1183 |
+
"eval_steps_per_second": 15.503,
|
| 1184 |
+
"step": 14000
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 6.84,
|
| 1188 |
+
"learning_rate": 6.314299197734781e-06,
|
| 1189 |
+
"loss": 0.014,
|
| 1190 |
+
"step": 14500
|
| 1191 |
+
},
|
| 1192 |
+
{
|
| 1193 |
+
"epoch": 6.84,
|
| 1194 |
+
"eval_LOC": {
|
| 1195 |
+
"f1": 0.8157271435338702,
|
| 1196 |
+
"number": 6349,
|
| 1197 |
+
"precision": 0.817793256292544,
|
| 1198 |
+
"recall": 0.8136714443219405
|
| 1199 |
+
},
|
| 1200 |
+
"eval_MISC": {
|
| 1201 |
+
"f1": 0.7454485269778219,
|
| 1202 |
+
"number": 1602,
|
| 1203 |
+
"precision": 0.7935165609584214,
|
| 1204 |
+
"recall": 0.7028714107365793
|
| 1205 |
+
},
|
| 1206 |
+
"eval_NORP": {
|
| 1207 |
+
"f1": 0.8912159238424924,
|
| 1208 |
+
"number": 5690,
|
| 1209 |
+
"precision": 0.877919863597613,
|
| 1210 |
+
"recall": 0.9049209138840071
|
| 1211 |
+
},
|
| 1212 |
+
"eval_PER": {
|
| 1213 |
+
"f1": 0.9255973540230596,
|
| 1214 |
+
"number": 16183,
|
| 1215 |
+
"precision": 0.9260267194458189,
|
| 1216 |
+
"recall": 0.9251683865785083
|
| 1217 |
+
},
|
| 1218 |
+
"eval_loss": 0.17196723818778992,
|
| 1219 |
+
"eval_overall_accuracy": 0.9721476236979166,
|
| 1220 |
+
"eval_overall_f1": 0.8864463947107882,
|
| 1221 |
+
"eval_overall_precision": 0.8872652759582116,
|
| 1222 |
+
"eval_overall_recall": 0.8856290236051502,
|
| 1223 |
+
"eval_runtime": 35.3824,
|
| 1224 |
+
"eval_samples_per_second": 119.721,
|
| 1225 |
+
"eval_steps_per_second": 14.979,
|
| 1226 |
+
"step": 14500
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 7.08,
|
| 1230 |
+
"learning_rate": 5.84237848041529e-06,
|
| 1231 |
+
"loss": 0.0118,
|
| 1232 |
+
"step": 15000
|
| 1233 |
+
},
|
| 1234 |
+
{
|
| 1235 |
+
"epoch": 7.08,
|
| 1236 |
+
"eval_LOC": {
|
| 1237 |
+
"f1": 0.8129576567971983,
|
| 1238 |
+
"number": 6349,
|
| 1239 |
+
"precision": 0.8217216411906677,
|
| 1240 |
+
"recall": 0.804378642305875
|
| 1241 |
+
},
|
| 1242 |
+
"eval_MISC": {
|
| 1243 |
+
"f1": 0.7294871794871794,
|
| 1244 |
+
"number": 1602,
|
| 1245 |
+
"precision": 0.7496706192358367,
|
| 1246 |
+
"recall": 0.7103620474406991
|
| 1247 |
+
},
|
| 1248 |
+
"eval_NORP": {
|
| 1249 |
+
"f1": 0.8883183568677792,
|
| 1250 |
+
"number": 5690,
|
| 1251 |
+
"precision": 0.8657214345287739,
|
| 1252 |
+
"recall": 0.9121265377855887
|
| 1253 |
+
},
|
| 1254 |
+
"eval_PER": {
|
| 1255 |
+
"f1": 0.9262795910221375,
|
| 1256 |
+
"number": 16183,
|
| 1257 |
+
"precision": 0.9205419260344196,
|
| 1258 |
+
"recall": 0.9320892294382994
|
| 1259 |
+
},
|
| 1260 |
+
"eval_loss": 0.18032006919384003,
|
| 1261 |
+
