Commit
•
5051d7c
1
Parent(s):
86a6322
Add sentence-level sentiment SloBERTa classifier
Browse files- config.json +39 -0
- pytorch_model.bin +3 -0
- rng_state.pth +3 -0
- sentencepiece.bpe.model +3 -0
- special_tokens_map.json +19 -0
- split_indices.json +0 -0
- tokenizer.json +0 -0
- tokenizer_config.json +26 -0
- trainer_state.json +2241 -0
- training_args.bin +3 -0
config.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "EMBEDDIA/sloberta",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"CamembertForSequenceClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.1,
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"classifier_dropout": null,
|
| 9 |
+
"eos_token_id": 2,
|
| 10 |
+
"gradient_checkpointing": false,
|
| 11 |
+
"hidden_act": "gelu",
|
| 12 |
+
"hidden_dropout_prob": 0.1,
|
| 13 |
+
"hidden_size": 768,
|
| 14 |
+
"id2label": {
|
| 15 |
+
"0": "negative",
|
| 16 |
+
"1": "neutral",
|
| 17 |
+
"2": "positive"
|
| 18 |
+
},
|
| 19 |
+
"initializer_range": 0.02,
|
| 20 |
+
"intermediate_size": 3072,
|
| 21 |
+
"label2id": {
|
| 22 |
+
"negative": 0,
|
| 23 |
+
"neutral": 1,
|
| 24 |
+
"positive": 2
|
| 25 |
+
},
|
| 26 |
+
"layer_norm_eps": 1e-05,
|
| 27 |
+
"max_position_embeddings": 514,
|
| 28 |
+
"model_type": "camembert",
|
| 29 |
+
"num_attention_heads": 12,
|
| 30 |
+
"num_hidden_layers": 12,
|
| 31 |
+
"pad_token_id": 1,
|
| 32 |
+
"position_embedding_type": "absolute",
|
| 33 |
+
"problem_type": "single_label_classification",
|
| 34 |
+
"torch_dtype": "float32",
|
| 35 |
+
"transformers_version": "4.25.1",
|
| 36 |
+
"type_vocab_size": 1,
|
| 37 |
+
"use_cache": true,
|
| 38 |
+
"vocab_size": 32005
|
| 39 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8441297dde8e83de2280edc30d806fb420869b0385df75a0f4b2317db17962bb
|
| 3 |
+
size 442568685
|
rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8aae008a2c67b657efc860d5a408a179eda460fc2031637ad0b8d2b788a19230
|
| 3 |
+
size 14567
|
sentencepiece.bpe.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:34b589385a2320549143ab23b0ccf82cc99a82685701cdabe0fad847bd0479ff
|
| 3 |
+
size 800013
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<s>NOTUSED",
|
| 4 |
+
"</s>NOTUSED"
|
| 5 |
+
],
|
| 6 |
+
"bos_token": "<s>",
|
| 7 |
+
"cls_token": "<s>",
|
| 8 |
+
"eos_token": "</s>",
|
| 9 |
+
"mask_token": {
|
| 10 |
+
"content": "<mask>",
|
| 11 |
+
"lstrip": true,
|
| 12 |
+
"normalized": true,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": "<pad>",
|
| 17 |
+
"sep_token": "</s>",
|
| 18 |
+
"unk_token": "<unk>"
|
| 19 |
+
}
|
split_indices.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<s>NOTUSED",
|
| 4 |
+
"</s>NOTUSED"
|
| 5 |
+
],
|
| 6 |
+
"bos_token": "<s>",
|
| 7 |
+
"cls_token": "<s>",
|
| 8 |
+
"do_lower_case": false,
|
| 9 |
+
"eos_token": "</s>",
|
| 10 |
+
"mask_token": {
|
| 11 |
+
"__type": "AddedToken",
|
| 12 |
+
"content": "<mask>",
|
| 13 |
+
"lstrip": true,
|
| 14 |
+
"normalized": true,
|
| 15 |
+
"rstrip": false,
|
| 16 |
+
"single_word": false
|
| 17 |
+
},
|
| 18 |
+
"model_max_length": 512,
|
| 19 |
+
"name_or_path": "EMBEDDIA/sloberta",
|
| 20 |
+
"pad_token": "<pad>",
|
| 21 |
+
"sep_token": "</s>",
|
| 22 |
+
"sp_model_kwargs": {},
|
| 23 |
+
"special_tokens_map_file": null,
|
| 24 |
+
"tokenizer_class": "CamembertTokenizer",
|
| 25 |
+
"unk_token": "<unk>"
|
| 26 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2241 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.6934678744913757,
|
| 3 |
+
"best_model_checkpoint": "sloberta_sentinews_sentence_2e-05_10e_lr2e-05/checkpoint-3560",
|
| 4 |
+
"epoch": 2.9966329966329965,
|
| 5 |
+
"global_step": 3560,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.03,
|
| 12 |
+
"learning_rate": 1.9932659932659936e-05,
|
| 13 |
+
"loss": 0.9684,
|
| 14 |
+
"step": 40
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.03,
|
| 18 |
+
"eval_accuracy": 0.5901716992303138,
|
| 19 |
+
"eval_f1_macro": 0.3290637211523573,
|
| 20 |
+
"eval_f1_negative": 0.25811001410437234,
|
| 21 |
+
"eval_f1_neutral": 0.7290811493526997,
|
| 22 |
+
"eval_f1_positive": 0.0,
|
| 23 |
+
"eval_loss": 0.9020117521286011,
|
| 24 |
+
"eval_precision_negative": 0.6455026455026455,
|
| 25 |
+
"eval_precision_neutral": 0.5861893881695862,
|
| 26 |
+
"eval_precision_positive": 0.0,
|
| 27 |
+
"eval_recall_negative": 0.16130453944468928,
|
| 28 |
+
"eval_recall_neutral": 0.9640918580375782,
|
| 29 |
+
"eval_recall_positive": 0.0,
|
| 30 |
+
"eval_runtime": 13.6446,
|
| 31 |
+
"eval_samples_per_second": 618.928,
|
| 32 |
+
"eval_steps_per_second": 3.225,
|
| 33 |
+
"step": 40
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"epoch": 0.07,
|
| 37 |
+
"learning_rate": 1.9865319865319866e-05,
|
| 38 |
+
"loss": 0.8398,
|
| 39 |
+
"step": 80
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.07,
|
| 43 |
+
"eval_accuracy": 0.6470100651272943,
|
| 44 |
+
"eval_f1_macro": 0.5536698584230076,
|
| 45 |
+
"eval_f1_negative": 0.5115520521880946,
|
| 46 |
+
"eval_f1_neutral": 0.7385869565217391,
|
| 47 |
+
"eval_f1_positive": 0.4108705665591893,
|
| 48 |
+
"eval_loss": 0.7759891152381897,
|
| 49 |
+
"eval_precision_negative": 0.6673758865248227,
|
| 50 |
+
"eval_precision_neutral": 0.65232,
|
| 51 |
+
"eval_precision_positive": 0.5681528662420382,
|
| 52 |
+
"eval_recall_negative": 0.41472014103129134,
|
| 53 |
+
"eval_recall_neutral": 0.8511482254697286,
|
| 54 |
+
"eval_recall_positive": 0.3217893217893218,
|
| 55 |
+
"eval_runtime": 13.475,
|
| 56 |
+
"eval_samples_per_second": 626.715,
|
| 57 |
+
"eval_steps_per_second": 3.265,
|
| 58 |
+
"step": 80
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.1,
|
| 62 |
+
"learning_rate": 1.97979797979798e-05,
|
| 63 |
+
"loss": 0.7748,
|
| 64 |
+
"step": 120
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"epoch": 0.1,
|
| 68 |
+
"eval_accuracy": 0.6824156305506217,
|
| 69 |
+
"eval_f1_macro": 0.6410968890440648,
|
| 70 |
+
"eval_f1_negative": 0.6465890942276981,
|
| 71 |
+
"eval_f1_neutral": 0.7394423944239442,
|
| 72 |
+
"eval_f1_positive": 0.5372591784805524,
|
| 73 |
+
"eval_loss": 0.7227517366409302,
|
| 74 |
+
"eval_precision_negative": 0.6702932828760644,
|
| 75 |
+
"eval_precision_neutral": 0.7263391059202577,
|
| 76 |
+
"eval_precision_positive": 0.5413919413919414,
|
| 77 |
+
"eval_recall_negative": 0.624504186866461,
|
| 78 |
+
"eval_recall_neutral": 0.753027139874739,
|
| 79 |
+
"eval_recall_positive": 0.5331890331890332,
|
| 80 |
+
"eval_runtime": 13.5517,
|
| 81 |
+
"eval_samples_per_second": 623.169,
|
| 82 |
+
"eval_steps_per_second": 3.247,
|
| 83 |
+
"step": 120
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"epoch": 0.13,
|
| 87 |
+
"learning_rate": 1.973063973063973e-05,
|
| 88 |
+
"loss": 0.7661,
|
| 89 |
+
"step": 160
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.13,
|
| 93 |
+
"eval_accuracy": 0.6966252220248668,
|
| 94 |
+
"eval_f1_macro": 0.6402639165706846,
|
| 95 |
+
"eval_f1_negative": 0.6606451612903226,
|
| 96 |
+
"eval_f1_neutral": 0.755861518507433,
|
| 97 |
+
"eval_f1_positive": 0.5042850699142984,
|
| 98 |
+
"eval_loss": 0.6990944147109985,
|
| 99 |
+
"eval_precision_negative": 0.6451070978580429,
|
| 100 |
+
"eval_precision_neutral": 0.7238677622778521,
|
| 101 |
+
"eval_precision_positive": 0.6726835138387485,
|
| 102 |
+
"eval_recall_negative": 0.6769501983252534,
|
| 103 |
+
"eval_recall_neutral": 0.7908141962421712,
|
| 104 |
+
"eval_recall_positive": 0.4033189033189033,
|
| 105 |
+
"eval_runtime": 13.529,
|
| 106 |
+
"eval_samples_per_second": 624.216,
|
| 107 |
+
"eval_steps_per_second": 3.252,
|
| 108 |
+
"step": 160
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"epoch": 0.17,
|
| 112 |
+
"learning_rate": 1.9663299663299665e-05,
|
| 113 |
+
"loss": 0.7329,
|
| 114 |
+
"step": 200
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.17,
|
| 118 |
+
"eval_accuracy": 0.6814683244523386,
|
| 119 |
+
"eval_f1_macro": 0.6584553155124641,
|
| 120 |
+
"eval_f1_negative": 0.6805819101834283,
|
| 121 |
+
"eval_f1_neutral": 0.7180338757887746,
|
| 122 |
+
"eval_f1_positive": 0.5767501605651895,
|
| 123 |
+
"eval_loss": 0.7035245299339294,
|
| 124 |
+
"eval_precision_negative": 0.6523848019401779,
|
| 125 |
+
"eval_precision_neutral": 0.7643176997407495,
|
| 126 |
+
"eval_precision_positive": 0.5196759259259259,
|
| 127 |
+
"eval_recall_negative": 0.7113265755839577,
|
| 128 |
+
"eval_recall_neutral": 0.677035490605428,
|
| 129 |
+
"eval_recall_positive": 0.6479076479076479,
|
| 130 |
+
"eval_runtime": 13.5578,
|
| 131 |
+
"eval_samples_per_second": 622.89,
|
| 132 |
+
"eval_steps_per_second": 3.245,
|
| 133 |
+
"step": 200
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"epoch": 0.2,
|
| 137 |
+
"learning_rate": 1.9595959595959596e-05,
|
| 138 |
+
"loss": 0.7084,
|
| 139 |
+
"step": 240
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"epoch": 0.2,
|
| 143 |
+
"eval_accuracy": 0.7044404973357016,
|
| 144 |
+
"eval_f1_macro": 0.6484202003006585,
|
| 145 |
+
"eval_f1_negative": 0.6737166776823088,
|
| 146 |
+
"eval_f1_neutral": 0.7610199644198458,
|
| 147 |
+
"eval_f1_positive": 0.5105239587998209,
|
| 148 |
+
"eval_loss": 0.6696082353591919,
|
| 149 |
+
"eval_precision_negative": 0.673568281938326,
|
| 150 |
+
"eval_precision_neutral": 0.7225975975975976,
|
| 151 |
+
"eval_precision_positive": 0.6729634002361276,
|
| 152 |
+
"eval_recall_negative": 0.6738651388276774,
|
| 153 |
+
"eval_recall_neutral": 0.8037578288100209,
|
| 154 |
+
"eval_recall_positive": 0.41125541125541126,
|
| 155 |
+
"eval_runtime": 13.468,
|
| 156 |
+
"eval_samples_per_second": 627.04,
|
| 157 |
+
"eval_steps_per_second": 3.267,
|
| 158 |
+
"step": 240
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 0.24,
|
| 162 |
+
"learning_rate": 1.952861952861953e-05,
|
| 163 |
+
"loss": 0.6878,
|
| 164 |
+
"step": 280
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.24,
|
| 168 |
+
"eval_accuracy": 0.6889283599763173,
|
| 169 |
+
"eval_f1_macro": 0.6646815029633141,
|
| 170 |
+
"eval_f1_negative": 0.6859417557091976,
|
| 171 |
+
"eval_f1_neutral": 0.7256559446701065,
|
| 172 |
+
"eval_f1_positive": 0.5824468085106382,
|
| 173 |
+
"eval_loss": 0.6821370720863342,
|
| 174 |
+
"eval_precision_negative": 0.6537539936102237,
|
| 175 |
+
"eval_precision_neutral": 0.7652234313498495,
|
| 176 |
+
"eval_precision_positive": 0.5400739827373613,
|
| 177 |
+
"eval_recall_negative": 0.7214631996474218,
|
| 178 |
+
"eval_recall_neutral": 0.6899791231732777,
|
| 179 |
+
"eval_recall_positive": 0.6320346320346321,
|
| 180 |
+
"eval_runtime": 13.4239,
|
| 181 |
+
"eval_samples_per_second": 629.1,
|
| 182 |
+
"eval_steps_per_second": 3.278,
|