"eval_overall_accuracy": 0.9713643391927084,
|
| 1262 |
+
"eval_overall_f1": 0.8848810437452033,
|
| 1263 |
+
"eval_overall_precision": 0.8806203094906023,
|
| 1264 |
+
"eval_overall_recall": 0.8891832081545065,
|
| 1265 |
+
"eval_runtime": 36.5943,
|
| 1266 |
+
"eval_samples_per_second": 115.756,
|
| 1267 |
+
"eval_steps_per_second": 14.483,
|
| 1268 |
+
"step": 15000
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 7.31,
|
| 1272 |
+
"learning_rate": 5.3704577630958e-06,
|
| 1273 |
+
"loss": 0.0103,
|
| 1274 |
+
"step": 15500
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"epoch": 7.31,
|
| 1278 |
+
"eval_LOC": {
|
| 1279 |
+
"f1": 0.8092074038917892,
|
| 1280 |
+
"number": 6349,
|
| 1281 |
+
"precision": 0.812807881773399,
|
| 1282 |
+
"recall": 0.8056386832572059
|
| 1283 |
+
},
|
| 1284 |
+
"eval_MISC": {
|
| 1285 |
+
"f1": 0.7470997679814386,
|
| 1286 |
+
"number": 1602,
|
| 1287 |
+
"precision": 0.7964664310954064,
|
| 1288 |
+
"recall": 0.7034956304619226
|
| 1289 |
+
},
|
| 1290 |
+
"eval_NORP": {
|
| 1291 |
+
"f1": 0.8900523560209423,
|
| 1292 |
+
"number": 5690,
|
| 1293 |
+
"precision": 0.8698204999161214,
|
| 1294 |
+
"recall": 0.9112478031634447
|
| 1295 |
+
},
|
| 1296 |
+
"eval_PER": {
|
| 1297 |
+
"f1": 0.9261114012719285,
|
| 1298 |
+
"number": 16183,
|
| 1299 |
+
"precision": 0.9209336429182451,
|
| 1300 |
+
"recall": 0.9313477105604647
|
| 1301 |
+
},
|
| 1302 |
+
"eval_loss": 0.17802822589874268,
|
| 1303 |
+
"eval_overall_accuracy": 0.9718373616536459,
|
| 1304 |
+
"eval_overall_f1": 0.8853806445146092,
|
| 1305 |
+
"eval_overall_precision": 0.8822706842017646,
|
| 1306 |
+
"eval_overall_recall": 0.8885126072961373,
|
| 1307 |
+
"eval_runtime": 34.0065,
|
| 1308 |
+
"eval_samples_per_second": 124.565,
|
| 1309 |
+
"eval_steps_per_second": 15.585,
|
| 1310 |
+
"step": 15500
|
| 1311 |
+
},
|
| 1312 |
+
{
|
| 1313 |
+
"epoch": 7.55,
|
| 1314 |
+
"learning_rate": 4.89853704577631e-06,
|
| 1315 |
+
"loss": 0.0081,
|
| 1316 |
+
"step": 16000
|
| 1317 |
+
},
|
| 1318 |
+
{
|
| 1319 |
+
"epoch": 7.55,
|
| 1320 |
+
"eval_LOC": {
|
| 1321 |
+
"f1": 0.8206599713055953,
|
| 1322 |
+
"number": 6349,
|
| 1323 |
+
"precision": 0.8307245441342586,
|
| 1324 |
+
"recall": 0.8108363521814459
|
| 1325 |
+
},
|
| 1326 |
+
"eval_MISC": {
|
| 1327 |
+
"f1": 0.7422750424448217,
|
| 1328 |
+
"number": 1602,
|
| 1329 |
+
"precision": 0.813849590469099,
|
| 1330 |
+
"recall": 0.6822721598002497
|
| 1331 |
+
},
|
| 1332 |
+
"eval_NORP": {
|
| 1333 |
+
"f1": 0.8894804637183341,
|
| 1334 |
+
"number": 5690,
|
| 1335 |
+
"precision": 0.8696893366918556,
|
| 1336 |
+
"recall": 0.9101933216168717
|
| 1337 |
+
},
|
| 1338 |
+
"eval_PER": {
|
| 1339 |
+
"f1": 0.9277675840978593,
|
| 1340 |
+
"number": 16183,
|
| 1341 |
+
"precision": 0.9183871163044136,
|
| 1342 |
+
"recall": 0.9373416548229624
|
| 1343 |
+
},
|
| 1344 |
+
"eval_loss": 0.1846681833267212,
|
| 1345 |
+
"eval_overall_accuracy": 0.9726308186848959,
|
| 1346 |
+
"eval_overall_f1": 0.8887291931278828,
|
| 1347 |
+
"eval_overall_precision": 0.8859456217512994,
|
| 1348 |
+
"eval_overall_recall": 0.8915303111587983,
|
| 1349 |
+
"eval_runtime": 35.2193,
|
| 1350 |
+
"eval_samples_per_second": 120.275,
|
| 1351 |
+
"eval_steps_per_second": 15.049,
|
| 1352 |
+
"step": 16000
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 7.79,
|
| 1356 |
+
"learning_rate": 4.42661632845682e-06,
|
| 1357 |
+
"loss": 0.0098,
|
| 1358 |
+
"step": 16500
|
| 1359 |
+
},
|
| 1360 |
+
{
|
| 1361 |
+
"epoch": 7.79,
|
| 1362 |
+
"eval_LOC": {
|
| 1363 |
+
"f1": 0.8185487061802642,
|
| 1364 |
+
"number": 6349,
|
| 1365 |
+
"precision": 0.8225190839694656,
|
| 1366 |
+
"recall": 0.8146164750354387
|
| 1367 |
+
},
|
| 1368 |
+
"eval_MISC": {
|
| 1369 |
+
"f1": 0.7496598639455783,
|
| 1370 |
+
"number": 1602,
|
| 1371 |
+
"precision": 0.8236173393124065,
|
| 1372 |
+
"recall": 0.6878901373283396
|
| 1373 |
+
},
|
| 1374 |
+
"eval_NORP": {
|
| 1375 |
+
"f1": 0.8895378800893318,
|
| 1376 |
+
"number": 5690,
|
| 1377 |
+
"precision": 0.8699596774193549,
|
| 1378 |
+
"recall": 0.9100175746924429
|
| 1379 |
+
},
|
| 1380 |
+
"eval_PER": {
|
| 1381 |
+
"f1": 0.9280763899584168,
|
| 1382 |
+
"number": 16183,
|
| 1383 |
+
"precision": 0.925254882692544,
|
| 1384 |
+
"recall": 0.9309151578817277
|
| 1385 |
+
},
|
| 1386 |
+
"eval_loss": 0.18522000312805176,
|
| 1387 |
+
"eval_overall_accuracy": 0.9727986653645834,
|
| 1388 |
+
"eval_overall_f1": 0.8885798538971919,
|
| 1389 |
+
"eval_overall_precision": 0.8880442062960482,
|
| 1390 |
+
"eval_overall_recall": 0.8891161480686696,
|
| 1391 |
+
"eval_runtime": 33.8843,
|
| 1392 |
+
"eval_samples_per_second": 125.014,
|
| 1393 |
+
"eval_steps_per_second": 15.641,
|
| 1394 |
+
"step": 16500
|
| 1395 |
+
},
|
| 1396 |
+
{
|
| 1397 |
+
"epoch": 8.02,
|
| 1398 |
+
"learning_rate": 3.954695611137329e-06,
|
| 1399 |
+
"loss": 0.0078,
|
| 1400 |
+
"step": 17000
|
| 1401 |
+
},
|
| 1402 |
+
{
|
| 1403 |
+
"epoch": 8.02,
|
| 1404 |
+
"eval_LOC": {
|
| 1405 |
+
"f1": 0.8156406785060126,
|
| 1406 |
+
"number": 6349,
|
| 1407 |
+
"precision": 0.8249033505154639,
|
| 1408 |
+
"recall": 0.806583713970704
|
| 1409 |
+
},
|
| 1410 |
+
"eval_MISC": {
|
| 1411 |
+