| 183 |
+
"step": 280
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"epoch": 0.27,
|
| 187 |
+
"learning_rate": 1.9461279461279464e-05,
|
| 188 |
+
"loss": 0.7059,
|
| 189 |
+
"step": 320
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.27,
|
| 193 |
+
"eval_accuracy": 0.7119005328596802,
|
| 194 |
+
"eval_f1_macro": 0.6649492798540481,
|
| 195 |
+
"eval_f1_negative": 0.6837046745303625,
|
| 196 |
+
"eval_f1_neutral": 0.763837267704671,
|
| 197 |
+
"eval_f1_positive": 0.5473058973271108,
|
| 198 |
+
"eval_loss": 0.662620484828949,
|
| 199 |
+
"eval_precision_negative": 0.6777825898657427,
|
| 200 |
+
"eval_precision_neutral": 0.736108422071636,
|
| 201 |
+
"eval_precision_positive": 0.6642636457260556,
|
| 202 |
+
"eval_recall_negative": 0.6897311591009255,
|
| 203 |
+
"eval_recall_neutral": 0.7937369519832985,
|
| 204 |
+
"eval_recall_positive": 0.4653679653679654,
|
| 205 |
+
"eval_runtime": 13.4494,
|
| 206 |
+
"eval_samples_per_second": 627.911,
|
| 207 |
+
"eval_steps_per_second": 3.272,
|
| 208 |
+
"step": 320
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"epoch": 0.3,
|
| 212 |
+
"learning_rate": 1.9393939393939395e-05,
|
| 213 |
+
"loss": 0.6745,
|
| 214 |
+
"step": 360
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 0.3,
|
| 218 |
+
"eval_accuracy": 0.709295441089402,
|
| 219 |
+
"eval_f1_macro": 0.6599969904410593,
|
| 220 |
+
"eval_f1_negative": 0.6773485513608428,
|
| 221 |
+
"eval_f1_neutral": 0.7627118644067796,
|
| 222 |
+
"eval_f1_positive": 0.5399305555555556,
|
| 223 |
+
"eval_loss": 0.6554310917854309,
|
| 224 |
+
"eval_precision_negative": 0.674682990817665,
|
| 225 |
+
"eval_precision_neutral": 0.7299618320610687,
|
| 226 |
+
"eval_precision_positive": 0.6775599128540305,
|
| 227 |
+
"eval_recall_negative": 0.6800352578228295,
|
| 228 |
+
"eval_recall_neutral": 0.7985386221294363,
|
| 229 |
+
"eval_recall_positive": 0.44877344877344877,
|
| 230 |
+
"eval_runtime": 13.4392,
|
| 231 |
+
"eval_samples_per_second": 628.385,
|
| 232 |
+
"eval_steps_per_second": 3.274,
|
| 233 |
+
"step": 360
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.34,
|
| 237 |
+
"learning_rate": 1.932659932659933e-05,
|
| 238 |
+
"loss": 0.6824,
|
| 239 |
+
"step": 400
|
| 240 |
+
},
|
| 241 |
+
{
|
| 242 |
+
"epoch": 0.34,
|
| 243 |
+
"eval_accuracy": 0.7095322676139728,
|
| 244 |
+
"eval_f1_macro": 0.6638552526690601,
|
| 245 |
+
"eval_f1_negative": 0.6730469619407968,
|
| 246 |
+
"eval_f1_neutral": 0.7641443071048136,
|
| 247 |
+
"eval_f1_positive": 0.5543744889615698,
|
| 248 |
+
"eval_loss": 0.6521803736686707,
|
| 249 |
+
"eval_precision_negative": 0.6798561151079137,
|
| 250 |
+
"eval_precision_neutral": 0.7366789381902732,
|
| 251 |
+
"eval_precision_positive": 0.6396226415094339,
|
| 252 |
+
"eval_recall_negative": 0.6663728514764213,
|
| 253 |
+
"eval_recall_neutral": 0.7937369519832985,
|
| 254 |
+
"eval_recall_positive": 0.48917748917748916,
|
| 255 |
+
"eval_runtime": 13.4346,
|
| 256 |
+
"eval_samples_per_second": 628.599,
|
| 257 |
+
"eval_steps_per_second": 3.275,
|
| 258 |
+
"step": 400
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 0.37,
|
| 262 |
+
"learning_rate": 1.925925925925926e-05,
|
| 263 |
+
"loss": 0.669,
|
| 264 |
+
"step": 440
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 0.37,
|
| 268 |
+
"eval_accuracy": 0.713084665482534,
|
| 269 |
+
"eval_f1_macro": 0.6622007799006476,
|
| 270 |
+
"eval_f1_negative": 0.6819873057561829,
|
| 271 |
+
"eval_f1_neutral": 0.7671314741035858,
|
| 272 |
+
"eval_f1_positive": 0.5374835598421743,
|
| 273 |
+
"eval_loss": 0.6492008566856384,
|
| 274 |
+
"eval_precision_negative": 0.6773913043478261,
|
| 275 |
+
"eval_precision_neutral": 0.7335238095238096,
|
| 276 |
+
"eval_precision_positive": 0.6849162011173184,
|
| 277 |
+
"eval_recall_negative": 0.6866460996033495,
|
| 278 |
+
"eval_recall_neutral": 0.8039665970772443,
|
| 279 |
+
"eval_recall_positive": 0.4422799422799423,
|
| 280 |
+
"eval_runtime": 13.4531,
|
| 281 |
+
"eval_samples_per_second": 627.738,
|
| 282 |
+
"eval_steps_per_second": 3.271,
|
| 283 |
+
"step": 440
|
| 284 |
+
},
|
| 285 |
+
{
|
| 286 |
+
"epoch": 0.4,
|
| 287 |
+
"learning_rate": 1.9191919191919194e-05,
|
| 288 |
+
"loss": 0.6835,
|
| 289 |
+
"step": 480
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 0.4,
|
| 293 |
+
"eval_accuracy": 0.7045589105979869,
|
| 294 |
+
"eval_f1_macro": 0.6620625908815416,
|
| 295 |
+
"eval_f1_negative": 0.6357811729916215,
|
| 296 |
+
"eval_f1_neutral": 0.7631345490389355,
|
| 297 |
+
"eval_f1_positive": 0.5872720506140677,
|
| 298 |
+
"eval_loss": 0.6506330966949463,
|
| 299 |
+
"eval_precision_negative": 0.7210732252655114,
|
| 300 |
+
"eval_precision_neutral": 0.722875816993464,
|
| 301 |
+
"eval_precision_positive": 0.606456571867794,
|
| 302 |
+
"eval_recall_negative": 0.5685323931247246,
|
| 303 |
+
"eval_recall_neutral": 0.8081419624217119,
|
| 304 |
+
"eval_recall_positive": 0.5692640692640693,
|
| 305 |
+
"eval_runtime": 13.434,
|
| 306 |
+
"eval_samples_per_second": 628.63,
|
| 307 |
+
"eval_steps_per_second": 3.275,
|
| 308 |
+
"step": 480
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 0.44,
|
| 312 |
+
"learning_rate": 1.9124579124579125e-05,
|
| 313 |
+
"loss": 0.659,
|
| 314 |
+
"step": 520
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 0.44,
|
| 318 |
+
"eval_accuracy": 0.7040852575488454,
|
| 319 |
+
"eval_f1_macro": 0.6783997964045833,
|
| 320 |
+
"eval_f1_negative": 0.699721226602947,
|
| 321 |
+
"eval_f1_neutral": 0.7399162442142384,
|
| 322 |
+
"eval_f1_positive": 0.595561918396564,
|
| 323 |
+
"eval_loss": 0.6650780439376831,
|
| 324 |
+
"eval_precision_negative": 0.6382128586996004,
|
| 325 |
+
"eval_precision_neutral": 0.7836134453781513,
|
| 326 |
+
"eval_precision_positive": 0.5909090909090909,
|
| 327 |
+
"eval_recall_negative": 0.7743499338915822,
|
| 328 |
+
"eval_recall_neutral": 0.7008350730688935,
|
| 329 |
+
"eval_recall_positive": 0.6002886002886003,
|
| 330 |
+
"eval_runtime": 13.4366,
|
| 331 |
+
"eval_samples_per_second": 628.508,
|
| 332 |
+
"eval_steps_per_second": 3.275,
|
| 333 |
+
"step": 520
|
| 334 |
+
},
|
| 335 |
+
{
|
| 336 |
+
"epoch": 0.47,
|
| 337 |
+
"learning_rate": 1.905723905723906e-05,
|
| 338 |
+
"loss": 0.6554,
|
| 339 |
+
"step": 560
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"epoch": 0.47,
|
| 343 |
+
"eval_accuracy": 0.7166370633510953,
|
| 344 |
+
"eval_f1_macro": 0.6722586905255515,
|
| 345 |
+
"eval_f1_negative": 0.6801165657924232,
|
| 346 |
+
"eval_f1_neutral": 0.7694463910301331,
|
| 347 |
+
"eval_f1_positive": 0.5672131147540983,
|
| 348 |
+
"eval_loss": 0.6467012763023376,
|
| 349 |
+
"eval_precision_negative": 0.6920620437956204,
|
| 350 |
+
"eval_precision_neutral": 0.7391806116560877,
|
| 351 |
+
"eval_precision_positive": 0.6565464895635673,
|
| 352 |
+
"eval_recall_negative": 0.6685764654032613,
|
| 353 |
+
"eval_recall_neutral": 0.8022964509394572,
|
| 354 |
+
"eval_recall_positive": 0.49927849927849927,
|
| 355 |
+
"eval_runtime": 13.4494,
|
| 356 |
+
"eval_samples_per_second": 627.909,
|
| 357 |
+
"eval_steps_per_second": 3.272,
|
| 358 |
+
"step": 560
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"epoch": 0.51,
|
| 362 |
+
"learning_rate": 1.8989898989898993e-05,
|
| 363 |
+
"loss": 0.659,
|
| 364 |
+
"step": 600
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"epoch": 0.51,
|
| 368 |
+
"eval_accuracy": 0.712137359384251,
|
| 369 |
+
"eval_f1_macro": 0.6705187144410446,
|
| 370 |
+
"eval_f1_negative": 0.6719478098788444,
|
| 371 |
+
"eval_f1_neutral": 0.7654419861848032,
|
| 372 |
+
"eval_f1_positive": 0.5741663472594865,
|
| 373 |
+
"eval_loss": 0.6443530917167664,
|
| 374 |
+
"eval_precision_negative": 0.71280276816609,
|
| 375 |
+
"eval_precision_neutral": 0.7353337180226966,
|
| 376 |
+
"eval_precision_positive": 0.6124284546197875,
|
| 377 |
+
"eval_recall_negative": 0.6355222565006611,
|
| 378 |
+
"eval_recall_neutral": 0.7981210855949895,
|
| 379 |
+
"eval_recall_positive": 0.5404040404040404,
|
| 380 |
+
"eval_runtime": 13.4345,
|
| 381 |
+
"eval_samples_per_second": 628.604,
|
| 382 |
+
"eval_steps_per_second": 3.275,
|
| 383 |
+
"step": 600
|
| 384 |
+
},
|
| 385 |
+
{
|
| 386 |
+
"epoch": 0.54,
|
| 387 |
+
"learning_rate": 1.8922558922558924e-05,
|
| 388 |
+
"loss": 0.6692,
|
| 389 |
+
"step": 640
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 0.54,
|
| 393 |
+
"eval_accuracy": 0.7047957371225577,
|
| 394 |
+
"eval_f1_macro": 0.6746517865902923,
|
| 395 |
+
"eval_f1_negative": 0.6761710794297353,
|
| 396 |
+
"eval_f1_neutral": 0.7508104151416919,
|
| 397 |
+
"eval_f1_positive": 0.5969738651994497,
|
| 398 |
+
"eval_loss": 0.6483765244483948,
|
| 399 |
+
"eval_precision_negative": 0.6948837209302325,
|
| 400 |
+
"eval_precision_neutral": 0.7521474963335428,
|
| 401 |
+
"eval_precision_positive": 0.5703022339027596,
|
| 402 |
+
"eval_recall_negative": 0.6584398413397973,
|
| 403 |
+
"eval_recall_neutral": 0.7494780793319415,
|
| 404 |
+
"eval_recall_positive": 0.6262626262626263,
|
| 405 |
+
"eval_runtime": 13.5781,
|
| 406 |
+
"eval_samples_per_second": 621.958,
|
| 407 |
+
"eval_steps_per_second": 3.241,
|
| 408 |
+
"step": 640
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"epoch": 0.57,
|
| 412 |
+
"learning_rate": 1.8855218855218858e-05,
|
| 413 |
+
"loss": 0.6343,
|
| 414 |
+
"step": 680
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 0.57,
|
| 418 |
+
"eval_accuracy": 0.7084665482534044,
|
| 419 |
+
"eval_f1_macro": 0.6745823109427566,
|
| 420 |
+
"eval_f1_negative": 0.6674407803065491,
|
| 421 |
+
"eval_f1_neutral": 0.7580216636010628,
|
| 422 |
+
"eval_f1_positive": 0.5982844889206576,
|
| 423 |
+
"eval_loss": 0.6420348286628723,
|
| 424 |
+
"eval_precision_negative": 0.7054491899852725,
|
| 425 |
+
"eval_precision_neutral": 0.7423939151321057,
|
| 426 |
+
"eval_precision_positive": 0.5927762039660056,
|
| 427 |
+
"eval_recall_negative": 0.6333186425738211,
|
| 428 |
+
"eval_recall_neutral": 0.774321503131524,
|
| 429 |
+
"eval_recall_positive": 0.6038961038961039,
|
| 430 |
+
"eval_runtime": 13.5238,
|
| 431 |
+
"eval_samples_per_second": 624.455,
|
| 432 |
+
"eval_steps_per_second": 3.254,
|
| 433 |
+
"step": 680
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.61,
|
| 437 |
+
"learning_rate": 1.8787878787878792e-05,
|
| 438 |
+
"loss": 0.6707,
|
| 439 |
+
"step": 720
|
| 440 |
+
},
|
| 441 |
+
{
|
| 442 |
+
"epoch": 0.61,
|
| 443 |
+
"eval_accuracy": 0.7161634103019539,
|
| 444 |
+
"eval_f1_macro": 0.6635859690261697,
|
| 445 |
+
"eval_f1_negative": 0.6552984165651644,
|
| 446 |
+
"eval_f1_neutral": 0.7757296466973886,
|
| 447 |
+