"f1": 0.7453208556149733,
|
| 1412 |
+
"number": 1602,
|
| 1413 |
+
"precision": 0.802158273381295,
|
| 1414 |
+
"recall": 0.6960049937578028
|
| 1415 |
+
},
|
| 1416 |
+
"eval_NORP": {
|
| 1417 |
+
"f1": 0.8898399383719935,
|
| 1418 |
+
"number": 5690,
|
| 1419 |
+
"precision": 0.8673452361087935,
|
| 1420 |
+
"recall": 0.9135325131810194
|
| 1421 |
+
},
|
| 1422 |
+
"eval_PER": {
|
| 1423 |
+
"f1": 0.9259453619385708,
|
| 1424 |
+
"number": 16183,
|
| 1425 |
+
"precision": 0.9251171971379225,
|
| 1426 |
+
"recall": 0.926775010813817
|
| 1427 |
+
},
|
| 1428 |
+
"eval_loss": 0.19055458903312683,
|
| 1429 |
+
"eval_overall_accuracy": 0.9721832275390625,
|
| 1430 |
+
"eval_overall_f1": 0.886578227983967,
|
| 1431 |
+
"eval_overall_precision": 0.8868905814850854,
|
| 1432 |
+
"eval_overall_recall": 0.8862660944206009,
|
| 1433 |
+
"eval_runtime": 35.3527,
|
| 1434 |
+
"eval_samples_per_second": 119.821,
|
| 1435 |
+
"eval_steps_per_second": 14.992,
|
| 1436 |
+
"step": 17000
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 8.26,
|
| 1440 |
+
"learning_rate": 3.482774893817839e-06,
|
| 1441 |
+
"loss": 0.0063,
|
| 1442 |
+
"step": 17500
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 8.26,
|
| 1446 |
+
"eval_LOC": {
|
| 1447 |
+
"f1": 0.8183385579937305,
|
| 1448 |
+
"number": 6349,
|
| 1449 |
+
"precision": 0.8143815317423179,
|
| 1450 |
+
"recall": 0.8223342258623405
|
| 1451 |
+
},
|
| 1452 |
+
"eval_MISC": {
|
| 1453 |
+
"f1": 0.7419786096256684,
|
| 1454 |
+
"number": 1602,
|
| 1455 |
+
"precision": 0.7985611510791367,
|
| 1456 |
+
"recall": 0.6928838951310862
|
| 1457 |
+
},
|
| 1458 |
+
"eval_NORP": {
|
| 1459 |
+
"f1": 0.8902174854865263,
|
| 1460 |
+
"number": 5690,
|
| 1461 |
+
"precision": 0.8779695778499402,
|
| 1462 |
+
"recall": 0.9028119507908612
|
| 1463 |
+
},
|
| 1464 |
+
"eval_PER": {
|
| 1465 |
+
"f1": 0.926946847960445,
|
| 1466 |
+
"number": 16183,
|
| 1467 |
+
"precision": 0.927118748840947,
|
| 1468 |
+
"recall": 0.926775010813817
|
| 1469 |
+
},
|
| 1470 |
+
"eval_loss": 0.19118891656398773,
|
| 1471 |
+
"eval_overall_accuracy": 0.9723765055338541,
|
| 1472 |
+
"eval_overall_f1": 0.8873317352019178,
|
| 1473 |
+
"eval_overall_precision": 0.8872573669918535,
|
| 1474 |
+
"eval_overall_recall": 0.8874061158798283,
|
| 1475 |
+
"eval_runtime": 35.4944,
|
| 1476 |
+
"eval_samples_per_second": 119.343,
|
| 1477 |
+
"eval_steps_per_second": 14.932,
|
| 1478 |
+
"step": 17500
|
| 1479 |
+
},
|
| 1480 |
+
{
|
| 1481 |
+
"epoch": 8.49,
|
| 1482 |
+
"learning_rate": 3.0108541764983485e-06,
|
| 1483 |
+
"loss": 0.0059,
|
| 1484 |
+
"step": 18000