"eval_f1_positive": 0.5597298438159561,
|
| 448 |
+
"eval_loss": 0.630727231502533,
|
| 449 |
+
"eval_precision_negative": 0.7325708061002179,
|
| 450 |
+
"eval_precision_neutral": 0.7180945609669392,
|
| 451 |
+
"eval_precision_positive": 0.6744659206510681,
|
| 452 |
+
"eval_recall_negative": 0.5927721463199648,
|
| 453 |
+
"eval_recall_neutral": 0.8434237995824635,
|
| 454 |
+
"eval_recall_positive": 0.47835497835497837,
|
| 455 |
+
"eval_runtime": 13.5358,
|
| 456 |
+
"eval_samples_per_second": 623.902,
|
| 457 |
+
"eval_steps_per_second": 3.251,
|
| 458 |
+
"step": 720
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 0.64,
|
| 462 |
+
"learning_rate": 1.8720538720538723e-05,
|
| 463 |
+
"loss": 0.6555,
|
| 464 |
+
"step": 760
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.64,
|
| 468 |
+
"eval_accuracy": 0.7173475429248076,
|
| 469 |
+
"eval_f1_macro": 0.6760934295779212,
|
| 470 |
+
"eval_f1_negative": 0.6785469499657298,
|
| 471 |
+
"eval_f1_neutral": 0.7685231476852316,
|
| 472 |
+
"eval_f1_positive": 0.5812101910828025,
|
| 473 |
+
"eval_loss": 0.6327303051948547,
|
| 474 |
+
"eval_precision_negative": 0.7044592030360531,
|
| 475 |
+
"eval_precision_neutral": 0.7374784110535406,
|
| 476 |
+
"eval_precision_positive": 0.6483126110124334,
|
| 477 |
+
"eval_recall_negative": 0.6544733362714852,
|
| 478 |
+
"eval_recall_neutral": 0.8022964509394572,
|
| 479 |
+
"eval_recall_positive": 0.5266955266955267,
|
| 480 |
+
"eval_runtime": 13.5283,
|
| 481 |
+
"eval_samples_per_second": 624.249,
|
| 482 |
+
"eval_steps_per_second": 3.252,
|
| 483 |
+
"step": 760
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"epoch": 0.67,
|
| 487 |
+
"learning_rate": 1.8653198653198653e-05,
|
| 488 |
+
"loss": 0.6539,
|
| 489 |
+
"step": 800
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 0.67,
|
| 493 |
+
"eval_accuracy": 0.713676731793961,
|
| 494 |
+
"eval_f1_macro": 0.6798716085872009,
|
| 495 |
+
"eval_f1_negative": 0.6834306241580603,
|
| 496 |
+
"eval_f1_neutral": 0.7602507450416196,
|
| 497 |
+
"eval_f1_positive": 0.5959334565619224,
|
| 498 |
+
"eval_loss": 0.63933926820755,
|
| 499 |
+
"eval_precision_negative": 0.6965675057208238,
|
| 500 |
+
"eval_precision_neutral": 0.7486338797814208,
|
| 501 |
+
"eval_precision_positive": 0.6110689916603488,
|
| 502 |
+
"eval_recall_negative": 0.6707800793301014,
|
| 503 |
+
"eval_recall_neutral": 0.7722338204592902,
|
| 504 |
+
"eval_recall_positive": 0.5815295815295816,
|
| 505 |
+
"eval_runtime": 13.556,
|
| 506 |
+
"eval_samples_per_second": 622.971,
|
| 507 |
+
"eval_steps_per_second": 3.246,
|
| 508 |
+
"step": 800
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"epoch": 0.71,
|
| 512 |
+
"learning_rate": 1.8585858585858588e-05,
|
| 513 |
+
"loss": 0.6586,
|
| 514 |
+
"step": 840
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"epoch": 0.71,
|
| 518 |
+
"eval_accuracy": 0.7122557726465364,
|
| 519 |
+
"eval_f1_macro": 0.6643997026372683,
|
| 520 |
+
"eval_f1_negative": 0.6466569058077111,
|
| 521 |
+
"eval_f1_neutral": 0.7709443099273607,
|
| 522 |
+
"eval_f1_positive": 0.575597892176733,
|
| 523 |
+
"eval_loss": 0.6333789229393005,
|
| 524 |
+
"eval_precision_negative": 0.7244395844723893,
|
| 525 |
+
"eval_precision_neutral": 0.7190605239385727,
|
| 526 |
+
"eval_precision_positive": 0.6567992599444958,
|
| 527 |
+
"eval_recall_negative": 0.5839576906126047,
|
| 528 |
+
"eval_recall_neutral": 0.8308977035490606,
|
| 529 |
+
"eval_recall_positive": 0.5122655122655123,
|
| 530 |
+
"eval_runtime": 13.615,
|
| 531 |
+
"eval_samples_per_second": 620.272,
|
| 532 |
+
"eval_steps_per_second": 3.232,
|
| 533 |
+
"step": 840
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.74,
|
| 537 |
+
"learning_rate": 1.851851851851852e-05,
|
| 538 |
+
"loss": 0.6579,
|
| 539 |
+
"step": 880
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.74,
|
| 543 |
+
"eval_accuracy": 0.7187685020722321,
|
| 544 |
+
"eval_f1_macro": 0.6807447825123648,
|
| 545 |
+
"eval_f1_negative": 0.6873633581110626,
|
| 546 |
+
"eval_f1_neutral": 0.7675631455159013,
|
| 547 |
+
"eval_f1_positive": 0.5873078439101301,
|
| 548 |
+
"eval_loss": 0.6417641639709473,
|
| 549 |
+
"eval_precision_negative": 0.6819956616052061,
|
| 550 |
+
"eval_precision_neutral": 0.7522549609140108,
|
| 551 |
+
"eval_precision_positive": 0.6472632493483927,
|
| 552 |
+
"eval_recall_negative": 0.6928162185985015,
|
| 553 |
+
"eval_recall_neutral": 0.7835073068893528,
|
| 554 |
+
"eval_recall_positive": 0.5375180375180375,
|
| 555 |
+
"eval_runtime": 13.429,
|
| 556 |
+
"eval_samples_per_second": 628.863,
|
| 557 |
+
"eval_steps_per_second": 3.276,
|
| 558 |
+
"step": 880
|
| 559 |
+
},
|
| 560 |
+
{
|
| 561 |
+
"epoch": 0.77,
|
| 562 |
+
"learning_rate": 1.8451178451178452e-05,
|
| 563 |
+
"loss": 0.6408,
|
| 564 |
+
"step": 920
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"epoch": 0.77,
|
| 568 |
+
"eval_accuracy": 0.7156897572528124,
|
| 569 |
+
"eval_f1_macro": 0.6733312463576527,
|
| 570 |
+
"eval_f1_negative": 0.6765443355368133,
|
| 571 |
+
"eval_f1_neutral": 0.7680304091227369,
|
| 572 |
+
"eval_f1_positive": 0.5754189944134079,
|
| 573 |
+
"eval_loss": 0.6319063901901245,
|
| 574 |
+
"eval_precision_negative": 0.7006610009442871,
|
| 575 |
+
"eval_precision_neutral": 0.7372767428461686,
|
| 576 |
+
"eval_precision_positive": 0.64375,
|
| 577 |
+
"eval_recall_negative": 0.6540326134861172,
|
| 578 |
+
"eval_recall_neutral": 0.8014613778705637,
|
| 579 |
+
"eval_recall_positive": 0.5202020202020202,
|
| 580 |
+
"eval_runtime": 13.4319,
|
| 581 |
+
"eval_samples_per_second": 628.725,
|
| 582 |
+
"eval_steps_per_second": 3.276,
|
| 583 |
+
"step": 920
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 0.81,
|
| 587 |
+
"learning_rate": 1.8383838383838387e-05,
|
| 588 |
+
"loss": 0.6458,
|
| 589 |
+
"step": 960
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.81,
|
| 593 |
+
"eval_accuracy": 0.7116637063351096,
|
| 594 |
+
"eval_f1_macro": 0.6808189121483205,
|
| 595 |
+
"eval_f1_negative": 0.6855624446412755,
|
| 596 |
+
"eval_f1_neutral": 0.7558841907935847,
|
| 597 |
+
"eval_f1_positive": 0.601010101010101,
|
| 598 |
+
"eval_loss": 0.6365010142326355,
|
| 599 |
+
"eval_precision_negative": 0.6889185580774366,
|
| 600 |
+
"eval_precision_neutral": 0.7541562759767249,
|
| 601 |
+
"eval_precision_positive": 0.601010101010101,
|
| 602 |
+
"eval_recall_negative": 0.6822388717496695,
|
| 603 |
+
"eval_recall_neutral": 0.7576200417536535,
|
| 604 |
+
"eval_recall_positive": 0.601010101010101,
|
| 605 |
+
"eval_runtime": 13.4342,
|
| 606 |
+
"eval_samples_per_second": 628.619,
|
| 607 |
+
"eval_steps_per_second": 3.275,
|
| 608 |
+
"step": 960
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 0.84,
|
| 612 |
+
"learning_rate": 1.831649831649832e-05,
|
| 613 |
+
"loss": 0.6663,
|
| 614 |
+
"step": 1000
|
| 615 |
+
},
|
| 616 |
+
{
|
| 617 |
+
"epoch": 0.84,
|
| 618 |
+
"eval_accuracy": 0.7177027827116637,
|
| 619 |
+
"eval_f1_macro": 0.6745871543495067,
|
| 620 |
+
"eval_f1_negative": 0.6834215167548501,
|
| 621 |
+
"eval_f1_neutral": 0.7686739720518749,
|
| 622 |
+
"eval_f1_positive": 0.5716659742417948,
|
| 623 |
+
"eval_loss": 0.6319110989570618,
|
| 624 |
+
"eval_precision_negative": 0.6837229819144244,
|
| 625 |
+
"eval_precision_neutral": 0.7413224743067675,
|
| 626 |
+
"eval_precision_positive": 0.67384916748286,
|
| 627 |
+
"eval_recall_negative": 0.6831203173204055,
|
| 628 |
+
"eval_recall_neutral": 0.7981210855949895,
|
| 629 |
+
"eval_recall_positive": 0.4963924963924964,
|
| 630 |
+
"eval_runtime": 13.4422,
|
| 631 |
+
"eval_samples_per_second": 628.246,
|
| 632 |
+
"eval_steps_per_second": 3.273,
|
| 633 |
+
"step": 1000
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 0.88,
|
| 637 |
+
"learning_rate": 1.824915824915825e-05,
|
| 638 |
+
"loss": 0.6456,
|
| 639 |
+
"step": 1040
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 0.88,
|
| 643 |
+
"eval_accuracy": 0.7104795737122558,
|
| 644 |
+
"eval_f1_macro": 0.6846884030136478,
|
| 645 |
+
"eval_f1_negative": 0.6933867735470942,
|
| 646 |
+
"eval_f1_neutral": 0.7508249068653539,
|
| 647 |
+
"eval_f1_positive": 0.6098535286284953,
|
| 648 |
+
"eval_loss": 0.6414014101028442,
|
| 649 |
+
"eval_precision_negative": 0.7007200720072008,
|
| 650 |
+
"eval_precision_neutral": 0.7659066232356134,
|
| 651 |
+
"eval_precision_positive": 0.5661310259579728,
|
| 652 |
+
"eval_recall_negative": 0.6862053768179814,
|
| 653 |
+
"eval_recall_neutral": 0.7363256784968685,
|
| 654 |
+
"eval_recall_positive": 0.6608946608946609,
|
| 655 |
+
"eval_runtime": 13.4341,
|
| 656 |
+
"eval_samples_per_second": 628.625,
|
| 657 |
+
"eval_steps_per_second": 3.275,
|
| 658 |
+
"step": 1040
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"epoch": 0.91,
|
| 662 |
+
"learning_rate": 1.8181818181818182e-05,
|
| 663 |
+
"loss": 0.6416,
|
| 664 |
+
"step": 1080
|
| 665 |
+
},
|
| 666 |
+
{
|
| 667 |
+
"epoch": 0.91,
|
| 668 |
+
"eval_accuracy": 0.7153345174659562,
|
| 669 |
+
"eval_f1_macro": 0.6849725647884108,
|
| 670 |
+
"eval_f1_negative": 0.6939308244507288,
|
| 671 |
+
"eval_f1_neutral": 0.7585337674614012,
|
| 672 |
+
"eval_f1_positive": 0.6024531024531025,
|
| 673 |
+
"eval_loss": 0.6341109871864319,
|
| 674 |
+
"eval_precision_negative": 0.6851374570446735,
|
| 675 |
+
"eval_precision_neutral": 0.7632635806383429,
|
| 676 |
+
"eval_precision_positive": 0.6024531024531025,
|
| 677 |
+
"eval_recall_negative": 0.7029528426619657,
|
| 678 |
+
"eval_recall_neutral": 0.7538622129436325,
|
| 679 |
+
"eval_recall_positive": 0.6024531024531025,
|
| 680 |
+
"eval_runtime": 13.5517,
|
| 681 |
+
"eval_samples_per_second": 623.169,
|
| 682 |
+
"eval_steps_per_second": 3.247,
|
| 683 |
+
"step": 1080
|
| 684 |
+
},
|
| 685 |
+
{
|
| 686 |
+
"epoch": 0.94,
|
| 687 |
+
"learning_rate": 1.8114478114478116e-05,
|
| 688 |
+
"loss": 0.6387,
|
| 689 |
+
"step": 1120
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.94,
|
| 693 |
+
"eval_accuracy": 0.7132030787448195,
|
| 694 |
+
"eval_f1_macro": 0.6812920768310707,
|
| 695 |
+
"eval_f1_negative": 0.6735266604303087,
|
| 696 |
+
"eval_f1_neutral": 0.7610456176319835,
|
| 697 |
+
"eval_f1_positive": 0.60930395243092,
|
| 698 |
+
"eval_loss": 0.6404346227645874,
|
| 699 |
+
"eval_precision_negative": 0.7174887892376681,
|
| 700 |
+
"eval_precision_neutral": 0.7476334340382679,
|
| 701 |
+
"eval_precision_positive": 0.5913102511880516,
|
| 702 |
+
"eval_recall_negative": 0.6346408109299251,
|
| 703 |
+
"eval_recall_neutral": 0.7749478079331942,
|
| 704 |
+
"eval_recall_positive": 0.6284271284271284,
|
| 705 |
+
"eval_runtime": 13.5098,
|
| 706 |
+
"eval_samples_per_second": 625.102,
|
| 707 |
+
"eval_steps_per_second": 3.257,
|
| 708 |
+
"step": 1120
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 0.98,
|
| 712 |
+
"learning_rate": 1.804713804713805e-05,
|
| 713 |
+
"loss": 0.6483,
|
| 714 |
+
"step": 1160
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 0.98,
|
| 718 |
+
"eval_accuracy": 0.7200710479573712,
|
| 719 |
+
"eval_f1_macro": 0.6806619404824857,
|
| 720 |
+
"eval_f1_negative": 0.6825213576541214,
|
| 721 |
+
"eval_f1_neutral": 0.7708228674771426,
|
| 722 |
+
"eval_f1_positive": 0.5886415963161934,
|
| 723 |
+
"eval_loss": 0.6273576021194458,
|
| 724 |
+
"eval_precision_negative": 0.7167798254122212,
|
| 725 |
+
"eval_precision_neutral": 0.7429788882432694,
|
| 726 |
+
"eval_precision_positive": 0.6286885245901639,
|
| 727 |
+
"eval_recall_negative": 0.6513882767739092,
|
| 728 |
+
"eval_recall_neutral": 0.8008350730688936,
|
| 729 |
+
"eval_recall_positive": 0.5533910533910534,
|
| 730 |
+
"eval_runtime": 13.5101,
|
| 731 |
+
"eval_samples_per_second": 625.087,
|
| 732 |
+
"eval_steps_per_second": 3.257,
|
| 733 |
+
"step": 1160
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 1.01,
|
| 737 |
+
"learning_rate": 1.797979797979798e-05,
|
| 738 |
+
"loss": 0.6418,
|
| 739 |
+
"step": 1200
|
| 740 |
+
},
|
| 741 |
+
{
|
| 742 |
+
"epoch": 1.01,
|
| 743 |
+
"eval_accuracy": 0.7216104203670811,
|
| 744 |
+
"eval_f1_macro": 0.6749026822057264,
|
| 745 |
+
"eval_f1_negative": 0.6914728682170542,
|
| 746 |
+
"eval_f1_neutral": 0.7730991818000399,
|
| 747 |
+
"eval_f1_positive": 0.560135996600085,
|
| 748 |
+
"eval_loss": 0.6303337812423706,
|
| 749 |
+
"eval_precision_negative": 0.695013357079252,
|
| 750 |
+
"eval_precision_neutral": 0.7404434250764526,
|
| 751 |
+
"eval_precision_positive": 0.6814891416752844,
|
| 752 |
+
"eval_recall_negative": 0.6879682679594535,
|
| 753 |
+
"eval_recall_neutral": 0.808768267223382,
|
| 754 |
+
"eval_recall_positive": 0.4754689754689755,
|
| 755 |
+
"eval_runtime": 14.4031,
|
| 756 |
+
"eval_samples_per_second": 586.332,
|
| 757 |
+
"eval_steps_per_second": 3.055,
|
| 758 |
+
"step": 1200
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 1.04,
|
| 762 |
+
"learning_rate": 1.7912457912457915e-05,
|
| 763 |
+
"loss": 0.6201,
|
| 764 |
+
"step": 1240
|
| 765 |
+
},
|
| 766 |
+
{
|
| 767 |
+
"epoch": 1.04,
|
| 768 |
+
"eval_accuracy": 0.7193605683836589,
|
| 769 |
+
"eval_f1_macro": 0.6783071530429613,
|
| 770 |
+
"eval_f1_negative": 0.6780135229657263,
|
| 771 |
+
"eval_f1_neutral": 0.7719368378972615,
|
| 772 |
+
"eval_f1_positive": 0.584971098265896,
|
| 773 |
+
"eval_loss": 0.6306412220001221,
|
| 774 |
+
"eval_precision_negative": 0.7198019801980198,
|
| 775 |
+
"eval_precision_neutral": 0.7404141104294478,
|
| 776 |
+
"eval_precision_positive": 0.6277915632754343,
|
| 777 |
+
"eval_recall_negative": 0.6408109299250772,
|
| 778 |
+
"eval_recall_neutral": 0.8062630480167015,
|
| 779 |
+
"eval_recall_positive": 0.5476190476190477,
|
| 780 |
+
"eval_runtime": 13.6129,
|
| 781 |
+
"eval_samples_per_second": 620.369,
|
| 782 |
+
"eval_steps_per_second": 3.232,
|
| 783 |
+
"step": 1240
|
| 784 |
+
},
|
| 785 |
+
{
|
| 786 |
+
"epoch": 1.08,
|
| 787 |
+
"learning_rate": 1.7845117845117846e-05,
|
| 788 |
+
"loss": 0.5971,
|
| 789 |
+
"step": 1280
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"epoch": 1.08,
|
| 793 |
+
"eval_accuracy": 0.7166370633510953,
|
| 794 |
+
"eval_f1_macro": 0.6745910338462799,
|
| 795 |
+
"eval_f1_negative": 0.6869394005688033,
|
| 796 |
+
"eval_f1_neutral": 0.7662285136501515,
|
| 797 |
+
"eval_f1_positive": 0.5706051873198847,
|
| 798 |
+
"eval_loss": 0.6360566020011902,
|
| 799 |
+
"eval_precision_negative": 0.682015638575152,
|
| 800 |
+
"eval_precision_neutral": 0.7429411764705882,
|
| 801 |
+
"eval_precision_positive": 0.6644295302013423,
|
| 802 |
+
"eval_recall_negative": 0.6919347730277655,
|
| 803 |
+
"eval_recall_neutral": 0.7910229645093946,
|
| 804 |
+
"eval_recall_positive": 0.5,
|
| 805 |
+
"eval_runtime": 13.5493,
|
| 806 |
+
"eval_samples_per_second": 623.278,
|
| 807 |
+
"eval_steps_per_second": 3.247,
|
| 808 |
+
"step": 1280
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 1.11,
|
| 812 |
+
"learning_rate": 1.7777777777777777e-05,
|
| 813 |
+
"loss": 0.6101,
|
| 814 |
+
"step": 1320
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"epoch": 1.11,
|
| 818 |
+
"eval_accuracy": 0.7226761397276495,
|
| 819 |
+
"eval_f1_macro": 0.6791712549911954,
|
| 820 |
+
"eval_f1_negative": 0.6990124516960068,
|
| 821 |
+
"eval_f1_neutral": 0.7709451929895654,
|
| 822 |
+
"eval_f1_positive": 0.5675561202880136,
|
| 823 |
+
"eval_loss": 0.6358277201652527,
|
| 824 |
+
"eval_precision_negative": 0.6814566764336543,
|
| 825 |
+
"eval_precision_neutral": 0.748868333005314,
|
| 826 |
+
"eval_precision_positive": 0.6871794871794872,
|
| 827 |
+
"eval_recall_negative": 0.7174966945791097,
|
| 828 |
+
"eval_recall_neutral": 0.7943632567849687,
|
| 829 |
+
"eval_recall_positive": 0.4834054834054834,
|
| 830 |
+
"eval_runtime": 13.4061,
|
| 831 |
+
"eval_samples_per_second": 629.938,
|
| 832 |
+
"eval_steps_per_second": 3.282,
|
| 833 |
+
"step": 1320
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 1.14,
|
| 837 |
+
"learning_rate": 1.771043771043771e-05,
|
| 838 |
+
"loss": 0.5977,
|
| 839 |
+
"step": 1360
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 1.14,
|
| 843 |
+
"eval_accuracy": 0.7124925991711072,
|
| 844 |
+
"eval_f1_macro": 0.6812595953156201,
|
| 845 |
+
"eval_f1_negative": 0.6902101028162717,
|
| 846 |
+
"eval_f1_neutral": 0.756903198916328,
|
| 847 |
+
"eval_f1_positive": 0.5966654842142604,
|
| 848 |
+
"eval_loss": 0.640369713306427,
|
| 849 |
+
"eval_precision_negative": 0.7002267573696145,
|
| 850 |
+
"eval_precision_neutral": 0.7555648013313917,
|
| 851 |
+
"eval_precision_positive": 0.586880669923238,
|
| 852 |
+
"eval_recall_negative": 0.6804759806081975,
|
| 853 |
+
"eval_recall_neutral": 0.7582463465553236,
|
| 854 |
+
"eval_recall_positive": 0.6067821067821068,
|
| 855 |
+
"eval_runtime": 13.458,
|
| 856 |
+
"eval_samples_per_second": 627.51,
|
| 857 |
+
"eval_steps_per_second": 3.269,
|
| 858 |
+
"step": 1360
|
| 859 |
+
},
|
| 860 |
+
{
|
| 861 |
+
"epoch": 1.18,
|
| 862 |
+
"learning_rate": 1.7643097643097645e-05,
|
| 863 |
+
"loss": 0.6229,
|
| 864 |
+
"step": 1400
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 1.18,
|
| 868 |
+
"eval_accuracy": 0.720307874481942,
|
| 869 |
+
"eval_f1_macro": 0.6787577486913711,
|
| 870 |
+
"eval_f1_negative": 0.6789838337182449,
|
| 871 |
+
"eval_f1_neutral": 0.7721746063384493,
|
| 872 |
+
"eval_f1_positive": 0.5851148060174189,
|
| 873 |
+
"eval_loss": 0.625924825668335,
|
| 874 |
+
"eval_precision_negative": 0.7132459970887919,
|
| 875 |
+
"eval_precision_neutral": 0.7387490465293669,
|
| 876 |
+
"eval_precision_positive": 0.6482456140350877,
|
| 877 |
+
"eval_recall_negative": 0.6478624944909652,
|
| 878 |
+
"eval_recall_neutral": 0.808768267223382,
|
| 879 |
+
"eval_recall_positive": 0.5331890331890332,
|
| 880 |
+
"eval_runtime": 13.413,
|
| 881 |
+
"eval_samples_per_second": 629.615,
|
| 882 |
+
"eval_steps_per_second": 3.28,
|
| 883 |
+
"step": 1400
|
| 884 |
+
},
|
| 885 |
+
{
|
| 886 |
+
"epoch": 1.21,
|
| 887 |
+
"learning_rate": 1.7575757575757576e-05,
|
| 888 |
+
"loss": 0.5971,
|
| 889 |
+
"step": 1440
|
| 890 |
+
},
|
| 891 |
+
{
|
| 892 |
+
"epoch": 1.21,
|
| 893 |
+
"eval_accuracy": 0.7111900532859681,
|
| 894 |
+
"eval_f1_macro": 0.685070727390424,
|
| 895 |
+
"eval_f1_negative": 0.6965195246179966,
|
| 896 |
+
"eval_f1_neutral": 0.7495985440530993,
|
| 897 |
+
"eval_f1_positive": 0.6090941135001762,
|
| 898 |
+
"eval_loss": 0.6486834287643433,
|
| 899 |
+
"eval_precision_negative": 0.6717151043798608,
|
| 900 |
+
"eval_precision_neutral": 0.7692814765985497,
|
| 901 |
+
"eval_precision_positive": 0.5954514128187457,
|
| 902 |
+
"eval_recall_negative": 0.7232260907888938,
|
| 903 |
+
"eval_recall_neutral": 0.7308977035490606,
|
| 904 |
+
"eval_recall_positive": 0.6233766233766234,
|
| 905 |
+
"eval_runtime": 13.4217,
|
| 906 |
+
"eval_samples_per_second": 629.204,
|
| 907 |
+
"eval_steps_per_second": 3.278,
|
| 908 |
+
"step": 1440
|
| 909 |
+
},
|
| 910 |
+
{
|
| 911 |
+
"epoch": 1.25,
|
| 912 |
+
"learning_rate": 1.750841750841751e-05,
|
| 913 |
+
"loss": 0.6006,
|
| 914 |
+
"step": 1480
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 1.25,
|
| 918 |
+
"eval_accuracy": 0.7155713439905269,
|
| 919 |
+
"eval_f1_macro": 0.683204217410409,
|
| 920 |
+
"eval_f1_negative": 0.6757380073800737,
|
| 921 |
+
"eval_f1_neutral": 0.7631821439541312,
|
| 922 |
+
"eval_f1_positive": 0.6106925008970219,
|
| 923 |
+
"eval_loss": 0.6353930830955505,
|
| 924 |
+
"eval_precision_negative": 0.7087566521528785,
|
| 925 |
+
"eval_precision_neutral": 0.7488446855535463,
|
| 926 |
+
"eval_precision_positive": 0.6074232690935046,
|
| 927 |
+
"eval_recall_negative": 0.6456588805641251,
|
| 928 |
+
"eval_recall_neutral": 0.7780793319415449,
|
| 929 |
+
"eval_recall_positive": 0.613997113997114,
|
| 930 |
+
"eval_runtime": 13.4559,
|
| 931 |
+
"eval_samples_per_second": 627.605,
|
| 932 |
+
"eval_steps_per_second": 3.27,
|
| 933 |
+
"step": 1480
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"epoch": 1.28,
|
| 937 |
+
"learning_rate": 1.7441077441077444e-05,
|
| 938 |
+
"loss": 0.6104,
|
| 939 |
+
"step": 1520
|
| 940 |
+
},
|
| 941 |
+
{
|
| 942 |
+
"epoch": 1.28,
|
| 943 |
+
"eval_accuracy": 0.7127294256956779,
|
| 944 |
+
"eval_f1_macro": 0.6839464489013288,
|
| 945 |
+
"eval_f1_negative": 0.6825213576541214,
|
| 946 |
+
"eval_f1_neutral": 0.7576642335766423,
|
| 947 |
+
"eval_f1_positive": 0.6116537554732233,
|
| 948 |
+
"eval_loss": 0.6454949975013733,
|
| 949 |
+
"eval_precision_negative": 0.7167798254122212,
|
| 950 |
+
"eval_precision_neutral": 0.756875,
|
| 951 |
+
"eval_precision_positive": 0.5735944409349337,
|
| 952 |
+
"eval_recall_negative": 0.6513882767739092,
|
| 953 |
+
"eval_recall_neutral": 0.758455114822547,
|
| 954 |
+
"eval_recall_positive": 0.6551226551226551,
|
| 955 |
+
"eval_runtime": 13.4867,
|
| 956 |
+
"eval_samples_per_second": 626.174,
|
| 957 |
+
"eval_steps_per_second": 3.262,
|
| 958 |
+
"step": 1520
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 1.31,
|
| 962 |
+
"learning_rate": 1.7373737373737375e-05,
|
| 963 |
+
"loss": 0.6101,
|
| 964 |
+
"step": 1560
|
| 965 |
+
},
|
| 966 |
+
{
|
| 967 |
+
"epoch": 1.31,
|
| 968 |
+
"eval_accuracy": 0.7206631142687981,
|
| 969 |
+
"eval_f1_macro": 0.6779869326137852,
|
| 970 |
+
"eval_f1_negative": 0.6807543740059078,
|
| 971 |
+