|
| 1485 |
+
},
|
| 1486 |
+
{
|
| 1487 |
+
"epoch": 8.49,
|
| 1488 |
+
"eval_LOC": {
|
| 1489 |
+
"f1": 0.8186487554326354,
|
| 1490 |
+
"number": 6349,
|
| 1491 |
+
"precision": 0.8214398985093562,
|
| 1492 |
+
"recall": 0.8158765159867696
|
| 1493 |
+
},
|
| 1494 |
+
"eval_MISC": {
|
| 1495 |
+
"f1": 0.7347328244274809,
|
| 1496 |
+
"number": 1602,
|
| 1497 |
+
"precision": 0.7490272373540856,
|
| 1498 |
+
"recall": 0.7209737827715356
|
| 1499 |
+
},
|
| 1500 |
+
"eval_NORP": {
|
| 1501 |
+
"f1": 0.8874354561101548,
|
| 1502 |
+
"number": 5690,
|
| 1503 |
+
"precision": 0.869477234401349,
|
| 1504 |
+
"recall": 0.9061511423550088
|
| 1505 |
+
},
|
| 1506 |
+
"eval_PER": {
|
| 1507 |
+
"f1": 0.9275050685015666,
|
| 1508 |
+
"number": 16183,
|
| 1509 |
+
"precision": 0.9221794636857858,
|
| 1510 |
+
"recall": 0.9328925415559538
|
| 1511 |
+
},
|
| 1512 |
+
"eval_loss": 0.19088691473007202,
|
| 1513 |
+
"eval_overall_accuracy": 0.9722646077473959,
|
| 1514 |
+
"eval_overall_f1": 0.886665666216464,
|
| 1515 |
+
"eval_overall_precision": 0.8818866297389631,
|
| 1516 |
+
"eval_overall_recall": 0.8914967811158798,
|
| 1517 |
+
"eval_runtime": 34.1683,
|
| 1518 |
+
"eval_samples_per_second": 123.975,
|
| 1519 |
+
"eval_steps_per_second": 15.511,
|
| 1520 |
+
"step": 18000
|
| 1521 |
+
},
|
| 1522 |
+
{
|
| 1523 |
+
"epoch": 8.73,
|
| 1524 |
+
"learning_rate": 2.5389334591788585e-06,
|
| 1525 |
+
"loss": 0.0059,
|
| 1526 |
+
"step": 18500
|
| 1527 |
+
},
|
| 1528 |
+
{
|
| 1529 |
+
"epoch": 8.73,
|
| 1530 |
+
"eval_LOC": {
|
| 1531 |
+
"f1": 0.8207652008255277,
|
| 1532 |
+
"number": 6349,
|
| 1533 |
+
"precision": 0.8273323731797088,
|
| 1534 |
+
"recall": 0.8143014647976059
|
| 1535 |
+
},
|
| 1536 |
+
"eval_MISC": {
|
| 1537 |
+
"f1": 0.7507487520798669,
|
| 1538 |
+
"number": 1602,
|
| 1539 |
+
"precision": 0.8039914468995011,
|
| 1540 |
+
"recall": 0.704119850187266
|
| 1541 |
+
},
|
| 1542 |
+
"eval_NORP": {
|
| 1543 |
+
"f1": 0.8878091106290672,
|
| 1544 |
+
"number": 5690,
|
| 1545 |
+
"precision": 0.8767780634104542,
|
| 1546 |
+
"recall": 0.8991212653778559
|
| 1547 |
+
},
|
| 1548 |
+
"eval_PER": {
|
| 1549 |
+
"f1": 0.9267059295118868,
|
| 1550 |
+
"number": 16183,
|
| 1551 |
+
"precision": 0.9248522895125554,
|
| 1552 |
+
"recall": 0.9285670147685843
|
| 1553 |
+
},
|
| 1554 |
+
"eval_loss": 0.19563351571559906,
|
| 1555 |
+
"eval_overall_accuracy": 0.9725545247395834,
|
| 1556 |
+
"eval_overall_f1": 0.8878926778488558,
|
| 1557 |
+
"eval_overall_precision": 0.8892214561964016,
|
| 1558 |
+
"eval_overall_recall": 0.886567864806867,
|
| 1559 |
+
"eval_runtime": 35.4075,
|
| 1560 |
+
"eval_samples_per_second": 119.636,
|
| 1561 |
+
"eval_steps_per_second": 14.969,
|
| 1562 |
+
"step": 18500
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 8.97,
|
| 1566 |
+
"learning_rate": 2.0670127418593677e-06,
|
| 1567 |
+
"loss": 0.0069,
|
| 1568 |
+
"step": 19000
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 8.97,
|
| 1572 |
+
"eval_LOC": {
|
| 1573 |
+
"f1": 0.8209361974991985,
|
| 1574 |
+
"number": 6349,
|
| 1575 |
+
"precision": 0.8358087155214624,
|
| 1576 |
+
"recall": 0.806583713970704
|
| 1577 |
+
},
|
| 1578 |
+
"eval_MISC": {
|
| 1579 |
+
"f1": 0.7553444180522565,
|
| 1580 |
+
"number": 1602,
|
| 1581 |
+
"precision": 0.8275092936802974,
|
| 1582 |
+
"recall": 0.6947565543071161
|
| 1583 |
+
},
|
| 1584 |
+
"eval_NORP": {
|
| 1585 |
+
"f1": 0.8875515818431913,
|
| 1586 |
+
"number": 5690,
|
| 1587 |
+
"precision": 0.8687310669808146,
|
| 1588 |
+
"recall": 0.9072056239015818
|
| 1589 |
+
},
|
| 1590 |
+
"eval_PER": {
|
| 1591 |
+
"f1": 0.9272833399895548,
|
| 1592 |
+
"number": 16183,
|
| 1593 |
+
"precision": 0.9220430107526881,
|
| 1594 |
+
"recall": 0.932583575356856
|
| 1595 |
+
},
|
| 1596 |
+
"eval_loss": 0.19536001980304718,
|
| 1597 |
+
"eval_overall_accuracy": 0.972808837890625,
|
| 1598 |
+
"eval_overall_f1": 0.8887695869543335,
|
| 1599 |
+
"eval_overall_precision": 0.8893962796319925,
|
| 1600 |
+
"eval_overall_recall": 0.8881437768240343,
|
| 1601 |
+
"eval_runtime": 35.4791,
|
| 1602 |
+
"eval_samples_per_second": 119.394,
|
| 1603 |
+
"eval_steps_per_second": 14.938,
|
| 1604 |
+
"step": 19000
|
| 1605 |
+
},
|
| 1606 |
+
{
|
| 1607 |
+
"epoch": 9.2,
|
| 1608 |
+
"learning_rate": 1.5950920245398775e-06,
|
| 1609 |
+
"loss": 0.006,
|
| 1610 |
+
"step": 19500
|
| 1611 |
+
},
|
| 1612 |
+
{
|
| 1613 |
+
"epoch": 9.2,
|
| 1614 |
+
"eval_LOC": {
|
| 1615 |
+
"f1": 0.8205210164615631,
|
| 1616 |
+
"number": 6349,
|
| 1617 |
+
"precision": 0.8327656123276561,
|
| 1618 |
+
"recall": 0.8086312805166168
|
| 1619 |
+
},
|
| 1620 |
+
"eval_MISC": {
|
| 1621 |
+
"f1": 0.7514871116986119,
|
| 1622 |
+
"number": 1602,
|
| 1623 |
+
"precision": 0.7984550561797753,
|
| 1624 |
+
"recall": 0.7097378277153558
|
| 1625 |
+
},
|
| 1626 |
+
"eval_NORP": {
|
| 1627 |
+
"f1": 0.8861538461538462,
|
| 1628 |
+
"number": 5690,
|
| 1629 |
+
"precision": 0.8625623960066556,
|
| 1630 |
+
"recall": 0.9110720562390158
|
| 1631 |
+
},
|
| 1632 |
+
"eval_PER": {
|
| 1633 |
+
"f1": 0.9265025233216089,
|
| 1634 |
+
"number": 16183,
|
| 1635 |
+
"precision": 0.9172722868217055,
|
| 1636 |
+
"recall": 0.9359204103071124
|
| 1637 |