"eval_f1_neutral": 0.772482552342971,
|
| 972 |
+
"eval_f1_positive": 0.5807238714924766,
|
| 973 |
+
"eval_loss": 0.6268090605735779,
|
| 974 |
+
"eval_precision_negative": 0.702626641651032,
|
| 975 |
+
"eval_precision_neutral": 0.7393129770992366,
|
| 976 |
+
"eval_precision_positive": 0.6654240447343895,
|
| 977 |
+
"eval_recall_negative": 0.6602027324812693,
|
| 978 |
+
"eval_recall_neutral": 0.808768267223382,
|
| 979 |
+
"eval_recall_positive": 0.5151515151515151,
|
| 980 |
+
"eval_runtime": 13.4458,
|
| 981 |
+
"eval_samples_per_second": 628.077,
|
| 982 |
+
"eval_steps_per_second": 3.272,
|
| 983 |
+
"step": 1560
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 1.35,
|
| 987 |
+
"learning_rate": 1.7306397306397305e-05,
|
| 988 |
+
"loss": 0.6022,
|
| 989 |
+
"step": 1600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 1.35,
|
| 993 |
+
"eval_accuracy": 0.7143872113676731,
|
| 994 |
+
"eval_f1_macro": 0.6836129674253361,
|
| 995 |
+
"eval_f1_negative": 0.6836007130124777,
|
| 996 |
+
"eval_f1_neutral": 0.7594515995014541,
|
| 997 |
+
"eval_f1_positive": 0.6077865897620764,
|
| 998 |
+
"eval_loss": 0.6394125819206238,
|
| 999 |
+
"eval_precision_negative": 0.6913023884632717,
|
| 1000 |
+
"eval_precision_neutral": 0.7556841670111616,
|
| 1001 |
+
"eval_precision_positive": 0.6073487031700289,
|
| 1002 |
+
"eval_recall_negative": 0.6760687527545174,
|
| 1003 |
+
"eval_recall_neutral": 0.7632567849686848,
|
| 1004 |
+
"eval_recall_positive": 0.6082251082251082,
|
| 1005 |
+
"eval_runtime": 13.4382,
|
| 1006 |
+
"eval_samples_per_second": 628.432,
|
| 1007 |
+
"eval_steps_per_second": 3.274,
|
| 1008 |
+
"step": 1600
|
| 1009 |
+
},
|
| 1010 |
+
{
|
| 1011 |
+
"epoch": 1.38,
|
| 1012 |
+
"learning_rate": 1.723905723905724e-05,
|
| 1013 |
+
"loss": 0.6236,
|
| 1014 |
+
"step": 1640
|
| 1015 |
+
},
|
| 1016 |
+
{
|
| 1017 |
+
"epoch": 1.38,
|
| 1018 |
+
"eval_accuracy": 0.7134399052693902,
|
| 1019 |
+
"eval_f1_macro": 0.6861857480530462,
|
| 1020 |
+
"eval_f1_negative": 0.683572710951526,
|
| 1021 |
+
"eval_f1_neutral": 0.7565409267626353,
|
| 1022 |
+
"eval_f1_positive": 0.6184436064449776,
|
| 1023 |
+
"eval_loss": 0.6394263505935669,
|
| 1024 |
+
"eval_precision_negative": 0.6963877457704618,
|
| 1025 |
+
"eval_precision_neutral": 0.7615823989845568,
|
| 1026 |
+
"eval_precision_positive": 0.589157413455258,
|
| 1027 |
+
"eval_recall_negative": 0.6712208021154694,
|
| 1028 |
+
"eval_recall_neutral": 0.7515657620041754,
|
| 1029 |
+
"eval_recall_positive": 0.6507936507936508,
|
| 1030 |
+
"eval_runtime": 13.4189,
|
| 1031 |
+
"eval_samples_per_second": 629.337,
|
| 1032 |
+
"eval_steps_per_second": 3.279,
|
| 1033 |
+
"step": 1640
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"epoch": 1.41,
|
| 1037 |
+
"learning_rate": 1.7171717171717173e-05,
|
| 1038 |
+
"loss": 0.6039,
|
| 1039 |
+
"step": 1680
|
| 1040 |
+
},
|
| 1041 |
+
{
|
| 1042 |
+
"epoch": 1.41,
|
| 1043 |
+
"eval_accuracy": 0.7182948490230906,
|
| 1044 |
+
"eval_f1_macro": 0.6829107741796993,
|
| 1045 |
+
"eval_f1_negative": 0.6881057268722467,
|
| 1046 |
+
"eval_f1_neutral": 0.7655158240854912,
|
| 1047 |
+
"eval_f1_positive": 0.5951107715813598,
|
| 1048 |
+
"eval_loss": 0.6307365298271179,
|
| 1049 |
+
"eval_precision_negative": 0.6878027300748569,
|
| 1050 |
+
"eval_precision_neutral": 0.7537434237150951,
|
| 1051 |
+
"eval_precision_positive": 0.6323051948051948,
|
| 1052 |
+
"eval_recall_negative": 0.6884089907448215,
|
| 1053 |
+
"eval_recall_neutral": 0.7776617954070981,
|
| 1054 |
+
"eval_recall_positive": 0.562049062049062,
|
| 1055 |
+
"eval_runtime": 13.423,
|
| 1056 |
+
"eval_samples_per_second": 629.145,
|
| 1057 |
+
"eval_steps_per_second": 3.278,
|
| 1058 |
+
"step": 1680
|
| 1059 |
+
},
|
| 1060 |
+
{
|
| 1061 |
+
"epoch": 1.45,
|
| 1062 |
+
"learning_rate": 1.7104377104377104e-05,
|
| 1063 |
+
"loss": 0.6059,
|
| 1064 |
+
"step": 1720
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 1.45,
|
| 1068 |
+
"eval_accuracy": 0.7173475429248076,
|
| 1069 |
+
"eval_f1_macro": 0.6865160173058559,
|
| 1070 |
+
"eval_f1_negative": 0.6937445699391833,
|
| 1071 |
+
"eval_f1_neutral": 0.7605015673981192,
|
| 1072 |
+
"eval_f1_positive": 0.6053019145802652,
|
| 1073 |
+
"eval_loss": 0.6336551904678345,
|
| 1074 |
+
"eval_precision_negative": 0.6839400428265524,
|
| 1075 |
+
"eval_precision_neutral": 0.7612970711297071,
|
| 1076 |
+
"eval_precision_positive": 0.6180451127819548,
|
| 1077 |
+
"eval_recall_negative": 0.7038342882327017,
|
| 1078 |
+
"eval_recall_neutral": 0.7597077244258873,
|
| 1079 |
+
"eval_recall_positive": 0.5930735930735931,
|
| 1080 |
+
"eval_runtime": 13.4767,
|
| 1081 |
+
"eval_samples_per_second": 626.636,
|
| 1082 |
+
"eval_steps_per_second": 3.265,
|
| 1083 |
+
"step": 1720
|
| 1084 |
+
},
|
| 1085 |
+
{
|
| 1086 |
+
"epoch": 1.48,
|
| 1087 |
+
"learning_rate": 1.7037037037037038e-05,
|
| 1088 |
+
"loss": 0.6042,
|
| 1089 |
+
"step": 1760
|
| 1090 |
+
},
|
| 1091 |
+
{
|
| 1092 |
+
"epoch": 1.48,
|
| 1093 |
+
"eval_accuracy": 0.7197158081705151,
|
| 1094 |
+
"eval_f1_macro": 0.685440373432607,
|
| 1095 |
+
"eval_f1_negative": 0.694493006993007,
|
| 1096 |
+
"eval_f1_neutral": 0.7645846217274789,
|
| 1097 |
+
"eval_f1_positive": 0.5972434915773354,
|
| 1098 |
+
"eval_loss": 0.63006192445755,
|
| 1099 |
+
"eval_precision_negative": 0.6887732986562636,
|
| 1100 |
+
"eval_precision_neutral": 0.7550895765472313,
|
| 1101 |
+
"eval_precision_positive": 0.636215334420881,
|
| 1102 |
+
"eval_recall_negative": 0.7003085059497576,
|
| 1103 |
+
"eval_recall_neutral": 0.774321503131524,
|
| 1104 |
+
"eval_recall_positive": 0.5627705627705628,
|
| 1105 |
+
"eval_runtime": 13.4204,
|
| 1106 |
+
"eval_samples_per_second": 629.266,
|
| 1107 |
+
"eval_steps_per_second": 3.279,
|
| 1108 |
+
"step": 1760
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"epoch": 1.52,
|
| 1112 |
+
"learning_rate": 1.6969696969696972e-05,
|
| 1113 |
+
"loss": 0.5907,
|
| 1114 |
+
"step": 1800
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"epoch": 1.52,
|
| 1118 |
+
"eval_accuracy": 0.7238602723505032,
|
| 1119 |
+
"eval_f1_macro": 0.6788816276493144,
|
| 1120 |
+
"eval_f1_negative": 0.6713483146067416,
|
| 1121 |
+
"eval_f1_neutral": 0.7784207945071114,
|
| 1122 |
+
"eval_f1_positive": 0.58687577383409,
|
| 1123 |
+
"eval_loss": 0.6279187202453613,
|
| 1124 |
+
"eval_precision_negative": 0.7159261108337494,
|
| 1125 |
+
"eval_precision_neutral": 0.7341350601295097,
|
| 1126 |
+
"eval_precision_positive": 0.6856316297010607,
|
| 1127 |
+
"eval_recall_negative": 0.6319964742177171,
|
| 1128 |
+
"eval_recall_neutral": 0.82839248434238,
|
| 1129 |
+
"eval_recall_positive": 0.512987012987013,
|
| 1130 |
+
"eval_runtime": 13.4379,
|
| 1131 |
+
"eval_samples_per_second": 628.446,
|
| 1132 |
+
"eval_steps_per_second": 3.274,
|
| 1133 |
+
"step": 1800
|
| 1134 |
+
},
|
| 1135 |
+
{
|
| 1136 |
+
"epoch": 1.55,
|
| 1137 |
+
"learning_rate": 1.6902356902356903e-05,
|
| 1138 |
+
"loss": 0.6135,
|
| 1139 |
+
"step": 1840
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 1.55,
|
| 1143 |
+
"eval_accuracy": 0.7229129662522202,
|
| 1144 |
+
"eval_f1_macro": 0.6795419420133372,
|
| 1145 |
+
"eval_f1_negative": 0.6882803943044906,
|
| 1146 |
+
"eval_f1_neutral": 0.7737461051361945,
|
| 1147 |
+
"eval_f1_positive": 0.5765993265993266,
|
| 1148 |
+
"eval_loss": 0.6268677711486816,
|
| 1149 |
+
"eval_precision_negative": 0.6842334494773519,
|
| 1150 |
+
"eval_precision_neutral": 0.7460748207016864,
|
| 1151 |
+
"eval_precision_positive": 0.6919191919191919,
|
| 1152 |
+
"eval_recall_negative": 0.6923754958131335,
|
| 1153 |
+
"eval_recall_neutral": 0.8035490605427975,
|
| 1154 |
+
"eval_recall_positive": 0.49422799422799424,
|
| 1155 |
+
"eval_runtime": 13.44,
|
| 1156 |
+
"eval_samples_per_second": 628.349,
|
| 1157 |
+
"eval_steps_per_second": 3.274,
|
| 1158 |
+
"step": 1840
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"epoch": 1.58,
|
| 1162 |
+
"learning_rate": 1.6835016835016837e-05,
|
| 1163 |
+
"loss": 0.5996,
|
| 1164 |
+
"step": 1880
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"epoch": 1.58,
|
| 1168 |
+
"eval_accuracy": 0.7220840734162226,
|
| 1169 |
+
"eval_f1_macro": 0.6869855176473937,
|
| 1170 |
+
"eval_f1_negative": 0.6864864864864865,
|
| 1171 |
+
"eval_f1_neutral": 0.7696068445711958,
|
| 1172 |
+
"eval_f1_positive": 0.6048632218844986,
|
| 1173 |
+
"eval_loss": 0.6290135979652405,
|
| 1174 |
+
"eval_precision_negative": 0.7019806540764625,
|
| 1175 |
+
"eval_precision_neutral": 0.7513922036595068,
|
| 1176 |
+
"eval_precision_positive": 0.6388443017656501,
|
| 1177 |
+
"eval_recall_negative": 0.6716615249008374,
|
| 1178 |
+
"eval_recall_neutral": 0.7887265135699374,
|
| 1179 |
+
"eval_recall_positive": 0.5743145743145743,
|
| 1180 |
+
"eval_runtime": 13.4252,
|
| 1181 |
+
"eval_samples_per_second": 629.039,
|
| 1182 |
+
"eval_steps_per_second": 3.277,
|
| 1183 |
+
"step": 1880
|
| 1184 |
+
},
|
| 1185 |
+
{
|
| 1186 |
+
"epoch": 1.62,
|
| 1187 |
+
"learning_rate": 1.6767676767676768e-05,
|
| 1188 |
+
"loss": 0.6136,
|
| 1189 |
+
"step": 1920
|
| 1190 |
+
},
|
| 1191 |
+
{
|
| 1192 |
+
"epoch": 1.62,
|
| 1193 |
+
"eval_accuracy": 0.7223208999407934,
|
| 1194 |
+
"eval_f1_macro": 0.6908988856510865,
|
| 1195 |
+
"eval_f1_negative": 0.7006585935840238,
|
| 1196 |
+
"eval_f1_neutral": 0.7644044755829762,
|
| 1197 |
+
"eval_f1_positive": 0.6076335877862595,
|
| 1198 |
+
"eval_loss": 0.6263943910598755,
|
| 1199 |
+
"eval_precision_negative": 0.6763740771123872,
|
| 1200 |
+
"eval_precision_neutral": 0.7657657657657657,
|
| 1201 |
+
"eval_precision_positive": 0.6450567260940032,
|
| 1202 |
+
"eval_recall_negative": 0.7267518730718379,
|
| 1203 |
+
"eval_recall_neutral": 0.7630480167014614,
|
| 1204 |
+
"eval_recall_positive": 0.5743145743145743,
|
| 1205 |
+
"eval_runtime": 13.5533,
|
| 1206 |
+
"eval_samples_per_second": 623.095,
|
| 1207 |
+
"eval_steps_per_second": 3.246,
|
| 1208 |
+
"step": 1920
|
| 1209 |
+
},
|
| 1210 |
+
{
|
| 1211 |
+
"epoch": 1.65,
|
| 1212 |
+
"learning_rate": 1.6700336700336702e-05,
|
| 1213 |
+
"loss": 0.6036,
|
| 1214 |
+
"step": 1960
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 1.65,
|
| 1218 |
+
"eval_accuracy": 0.7141503848431024,
|
| 1219 |
+
"eval_f1_macro": 0.6879732554886911,
|
| 1220 |
+
"eval_f1_negative": 0.6988408851422551,
|
| 1221 |
+
"eval_f1_neutral": 0.7530068728522336,
|
| 1222 |
+
"eval_f1_positive": 0.6120720084715848,
|
| 1223 |
+
"eval_loss": 0.6364882588386536,
|
| 1224 |
+
"eval_precision_negative": 0.6696284329563813,
|
| 1225 |
+
"eval_precision_neutral": 0.7753206545776206,
|
| 1226 |
+
"eval_precision_positive": 0.5991706979958535,
|
| 1227 |
+
"eval_recall_negative": 0.7307183781401498,
|
| 1228 |
+
"eval_recall_neutral": 0.7319415448851775,
|
| 1229 |
+
"eval_recall_positive": 0.6255411255411255,
|
| 1230 |
+
"eval_runtime": 13.4324,
|
| 1231 |
+
"eval_samples_per_second": 628.703,
|
| 1232 |
+
"eval_steps_per_second": 3.276,
|
| 1233 |
+
"step": 1960
|
| 1234 |
+
},
|
| 1235 |
+
{
|
| 1236 |
+
"epoch": 1.68,
|
| 1237 |
+
"learning_rate": 1.6632996632996633e-05,
|
| 1238 |
+
"loss": 0.6123,
|
| 1239 |
+
"step": 2000
|
| 1240 |
+
},
|
| 1241 |
+
{
|
| 1242 |
+
"epoch": 1.68,
|
| 1243 |
+
"eval_accuracy": 0.7198342214328005,
|
| 1244 |
+
"eval_f1_macro": 0.6811378179180303,
|
| 1245 |
+
"eval_f1_negative": 0.694331550802139,
|
| 1246 |
+
"eval_f1_neutral": 0.7660226924256361,
|
| 1247 |
+
"eval_f1_positive": 0.5830592105263157,
|
| 1248 |
+
"eval_loss": 0.6256130933761597,
|
| 1249 |
+
"eval_precision_negative": 0.6745635910224439,
|
| 1250 |
+
"eval_precision_neutral": 0.7504506308832365,
|
| 1251 |
+
"eval_precision_positive": 0.6778202676864244,
|
| 1252 |
+
"eval_recall_negative": 0.7152930806522697,
|
| 1253 |
+
"eval_recall_neutral": 0.7822546972860125,
|
| 1254 |
+
"eval_recall_positive": 0.5115440115440115,
|
| 1255 |
+
"eval_runtime": 13.4059,
|
| 1256 |
+
"eval_samples_per_second": 629.947,
|
| 1257 |
+
"eval_steps_per_second": 3.282,
|
| 1258 |
+
"step": 2000
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"epoch": 1.72,
|
| 1262 |
+
"learning_rate": 1.6565656565656567e-05,
|
| 1263 |
+
"loss": 0.5898,
|
| 1264 |
+
"step": 2040
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"epoch": 1.72,
|
| 1268 |
+
"eval_accuracy": 0.7255180580224985,
|
| 1269 |
+
"eval_f1_macro": 0.6783018381472949,
|
| 1270 |
+
"eval_f1_negative": 0.6911698789780368,
|
| 1271 |
+
"eval_f1_neutral": 0.777568611909244,
|
| 1272 |
+
"eval_f1_positive": 0.5661670235546039,
|
| 1273 |
+
"eval_loss": 0.6203241944313049,
|
| 1274 |
+
"eval_precision_negative": 0.7031463748290013,
|
| 1275 |
+
"eval_precision_neutral": 0.7399585140486518,
|
| 1276 |
+
"eval_precision_positive": 0.6965226554267651,
|
| 1277 |
+
"eval_recall_negative": 0.6795945350374615,
|
| 1278 |
+
"eval_recall_neutral": 0.8192066805845512,
|
| 1279 |
+
"eval_recall_positive": 0.4769119769119769,
|
| 1280 |
+
"eval_runtime": 13.429,
|
| 1281 |
+
"eval_samples_per_second": 628.864,
|
| 1282 |
+
"eval_steps_per_second": 3.276,
|
| 1283 |
+
"step": 2040
|
| 1284 |
+
},
|
| 1285 |
+
{
|
| 1286 |
+
"epoch": 1.75,
|
| 1287 |
+
"learning_rate": 1.64983164983165e-05,
|
| 1288 |
+
"loss": 0.585,
|
| 1289 |
+
"step": 2080
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"epoch": 1.75,
|
| 1293 |
+
"eval_accuracy": 0.7277679100059207,
|
| 1294 |
+
"eval_f1_macro": 0.6856054160655031,
|
| 1295 |
+
"eval_f1_negative": 0.6840989399293286,
|
| 1296 |
+
"eval_f1_neutral": 0.7798319327731092,
|
| 1297 |
+
"eval_f1_positive": 0.5928853754940712,
|
| 1298 |
+
"eval_loss": 0.62270188331604,
|
| 1299 |
+
"eval_precision_negative": 0.7348178137651822,
|
| 1300 |
+
"eval_precision_neutral": 0.7406572769953051,
|
| 1301 |
+
"eval_precision_positive": 0.6555944055944056,
|
| 1302 |
+
"eval_recall_negative": 0.6399294843543412,
|
| 1303 |
+
"eval_recall_neutral": 0.8233820459290188,
|
| 1304 |
+
"eval_recall_positive": 0.5411255411255411,
|
| 1305 |
+
"eval_runtime": 13.4336,
|
| 1306 |
+
"eval_samples_per_second": 628.646,
|
| 1307 |
+
"eval_steps_per_second": 3.275,
|
| 1308 |
+
"step": 2080
|
| 1309 |
+
},
|
| 1310 |
+
{
|
| 1311 |
+
"epoch": 1.78,
|
| 1312 |
+
"learning_rate": 1.6430976430976432e-05,
|
| 1313 |
+
"loss": 0.6092,
|
| 1314 |
+
"step": 2120
|
| 1315 |
+
},
|
| 1316 |
+
{
|
| 1317 |
+
"epoch": 1.78,
|
| 1318 |
+
"eval_accuracy": 0.7194789816459444,
|
| 1319 |
+
"eval_f1_macro": 0.691219851642046,
|
| 1320 |
+
"eval_f1_negative": 0.6904706700672385,
|
| 1321 |
+
"eval_f1_neutral": 0.7632427932146946,
|
| 1322 |
+
"eval_f1_positive": 0.6199460916442048,
|
| 1323 |
+
"eval_loss": 0.6305855512619019,
|
| 1324 |
+
"eval_precision_negative": 0.7284735812133072,
|
| 1325 |
+
"eval_precision_neutral": 0.7609462544096286,
|
| 1326 |
+
"eval_precision_positive": 0.5815423514538559,
|
| 1327 |
+
"eval_recall_negative": 0.6562362274129573,
|
| 1328 |
+
"eval_recall_neutral": 0.765553235908142,
|
| 1329 |
+
"eval_recall_positive": 0.6637806637806638,
|
| 1330 |
+
"eval_runtime": 13.4354,
|
| 1331 |
+
"eval_samples_per_second": 628.563,
|
| 1332 |
+
"eval_steps_per_second": 3.275,
|
| 1333 |
+
"step": 2120
|
| 1334 |
+
},
|
| 1335 |
+
{
|
| 1336 |
+
"epoch": 1.82,
|
| 1337 |
+
"learning_rate": 1.6363636363636366e-05,
|
| 1338 |
+
"loss": 0.5902,
|
| 1339 |
+
"step": 2160
|
| 1340 |
+
},
|
| 1341 |
+
{
|
| 1342 |
+
"epoch": 1.82,
|
| 1343 |
+
"eval_accuracy": 0.7217288336293665,
|
| 1344 |
+
"eval_f1_macro": 0.6903674148692266,
|
| 1345 |
+
"eval_f1_negative": 0.699802067297119,
|
| 1346 |
+
"eval_f1_neutral": 0.7650182196772514,
|
| 1347 |
+
"eval_f1_positive": 0.606281957633309,
|
| 1348 |
+
"eval_loss": 0.6268182992935181,
|
| 1349 |
+
"eval_precision_negative": 0.6984196663740123,
|
| 1350 |
+
"eval_precision_neutral": 0.7630321910695742,
|
| 1351 |
+
"eval_precision_positive": 0.613905325443787,
|
| 1352 |
+
"eval_recall_negative": 0.7011899515204936,
|
| 1353 |
+
"eval_recall_neutral": 0.7670146137787056,
|
| 1354 |
+
"eval_recall_positive": 0.5988455988455988,
|
| 1355 |
+
"eval_runtime": 13.453,
|
| 1356 |
+
"eval_samples_per_second": 627.739,
|
| 1357 |
+
"eval_steps_per_second": 3.271,
|
| 1358 |
+
"step": 2160
|
| 1359 |
+
},
|
| 1360 |
+
{
|
| 1361 |
+
"epoch": 1.85,
|
| 1362 |
+
"learning_rate": 1.6296296296296297e-05,
|
| 1363 |
+
"loss": 0.6077,
|
| 1364 |
+
"step": 2200
|
| 1365 |
+
},
|
| 1366 |
+
{
|
| 1367 |
+
"epoch": 1.85,
|
| 1368 |
+
"eval_accuracy": 0.7239786856127887,
|
| 1369 |
+
"eval_f1_macro": 0.6863881436307776,
|
| 1370 |
+
"eval_f1_negative": 0.6860517946388005,
|
| 1371 |
+
"eval_f1_neutral": 0.7731126362535325,
|
| 1372 |
+
"eval_f1_positive": 0.5999999999999999,
|
| 1373 |
+
"eval_loss": 0.6207541227340698,
|
| 1374 |
+
"eval_precision_negative": 0.7079231129864041,
|
| 1375 |
+
"eval_precision_neutral": 0.7483391949980461,
|
| 1376 |
+
"eval_precision_positive": 0.6482412060301508,
|
| 1377 |
+
"eval_recall_negative": 0.6654914059056853,
|
| 1378 |
+
"eval_recall_neutral": 0.7995824634655533,
|
| 1379 |
+
"eval_recall_positive": 0.5584415584415584,
|
| 1380 |
+
"eval_runtime": 13.4434,
|
| 1381 |
+
"eval_samples_per_second": 628.189,
|
| 1382 |
+
"eval_steps_per_second": 3.273,
|
| 1383 |
+
"step": 2200
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 1.89,
|
| 1387 |
+
"learning_rate": 1.622895622895623e-05,
|
| 1388 |
+
"loss": 0.5971,
|
| 1389 |
+
"step": 2240
|
| 1390 |
+
},
|
| 1391 |
+
{
|
| 1392 |
+
"epoch": 1.89,
|
| 1393 |
+
"eval_accuracy": 0.7297809354647721,
|
| 1394 |
+
"eval_f1_macro": 0.6928803015060957,
|
| 1395 |
+
"eval_f1_negative": 0.6937485894831867,
|
| 1396 |
+
"eval_f1_neutral": 0.7777216213484281,
|
| 1397 |
+
"eval_f1_positive": 0.6071706936866719,
|
| 1398 |
+
"eval_loss": 0.6181867122650146,
|
| 1399 |
+
"eval_precision_negative": 0.7109158186864015,
|
| 1400 |
+
"eval_precision_neutral": 0.7538702723887909,
|
| 1401 |
+
"eval_precision_positive": 0.6601694915254237,
|
| 1402 |
+
"eval_recall_negative": 0.6773909211106214,
|
| 1403 |
+
"eval_recall_neutral": 0.8031315240083508,
|
| 1404 |
+
"eval_recall_positive": 0.562049062049062,
|
| 1405 |
+
"eval_runtime": 13.5165,
|
| 1406 |
+
"eval_samples_per_second": 624.792,
|
| 1407 |
+
"eval_steps_per_second": 3.255,
|
| 1408 |
+
"step": 2240
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"epoch": 1.92,
|
| 1412 |
+
"learning_rate": 1.616161616161616e-05,
|
| 1413 |
+
"loss": 0.6053,
|
| 1414 |
+
"step": 2280
|
| 1415 |
+
},
|
| 1416 |
+
{
|
| 1417 |
+
"epoch": 1.92,
|
| 1418 |
+
"eval_accuracy": 0.7235050325636472,
|
| 1419 |
+
"eval_f1_macro": 0.6869805131510361,
|
| 1420 |
+
"eval_f1_negative": 0.6891179839633448,
|
| 1421 |
+
"eval_f1_neutral": 0.7715961966417155,
|
| 1422 |
+
"eval_f1_positive": 0.6002273588480483,
|
| 1423 |
+
"eval_loss": 0.6221389174461365,
|
| 1424 |
+
"eval_precision_negative": 0.7175572519083969,
|
| 1425 |
+
"eval_precision_neutral": 0.7484301412872841,
|
| 1426 |
+
"eval_precision_positive": 0.6320830007980845,
|
| 1427 |
+
"eval_recall_negative": 0.6628470691934774,
|
| 1428 |
+
"eval_recall_neutral": 0.7962421711899791,
|
| 1429 |
+
"eval_recall_positive": 0.5714285714285714,
|
| 1430 |
+
"eval_runtime": 13.4474,
|
| 1431 |
+
"eval_samples_per_second": 628.005,
|
| 1432 |
+
"eval_steps_per_second": 3.272,
|
| 1433 |
+
"step": 2280
|
| 1434 |
+
},
|
| 1435 |
+
{
|
| 1436 |
+
"epoch": 1.95,
|
| 1437 |
+
"learning_rate": 1.6094276094276096e-05,
|
| 1438 |
+
"loss": 0.5959,
|
| 1439 |
+
"step": 2320
|
| 1440 |
+
},
|
| 1441 |
+
{
|
| 1442 |
+
"epoch": 1.95,
|
| 1443 |
+
"eval_accuracy": 0.7199526346950859,
|
| 1444 |
+
"eval_f1_macro": 0.682651116707718,
|
| 1445 |
+
"eval_f1_negative": 0.6665057915057916,
|
| 1446 |
+
"eval_f1_neutral": 0.7722100219604712,
|
| 1447 |
+
"eval_f1_positive": 0.6092375366568915,
|
| 1448 |
+
"eval_loss": 0.6313912272453308,
|
| 1449 |
+
"eval_precision_negative": 0.7365333333333334,
|
| 1450 |
+
"eval_precision_neutral": 0.7398622800306044,
|
| 1451 |
+
"eval_precision_positive": 0.6192250372578242,
|
| 1452 |
+
"eval_recall_negative": 0.6086381665932129,
|
| 1453 |
+
"eval_recall_neutral": 0.8075156576200417,
|
| 1454 |
+
"eval_recall_positive": 0.5995670995670995,
|
| 1455 |
+
"eval_runtime": 13.532,
|
| 1456 |
+
"eval_samples_per_second": 624.078,
|
| 1457 |
+
"eval_steps_per_second": 3.252,
|
| 1458 |
+
"step": 2320
|
| 1459 |
+
},
|
| 1460 |
+
{
|
| 1461 |
+
"epoch": 1.99,
|
| 1462 |
+
"learning_rate": 1.602693602693603e-05,
|
| 1463 |
+
"loss": 0.6196,
|
| 1464 |
+
"step": 2360
|
| 1465 |
+
},
|
| 1466 |
+
{
|
| 1467 |
+
"epoch": 1.99,
|
| 1468 |
+
"eval_accuracy": 0.709295441089402,
|
| 1469 |
+
"eval_f1_macro": 0.6881296895840747,
|
| 1470 |
+
"eval_f1_negative": 0.700795870079587,
|
| 1471 |
+
"eval_f1_neutral": 0.7440385036097135,
|
| 1472 |
+
"eval_f1_positive": 0.6195546950629235,
|
| 1473 |
+
"eval_loss": 0.6360026001930237,
|
| 1474 |
+
"eval_precision_negative": 0.684453781512605,
|
| 1475 |
+
"eval_precision_neutral": 0.7814797794117647,
|
| 1476 |
+
"eval_precision_positive": 0.5604203152364273,
|
| 1477 |
+
"eval_recall_negative": 0.7179374173644777,
|
| 1478 |
+
"eval_recall_neutral": 0.7100208768267223,
|
| 1479 |
+
"eval_recall_positive": 0.6926406926406926,
|
| 1480 |
+
"eval_runtime": 13.7523,
|