+
},
|
| 1638 |
+
"eval_loss": 0.19643273949623108,
|
| 1639 |
+
"eval_overall_accuracy": 0.9723968505859375,
|
| 1640 |
+
"eval_overall_f1": 0.8876616334362226,
|
| 1641 |
+
"eval_overall_precision": 0.8834313041745542,
|
| 1642 |
+
"eval_overall_recall": 0.8919326716738197,
|
| 1643 |
+
"eval_runtime": 34.2373,
|
| 1644 |
+
"eval_samples_per_second": 123.725,
|
| 1645 |
+
"eval_steps_per_second": 15.48,
|
| 1646 |
+
"step": 19500
|
| 1647 |
+
},
|
| 1648 |
+
{
|
| 1649 |
+
"epoch": 9.44,
|
| 1650 |
+
"learning_rate": 1.123171307220387e-06,
|
| 1651 |
+
"loss": 0.0046,
|
| 1652 |
+
"step": 20000
|
| 1653 |
+
},
|
| 1654 |
+
{
|
| 1655 |
+
"epoch": 9.44,
|
| 1656 |
+
"eval_LOC": {
|
| 1657 |
+
"f1": 0.8190899001109878,
|
| 1658 |
+
"number": 6349,
|
| 1659 |
+
"precision": 0.8245810055865922,
|
| 1660 |
+
"recall": 0.8136714443219405
|
| 1661 |
+
},
|
| 1662 |
+
"eval_MISC": {
|
| 1663 |
+
"f1": 0.7499167499167498,
|
| 1664 |
+
"number": 1602,
|
| 1665 |
+
"precision": 0.8037116345467523,
|
| 1666 |
+
"recall": 0.7028714107365793
|
| 1667 |
+
},
|
| 1668 |
+
"eval_NORP": {
|
| 1669 |
+
"f1": 0.8883740883740884,
|
| 1670 |
+
"number": 5690,
|
| 1671 |
+
"precision": 0.8678960603520537,
|
| 1672 |
+
"recall": 0.909841827768014
|
| 1673 |
+
},
|
| 1674 |
+
"eval_PER": {
|
| 1675 |
+
"f1": 0.9276177525033996,
|
| 1676 |
+
"number": 16183,
|
| 1677 |
+
"precision": 0.9279045322450998,
|
| 1678 |
+
"recall": 0.927331149972193
|
| 1679 |
+
},
|
| 1680 |
+
"eval_loss": 0.19752554595470428,
|
| 1681 |
+
"eval_overall_accuracy": 0.9726409912109375,
|
| 1682 |
+
"eval_overall_f1": 0.8880391762259341,
|
| 1683 |
+
"eval_overall_precision": 0.8883371359549054,
|
| 1684 |
+
"eval_overall_recall": 0.8877414163090128,
|
| 1685 |
+
"eval_runtime": 35.5214,
|
| 1686 |
+
"eval_samples_per_second": 119.252,
|
| 1687 |
+
"eval_steps_per_second": 14.921,
|
| 1688 |
+
"step": 20000
|
| 1689 |
+
},
|
| 1690 |
+
{
|
| 1691 |
+
"epoch": 9.67,
|
| 1692 |
+
"learning_rate": 6.512505899008966e-07,
|
| 1693 |
+
"loss": 0.0038,
|
| 1694 |
+
"step": 20500
|
| 1695 |
+
},
|
| 1696 |
+
{
|
| 1697 |
+
"epoch": 9.67,
|
| 1698 |
+
"eval_LOC": {
|
| 1699 |
+
"f1": 0.8203611954610835,
|
| 1700 |
+
"number": 6349,
|
| 1701 |
+
"precision": 0.832603406326034,
|
| 1702 |
+
"recall": 0.8084737753977004
|
| 1703 |
+
},
|
| 1704 |
+
"eval_MISC": {
|
| 1705 |
+
"f1": 0.7467811158798283,
|
| 1706 |
+
"number": 1602,
|
| 1707 |
+
"precision": 0.7925718290119131,
|
| 1708 |
+
"recall": 0.7059925093632958
|
| 1709 |
+
},
|
| 1710 |
+
"eval_NORP": {
|
| 1711 |
+
"f1": 0.8882978723404255,
|
| 1712 |
+
"number": 5690,