| 1481 |
+
"eval_samples_per_second": 614.079,
|
| 1482 |
+
"eval_steps_per_second": 3.199,
|
| 1483 |
+
"step": 2360
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.02,
|
| 1487 |
+
"learning_rate": 1.595959595959596e-05,
|
| 1488 |
+
"loss": 0.5686,
|
| 1489 |
+
"step": 2400
|
| 1490 |
+
},
|
| 1491 |
+
{
|
| 1492 |
+
"epoch": 2.02,
|
| 1493 |
+
"eval_accuracy": 0.7191237418590882,
|
| 1494 |
+
"eval_f1_macro": 0.6870370031512542,
|
| 1495 |
+
"eval_f1_negative": 0.6933101650738488,
|
| 1496 |
+
"eval_f1_neutral": 0.7630622208372286,
|
| 1497 |
+
"eval_f1_positive": 0.6047386235426853,
|
| 1498 |
+
"eval_loss": 0.6405708193778992,
|
| 1499 |
+
"eval_precision_negative": 0.6835117773019272,
|
| 1500 |
+
"eval_precision_neutral": 0.7593549720901385,
|
| 1501 |
+
"eval_precision_positive": 0.631578947368421,
|
| 1502 |
+
"eval_recall_negative": 0.7033935654473337,
|
| 1503 |
+
"eval_recall_neutral": 0.7668058455114822,
|
| 1504 |
+
"eval_recall_positive": 0.5800865800865801,
|
| 1505 |
+
"eval_runtime": 13.5092,
|
| 1506 |
+
"eval_samples_per_second": 625.128,
|
| 1507 |
+
"eval_steps_per_second": 3.257,
|
| 1508 |
+
"step": 2400
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 2.05,
|
| 1512 |
+
"learning_rate": 1.5892255892255895e-05,
|
| 1513 |
+
"loss": 0.5586,
|
| 1514 |
+
"step": 2440
|
| 1515 |
+
},
|
| 1516 |
+
{
|
| 1517 |
+
"epoch": 2.05,
|
| 1518 |
+
"eval_accuracy": 0.7223208999407934,
|
| 1519 |
+
"eval_f1_macro": 0.6862310990226811,
|
| 1520 |
+
"eval_f1_negative": 0.6942959001782532,
|
| 1521 |
+
"eval_f1_neutral": 0.7689950980392158,
|
| 1522 |
+
"eval_f1_positive": 0.5954022988505746,
|
| 1523 |
+
"eval_loss": 0.6309220790863037,
|
| 1524 |
+
"eval_precision_negative": 0.7021180712032447,
|
| 1525 |
+
"eval_precision_neutral": 0.7526989204318273,
|
| 1526 |
+
"eval_precision_positive": 0.6348039215686274,
|
| 1527 |
+
"eval_recall_negative": 0.6866460996033495,
|
| 1528 |
+
"eval_recall_neutral": 0.7860125260960334,
|
| 1529 |
+
"eval_recall_positive": 0.5606060606060606,
|
| 1530 |
+
"eval_runtime": 13.5411,
|
| 1531 |
+
"eval_samples_per_second": 623.656,
|
| 1532 |
+
"eval_steps_per_second": 3.249,
|
| 1533 |
+
"step": 2440
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 2.09,
|
| 1537 |
+
"learning_rate": 1.5824915824915825e-05,
|
| 1538 |
+
"loss": 0.5671,
|
| 1539 |
+
"step": 2480
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.09,
|
| 1543 |
+
"eval_accuracy": 0.7205447010065127,
|
| 1544 |
+
"eval_f1_macro": 0.683056392123643,
|
| 1545 |
+
"eval_f1_negative": 0.6914449087310313,
|
| 1546 |
+
"eval_f1_neutral": 0.7678189048638727,
|
| 1547 |
+
"eval_f1_positive": 0.5899053627760252,
|
| 1548 |
+
"eval_loss": 0.6297855377197266,
|
| 1549 |
+
"eval_precision_negative": 0.6900790166812993,
|
| 1550 |
+
"eval_precision_neutral": 0.7504484751843731,
|
| 1551 |
+
"eval_precision_positive": 0.6504347826086957,
|
| 1552 |
+
"eval_recall_negative": 0.6928162185985015,
|
| 1553 |
+
"eval_recall_neutral": 0.7860125260960334,
|
| 1554 |
+
"eval_recall_positive": 0.5396825396825397,
|
| 1555 |
+
"eval_runtime": 13.5165,
|
| 1556 |
+
"eval_samples_per_second": 624.793,
|
| 1557 |
+
"eval_steps_per_second": 3.255,
|
| 1558 |
+
"step": 2480
|
| 1559 |
+
},
|
| 1560 |
+
{
|
| 1561 |
+
"epoch": 2.12,
|
| 1562 |
+
"learning_rate": 1.575757575757576e-05,
|
| 1563 |
+
"loss": 0.5458,
|
| 1564 |
+
"step": 2520
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"epoch": 2.12,
|
| 1568 |
+
"eval_accuracy": 0.7214920071047958,
|
| 1569 |
+
"eval_f1_macro": 0.6872366815721698,
|
| 1570 |
+
"eval_f1_negative": 0.6905083220872694,
|
| 1571 |
+
"eval_f1_neutral": 0.7679083094555873,
|
| 1572 |
+
"eval_f1_positive": 0.6032934131736527,
|
| 1573 |
+
"eval_loss": 0.63502436876297,
|
| 1574 |
+
"eval_precision_negative": 0.7050987597611392,
|
| 1575 |
+
"eval_precision_neutral": 0.7531112003211562,
|
| 1576 |
+
"eval_precision_positive": 0.6267496111975117,
|
| 1577 |
+
"eval_recall_negative": 0.6765094755398854,
|
| 1578 |
+
"eval_recall_neutral": 0.7832985386221294,
|
| 1579 |
+
"eval_recall_positive": 0.5815295815295816,
|
| 1580 |
+
"eval_runtime": 13.5006,
|
| 1581 |
+
"eval_samples_per_second": 625.526,
|
| 1582 |
+
"eval_steps_per_second": 3.259,
|
| 1583 |
+
"step": 2520
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 2.15,
|
| 1587 |
+
"learning_rate": 1.569023569023569e-05,
|
| 1588 |
+
"loss": 0.5617,
|
| 1589 |
+
"step": 2560
|
| 1590 |
+
},
|
| 1591 |
+
{
|
| 1592 |
+
"epoch": 2.15,
|
| 1593 |
+
"eval_accuracy": 0.7103611604499704,
|
| 1594 |
+
"eval_f1_macro": 0.6868863746975769,
|
| 1595 |
+
"eval_f1_negative": 0.7016379846568526,
|
| 1596 |
+
"eval_f1_neutral": 0.7452902101709682,
|
| 1597 |
+
"eval_f1_positive": 0.6137309292649098,
|
| 1598 |
+
"eval_loss": 0.6412296891212463,
|
| 1599 |
+
"eval_precision_negative": 0.6624902114330462,
|
| 1600 |
+
"eval_precision_neutral": 0.7789665376735716,
|
| 1601 |
+
"eval_precision_positive": 0.5907877169559412,
|
| 1602 |
+
"eval_recall_negative": 0.745702952842662,
|
| 1603 |
+
"eval_recall_neutral": 0.7144050104384133,
|
| 1604 |
+
"eval_recall_positive": 0.6385281385281385,
|
| 1605 |
+
"eval_runtime": 13.561,
|
| 1606 |
+
"eval_samples_per_second": 622.74,
|
| 1607 |
+
"eval_steps_per_second": 3.245,
|
| 1608 |
+
"step": 2560
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"epoch": 2.19,
|
| 1612 |
+
"learning_rate": 1.5622895622895624e-05,
|
| 1613 |
+
"loss": 0.5582,
|
| 1614 |
+
"step": 2600
|
| 1615 |
+
},
|
| 1616 |
+
{
|
| 1617 |
+
"epoch": 2.19,
|
| 1618 |
+
"eval_accuracy": 0.7085849615156897,
|
| 1619 |
+
"eval_f1_macro": 0.687474213023077,
|
| 1620 |
+
"eval_f1_negative": 0.699581589958159,
|
| 1621 |
+
"eval_f1_neutral": 0.7421052631578948,
|
| 1622 |
+
"eval_f1_positive": 0.6207357859531774,
|
| 1623 |
+
"eval_loss": 0.6525644659996033,
|
| 1624 |
+
"eval_precision_negative": 0.6658701712465154,
|
| 1625 |
+
"eval_precision_neutral": 0.7815242494226328,
|
| 1626 |
+
"eval_precision_positive": 0.5785536159600998,
|
| 1627 |
+
"eval_recall_negative": 0.7368884971353019,
|
| 1628 |
+
"eval_recall_neutral": 0.7064718162839249,
|
| 1629 |
+
"eval_recall_positive": 0.6695526695526696,
|
| 1630 |
+
"eval_runtime": 13.5361,
|
| 1631 |
+
"eval_samples_per_second": 623.888,
|
| 1632 |
+
"eval_steps_per_second": 3.251,
|
| 1633 |
+
"step": 2600
|
| 1634 |
+
},
|
| 1635 |
+
{
|
| 1636 |
+
"epoch": 2.22,
|
| 1637 |
+
"learning_rate": 1.555555555555556e-05,
|
| 1638 |
+
"loss": 0.5543,
|
| 1639 |
+
"step": 2640
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.22,
|
| 1643 |
+
"eval_accuracy": 0.7201894612196567,
|
| 1644 |
+
"eval_f1_macro": 0.6869432555451471,
|
| 1645 |
+
"eval_f1_negative": 0.7009306260575296,
|
| 1646 |
+
"eval_f1_neutral": 0.762918885061071,
|
| 1647 |
+
"eval_f1_positive": 0.5969802555168409,
|
| 1648 |
+
"eval_loss": 0.640322744846344,
|
| 1649 |
+
"eval_precision_negative": 0.6738511590077267,
|
| 1650 |
+
"eval_precision_neutral": 0.7629985383169764,
|
| 1651 |
+
"eval_precision_positive": 0.6441102756892231,
|
| 1652 |
+
"eval_recall_negative": 0.7302776553547818,
|
| 1653 |
+
"eval_recall_neutral": 0.762839248434238,
|
| 1654 |
+
"eval_recall_positive": 0.5562770562770563,
|
| 1655 |
+
"eval_runtime": 13.5133,
|
| 1656 |
+
"eval_samples_per_second": 624.941,
|
| 1657 |
+
"eval_steps_per_second": 3.256,
|
| 1658 |
+
"step": 2640
|
| 1659 |
+
},
|
| 1660 |
+
{
|
| 1661 |
+
"epoch": 2.26,
|
| 1662 |
+
"learning_rate": 1.548821548821549e-05,
|
| 1663 |
+
"loss": 0.558,
|
| 1664 |
+
"step": 2680
|
| 1665 |
+
},
|
| 1666 |
+
{
|
| 1667 |
+
"epoch": 2.26,
|
| 1668 |
+
"eval_accuracy": 0.7206631142687981,
|
| 1669 |
+
"eval_f1_macro": 0.6874230644596926,
|
| 1670 |
+
"eval_f1_negative": 0.697239536954586,
|
| 1671 |
+
"eval_f1_neutral": 0.7653250773993808,
|
| 1672 |
+
"eval_f1_positive": 0.5997045790251108,
|
| 1673 |
+
"eval_loss": 0.6337417364120483,
|
| 1674 |
+
"eval_precision_negative": 0.7044534412955465,
|
| 1675 |
+
"eval_precision_neutral": 0.756734693877551,
|
| 1676 |
+
"eval_precision_positive": 0.6142208774583964,
|
| 1677 |
+
"eval_recall_negative": 0.6901718818862935,
|
| 1678 |
+
"eval_recall_neutral": 0.7741127348643007,
|
| 1679 |
+
"eval_recall_positive": 0.5858585858585859,
|
| 1680 |
+
"eval_runtime": 13.5113,
|
| 1681 |
+
"eval_samples_per_second": 625.032,
|
| 1682 |
+
"eval_steps_per_second": 3.257,
|
| 1683 |
+
"step": 2680
|
| 1684 |
+
},
|
| 1685 |
+
{
|
| 1686 |
+
"epoch": 2.29,
|
| 1687 |
+
"learning_rate": 1.5420875420875423e-05,
|
| 1688 |
+
"loss": 0.5506,
|
| 1689 |
+
"step": 2720
|
| 1690 |
+
},
|
| 1691 |
+
{
|
| 1692 |
+
"epoch": 2.29,
|
| 1693 |
+
"eval_accuracy": 0.7171107164002368,
|
| 1694 |
+
"eval_f1_macro": 0.6895025450757618,
|
| 1695 |
+
"eval_f1_negative": 0.7093775262732418,
|
| 1696 |
+
"eval_f1_neutral": 0.753454231433506,
|
| 1697 |
+
"eval_f1_positive": 0.6056758775205378,
|
| 1698 |
+
"eval_loss": 0.6484044194221497,
|
| 1699 |
+
"eval_precision_negative": 0.6550951847704367,
|
| 1700 |
+
"eval_precision_neutral": 0.7800625838176128,
|
| 1701 |
+
"eval_precision_positive": 0.6277089783281734,
|
| 1702 |
+
"eval_recall_negative": 0.7734684883208461,
|
| 1703 |
+
"eval_recall_neutral": 0.7286012526096033,
|
| 1704 |
+
"eval_recall_positive": 0.5851370851370852,
|
| 1705 |
+
"eval_runtime": 13.5135,
|
| 1706 |
+
"eval_samples_per_second": 624.931,
|
| 1707 |
+
"eval_steps_per_second": 3.256,
|
| 1708 |
+
"step": 2720
|
| 1709 |
+
},
|
| 1710 |
+
{
|
| 1711 |
+
"epoch": 2.32,
|
| 1712 |
+
"learning_rate": 1.5353535353535354e-05,
|
| 1713 |
+
"loss": 0.5569,
|
| 1714 |
+
"step": 2760
|
| 1715 |
+
},
|
| 1716 |
+
{
|
| 1717 |
+
"epoch": 2.32,
|
| 1718 |
+
"eval_accuracy": 0.7167554766133807,
|
| 1719 |
+
"eval_f1_macro": 0.6895340612929335,
|
| 1720 |
+
"eval_f1_negative": 0.7063525452250736,
|
| 1721 |
+
"eval_f1_neutral": 0.7544910179640718,
|
| 1722 |
+
"eval_f1_positive": 0.6077586206896551,
|
| 1723 |
+
"eval_loss": 0.6315791606903076,
|
| 1724 |
+
"eval_precision_negative": 0.6756539235412475,
|
| 1725 |
+
"eval_precision_neutral": 0.7733450241122315,
|
| 1726 |
+
"eval_precision_positive": 0.6051502145922747,
|
| 1727 |
+
"eval_recall_negative": 0.7399735566328779,
|
| 1728 |
+
"eval_recall_neutral": 0.7365344467640919,
|
| 1729 |
+
"eval_recall_positive": 0.6103896103896104,
|
| 1730 |
+
"eval_runtime": 13.5058,
|
| 1731 |
+
"eval_samples_per_second": 625.287,
|
| 1732 |
+
"eval_steps_per_second": 3.258,
|
| 1733 |
+
"step": 2760
|
| 1734 |
+
},
|
| 1735 |
+
{
|
| 1736 |
+
"epoch": 2.36,
|
| 1737 |
+
"learning_rate": 1.5286195286195288e-05,
|
| 1738 |
+
"loss": 0.5469,
|
| 1739 |
+