|
| 1713 |
+
"precision": 0.8677505866577271,
|
| 1714 |
+
"recall": 0.909841827768014
|
| 1715 |
+
},
|
| 1716 |
+
"eval_PER": {
|
| 1717 |
+
"f1": 0.9284789444478698,
|
| 1718 |
+
"number": 16183,
|
| 1719 |
+
"precision": 0.9264226391879422,
|
| 1720 |
+
"recall": 0.9305443984428103
|
| 1721 |
+
},
|
| 1722 |
+
"eval_loss": 0.1979084610939026,
|
| 1723 |
+
"eval_overall_accuracy": 0.972808837890625,
|
| 1724 |
+
"eval_overall_f1": 0.8887100290088367,
|
| 1725 |
+
"eval_overall_precision": 0.8888739811491632,
|
| 1726 |
+
"eval_overall_recall": 0.8885461373390557,
|
| 1727 |
+
"eval_runtime": 34.2132,
|
| 1728 |
+
"eval_samples_per_second": 123.812,
|
| 1729 |
+
"eval_steps_per_second": 15.491,
|
| 1730 |
+
"step": 20500
|
| 1731 |
+
},
|
| 1732 |
+
{
|
| 1733 |
+
"epoch": 9.91,
|
| 1734 |
+
"learning_rate": 1.7932987258140632e-07,
|
| 1735 |
+
"loss": 0.0057,
|
| 1736 |
+
"step": 21000
|
| 1737 |
+
},
|
| 1738 |
+
{
|
| 1739 |
+
"epoch": 9.91,
|
| 1740 |
+
"eval_LOC": {
|
| 1741 |
+
"f1": 0.8203578528827038,
|
| 1742 |
+
"number": 6349,
|
| 1743 |
+
"precision": 0.8284612913588179,
|
| 1744 |
+
"recall": 0.8124114033706096
|
| 1745 |
+
},
|
| 1746 |
+
"eval_MISC": {
|
| 1747 |
+
"f1": 0.748,
|
| 1748 |
+
"number": 1602,
|
| 1749 |
+
"precision": 0.8025751072961373,
|
| 1750 |
+
"recall": 0.700374531835206
|
| 1751 |
+
},
|
| 1752 |
+
"eval_NORP": {
|
| 1753 |
+
"f1": 0.8893664574916187,
|
| 1754 |
+
"number": 5690,
|
| 1755 |
+
"precision": 0.8704358068315665,
|
| 1756 |
+
"recall": 0.9091388400702988
|
| 1757 |
+
},
|
| 1758 |
+
"eval_PER": {
|
| 1759 |
+
"f1": 0.9278439602493672,
|
| 1760 |
+
"number": 16183,
|
| 1761 |
+
"precision": 0.9268142302238116,
|
| 1762 |
+
"recall": 0.9288759809676821
|
| 1763 |
+
},
|
| 1764 |
+
"eval_loss": 0.19716496765613556,
|
| 1765 |
+
"eval_overall_accuracy": 0.9727630615234375,
|
| 1766 |
+
"eval_overall_f1": 0.8886092937426606,
|
| 1767 |
+
"eval_overall_precision": 0.8891761230108104,
|
| 1768 |
+
"eval_overall_recall": 0.8880431866952789,
|
| 1769 |
+
"eval_runtime": 34.0058,
|
| 1770 |
+
"eval_samples_per_second": 124.567,
|
| 1771 |
+
"eval_steps_per_second": 15.586,
|
| 1772 |
+
"step": 21000
|
| 1773 |
+
}
|
| 1774 |
+
],
|
| 1775 |
+
"max_steps": 21190,
|
| 1776 |
+
"num_train_epochs": 10,
|
| 1777 |
+
"total_flos": 8856293355753336.0,
|
| 1778 |
+
"trial_name": null,
|
| 1779 |
+
"trial_params": null
|
| 1780 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e3fbc8399115666d9f122d90acf36c2334aa64ca15e6d86d505c9d53cc11e2d5
|
| 3 |
+
size 4088
|
vocab.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|