"step": 2800
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 2.36,
|
| 1743 |
+
"eval_accuracy": 0.7182948490230906,
|
| 1744 |
+
"eval_f1_macro": 0.69129222470535,
|
| 1745 |
+
"eval_f1_negative": 0.7031543764361813,
|
| 1746 |
+
"eval_f1_neutral": 0.7562953478446436,
|
| 1747 |
+
"eval_f1_positive": 0.6144269498352252,
|
| 1748 |
+
"eval_loss": 0.6461361646652222,
|
| 1749 |
+
"eval_precision_negative": 0.6683876092136616,
|
| 1750 |
+
"eval_precision_neutral": 0.7734613705805325,
|
| 1751 |
+
"eval_precision_positive": 0.6237918215613383,
|
| 1752 |
+
"eval_recall_negative": 0.7417364477743499,
|
| 1753 |
+
"eval_recall_neutral": 0.7398747390396659,
|
| 1754 |
+
"eval_recall_positive": 0.6053391053391053,
|
| 1755 |
+
"eval_runtime": 13.5002,
|
| 1756 |
+
"eval_samples_per_second": 625.546,
|
| 1757 |
+
"eval_steps_per_second": 3.259,
|
| 1758 |
+
"step": 2800
|
| 1759 |
+
},
|
| 1760 |
+
{
|
| 1761 |
+
"epoch": 2.39,
|
| 1762 |
+
"learning_rate": 1.521885521885522e-05,
|
| 1763 |
+
"loss": 0.5497,
|
| 1764 |
+
"step": 2840
|
| 1765 |
+
},
|
| 1766 |
+
{
|
| 1767 |
+
"epoch": 2.39,
|
| 1768 |
+
"eval_accuracy": 0.7205447010065127,
|
| 1769 |
+
"eval_f1_macro": 0.6901740903513232,
|
| 1770 |
+
"eval_f1_negative": 0.6989695242271432,
|
| 1771 |
+
"eval_f1_neutral": 0.7628887497390942,
|
| 1772 |
+
"eval_f1_positive": 0.608663997087732,
|
| 1773 |
+
"eval_loss": 0.6306228041648865,
|
| 1774 |
+
"eval_precision_negative": 0.6954624781849913,
|
| 1775 |
+
"eval_precision_neutral": 0.7627295492487479,
|
| 1776 |
+
"eval_precision_positive": 0.6142542248346804,
|
| 1777 |
+
"eval_recall_negative": 0.7025121198765977,
|
| 1778 |
+
"eval_recall_neutral": 0.7630480167014614,
|
| 1779 |
+
"eval_recall_positive": 0.6031746031746031,
|
| 1780 |
+
"eval_runtime": 13.5161,
|
| 1781 |
+
"eval_samples_per_second": 624.81,
|
| 1782 |
+
"eval_steps_per_second": 3.255,
|
| 1783 |
+
"step": 2840
|
| 1784 |
+
},
|
| 1785 |
+
{
|
| 1786 |
+
"epoch": 2.42,
|
| 1787 |
+
"learning_rate": 1.5151515151515153e-05,
|
| 1788 |
+
"loss": 0.5577,
|
| 1789 |
+
"step": 2880
|
| 1790 |
+
},
|
| 1791 |
+
{
|
| 1792 |
+
"epoch": 2.42,
|
| 1793 |
+
"eval_accuracy": 0.7149792776791001,
|
| 1794 |
+
"eval_f1_macro": 0.686526091020862,
|
| 1795 |
+
"eval_f1_negative": 0.6968478950708694,
|
| 1796 |
+
"eval_f1_neutral": 0.7555225148683093,
|
| 1797 |
+
"eval_f1_positive": 0.6072078631234074,
|
| 1798 |
+
"eval_loss": 0.6409549713134766,
|
| 1799 |
+
"eval_precision_negative": 0.6700569568755086,
|
| 1800 |
+
"eval_precision_neutral": 0.7689148292261133,
|
| 1801 |
+
"eval_precision_positive": 0.6127847171197649,
|
| 1802 |
+
"eval_recall_negative": 0.7258704275011018,
|
| 1803 |
+
"eval_recall_neutral": 0.7425887265135699,
|
| 1804 |
+
"eval_recall_positive": 0.6017316017316018,
|
| 1805 |
+
"eval_runtime": 13.6855,
|
| 1806 |
+
"eval_samples_per_second": 617.076,
|
| 1807 |
+
"eval_steps_per_second": 3.215,
|
| 1808 |
+
"step": 2880
|
| 1809 |
+
},
|
| 1810 |
+
{
|
| 1811 |
+
"epoch": 2.46,
|
| 1812 |
+
"learning_rate": 1.5084175084175085e-05,
|
| 1813 |
+
"loss": 0.557,
|
| 1814 |
+
"step": 2920
|
| 1815 |
+
},
|
| 1816 |
+
{
|
| 1817 |
+
"epoch": 2.46,
|
| 1818 |
+
"eval_accuracy": 0.7225577264653641,
|
| 1819 |
+
"eval_f1_macro": 0.6876490677339296,
|
| 1820 |
+
"eval_f1_negative": 0.7071207430340557,
|
| 1821 |
+
"eval_f1_neutral": 0.7645151166440004,
|
| 1822 |
+
"eval_f1_positive": 0.5913113435237329,
|
| 1823 |
+
"eval_loss": 0.6443095803260803,
|
| 1824 |
+
"eval_precision_negative": 0.6649844720496895,
|
| 1825 |
+
"eval_precision_neutral": 0.7661983644369889,
|
| 1826 |
+
"eval_precision_positive": 0.6681818181818182,
|
| 1827 |
+
"eval_recall_negative": 0.75495813133539,
|
| 1828 |
+
"eval_recall_neutral": 0.762839248434238,
|
| 1829 |
+
"eval_recall_positive": 0.5303030303030303,
|
| 1830 |
+
"eval_runtime": 13.5647,
|
| 1831 |
+
"eval_samples_per_second": 622.57,
|
| 1832 |
+
"eval_steps_per_second": 3.244,
|
| 1833 |
+
"step": 2920
|
| 1834 |
+
},
|
| 1835 |
+
{
|
| 1836 |
+
"epoch": 2.49,
|
| 1837 |
+
"learning_rate": 1.5016835016835018e-05,
|
| 1838 |
+
"loss": 0.5542,
|
| 1839 |
+
"step": 2960
|
| 1840 |
+
},
|
| 1841 |
+
{
|
| 1842 |
+
"epoch": 2.49,
|
| 1843 |
+
"eval_accuracy": 0.7161634103019539,
|
| 1844 |
+
"eval_f1_macro": 0.687092758801656,
|
| 1845 |
+
"eval_f1_negative": 0.7009579341940858,
|
| 1846 |
+
"eval_f1_neutral": 0.7560144773259526,
|
| 1847 |
+
"eval_f1_positive": 0.6043058648849294,
|
| 1848 |
+
"eval_loss": 0.6480274796485901,
|
| 1849 |
+
"eval_precision_negative": 0.6644295302013423,
|
| 1850 |
+
"eval_precision_neutral": 0.7712858384013901,
|
| 1851 |
+
"eval_precision_positive": 0.6223241590214067,
|
| 1852 |
+
"eval_recall_negative": 0.7417364477743499,
|
| 1853 |
+
"eval_recall_neutral": 0.7413361169102296,
|
| 1854 |
+
"eval_recall_positive": 0.5873015873015873,
|
| 1855 |
+
"eval_runtime": 13.5134,
|
| 1856 |
+
"eval_samples_per_second": 624.935,
|
| 1857 |
+
"eval_steps_per_second": 3.256,
|
| 1858 |
+
"step": 2960
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"epoch": 2.53,
|
| 1862 |
+
"learning_rate": 1.4949494949494952e-05,
|
| 1863 |
+
"loss": 0.567,
|
| 1864 |
+
"step": 3000
|
| 1865 |
+
},
|
| 1866 |
+
{
|
| 1867 |
+
"epoch": 2.53,
|
| 1868 |
+
"eval_accuracy": 0.7079928952042629,
|
| 1869 |
+
"eval_f1_macro": 0.6865435263558487,
|
| 1870 |
+
"eval_f1_negative": 0.7025898078529658,
|
| 1871 |
+
"eval_f1_neutral": 0.7416051965209732,
|
| 1872 |
+
"eval_f1_positive": 0.6154355746936071,
|
| 1873 |
+
"eval_loss": 0.6429863572120667,
|
| 1874 |
+
"eval_precision_negative": 0.6677252878126241,
|
| 1875 |
+
"eval_precision_neutral": 0.7845329606335896,
|
| 1876 |
+
"eval_precision_positive": 0.5688916105327618,
|
| 1877 |
+
"eval_recall_negative": 0.7412957249889819,
|
| 1878 |
+
"eval_recall_neutral": 0.7031315240083508,
|
| 1879 |
+
"eval_recall_positive": 0.6702741702741702,
|
| 1880 |
+
"eval_runtime": 13.5159,
|
| 1881 |
+
"eval_samples_per_second": 624.818,
|
| 1882 |
+
"eval_steps_per_second": 3.255,
|
| 1883 |
+
"step": 3000
|
| 1884 |
+
},
|
| 1885 |
+
{
|
| 1886 |
+
"epoch": 2.56,
|
| 1887 |
+
"learning_rate": 1.4882154882154884e-05,
|
| 1888 |
+
"loss": 0.5606,
|
| 1889 |
+
"step": 3040
|
| 1890 |
+
},
|
| 1891 |
+
{
|
| 1892 |
+
"epoch": 2.56,
|
| 1893 |
+
"eval_accuracy": 0.7155713439905269,
|
| 1894 |
+
"eval_f1_macro": 0.6890929890491521,
|
| 1895 |
+
"eval_f1_negative": 0.7017241379310344,
|
| 1896 |
+
"eval_f1_neutral": 0.7549994652978291,
|
| 1897 |
+
"eval_f1_positive": 0.6105553639185926,
|
| 1898 |
+
"eval_loss": 0.6457778811454773,
|
| 1899 |
+
"eval_precision_negative": 0.6866301138760017,
|
| 1900 |
+
"eval_precision_neutral": 0.7739530804648104,
|
| 1901 |
+
"eval_precision_positive": 0.5849306014540647,
|
| 1902 |
+
"eval_recall_negative": 0.7174966945791097,
|
| 1903 |
+
"eval_recall_neutral": 0.7369519832985386,
|
| 1904 |
+
"eval_recall_positive": 0.6385281385281385,
|
| 1905 |
+
"eval_runtime": 13.5186,
|
| 1906 |
+
"eval_samples_per_second": 624.694,
|
| 1907 |
+
"eval_steps_per_second": 3.255,
|
| 1908 |
+
"step": 3040
|
| 1909 |
+
},
|
| 1910 |
+
{
|
| 1911 |
+
"epoch": 2.59,
|
| 1912 |
+
"learning_rate": 1.4814814814814815e-05,
|
| 1913 |
+
"loss": 0.5492,
|
| 1914 |
+
"step": 3080
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"epoch": 2.59,
|
| 1918 |
+
"eval_accuracy": 0.713676731793961,
|
| 1919 |
+
"eval_f1_macro": 0.687090376547261,
|
| 1920 |
+
"eval_f1_negative": 0.6999178981937603,
|
| 1921 |
+
"eval_f1_neutral": 0.7514812021975654,
|
| 1922 |
+
"eval_f1_positive": 0.6098720292504571,
|
| 1923 |
+
"eval_loss": 0.646338164806366,
|
| 1924 |
+
"eval_precision_negative": 0.6550134460238187,
|
| 1925 |
+
"eval_precision_neutral": 0.7763187180057868,
|
| 1926 |
+
"eval_precision_positive": 0.6182357301704967,
|
| 1927 |
+
"eval_recall_negative": 0.751432349052446,
|
| 1928 |
+
"eval_recall_neutral": 0.7281837160751565,
|
| 1929 |
+
"eval_recall_positive": 0.6017316017316018,
|
| 1930 |
+
"eval_runtime": 13.5425,
|
| 1931 |
+
"eval_samples_per_second": 623.591,
|
| 1932 |
+
"eval_steps_per_second": 3.249,
|
| 1933 |
+
"step": 3080
|
| 1934 |
+
},
|
| 1935 |
+
{
|
| 1936 |
+
"epoch": 2.63,
|
| 1937 |
+
"learning_rate": 1.4747474747474747e-05,
|
| 1938 |
+
"loss": 0.5509,
|
| 1939 |
+
"step": 3120
|
| 1940 |
+
},
|
| 1941 |
+
{
|
| 1942 |
+
"epoch": 2.63,
|
| 1943 |
+
"eval_accuracy": 0.7187685020722321,
|
| 1944 |
+
"eval_f1_macro": 0.6907132499480086,
|
| 1945 |
+
"eval_f1_negative": 0.7029375258585022,
|
| 1946 |
+
"eval_f1_neutral": 0.7576566001493972,
|
| 1947 |
+
"eval_f1_positive": 0.6115456238361267,
|
| 1948 |
+
"eval_loss": 0.6359885334968567,
|
| 1949 |
+
"eval_precision_negative": 0.6623781676413255,
|
| 1950 |
+
"eval_precision_neutral": 0.7749399694389871,
|
| 1951 |
+
"eval_precision_positive": 0.6320246343341032,
|
| 1952 |
+
"eval_recall_negative": 0.748788012340238,
|
| 1953 |
+
"eval_recall_neutral": 0.7411273486430062,
|
| 1954 |
+
"eval_recall_positive": 0.5923520923520924,
|
| 1955 |
+
"eval_runtime": 13.4528,
|
| 1956 |
+
"eval_samples_per_second": 627.753,
|
| 1957 |
+
"eval_steps_per_second": 3.271,
|
| 1958 |
+
"step": 3120
|
| 1959 |
+
},
|
| 1960 |
+
{
|
| 1961 |
+
"epoch": 2.66,
|
| 1962 |
+
"learning_rate": 1.4680134680134681e-05,
|
| 1963 |
+
"loss": 0.5626,
|
| 1964 |
+
"step": 3160
|
| 1965 |
+
},
|
| 1966 |
+
{
|
| 1967 |
+
"epoch": 2.66,
|
| 1968 |
+
"eval_accuracy": 0.7175843694493783,
|
| 1969 |
+
"eval_f1_macro": 0.6878133884212742,
|
| 1970 |
+
"eval_f1_negative": 0.7006661115736886,
|
| 1971 |
+
"eval_f1_neutral": 0.7580354301474488,
|
| 1972 |
+
"eval_f1_positive": 0.6047386235426853,
|
| 1973 |
+
"eval_loss": 0.6416473984718323,
|
| 1974 |
+
"eval_precision_negative": 0.663905325443787,
|
| 1975 |
+
"eval_precision_neutral": 0.770541298253181,
|
| 1976 |
+
"eval_precision_positive": 0.631578947368421,
|
| 1977 |
+
"eval_recall_negative": 0.7417364477743499,
|
| 1978 |
+
"eval_recall_neutral": 0.7459290187891441,
|
| 1979 |
+
"eval_recall_positive": 0.5800865800865801,
|
| 1980 |
+
"eval_runtime": 13.5444,
|
| 1981 |
+
"eval_samples_per_second": 623.505,
|
| 1982 |
+
"eval_steps_per_second": 3.249,
|
| 1983 |
+
"step": 3160
|
| 1984 |
+
},
|
| 1985 |
+
{
|
| 1986 |
+
"epoch": 2.69,
|
| 1987 |
+
"learning_rate": 1.4612794612794614e-05,
|
| 1988 |
+
"loss": 0.5589,
|
| 1989 |
+
"step": 3200
|
| 1990 |
+
},
|
| 1991 |
+
{
|
| 1992 |
+
"epoch": 2.69,
|
| 1993 |
+
"eval_accuracy": 0.7187685020722321,
|
| 1994 |
+
"eval_f1_macro": 0.6920386854465886,
|
| 1995 |
+
"eval_f1_negative": 0.706980691703798,
|
| 1996 |
+
"eval_f1_neutral": 0.7570633561643835,
|
| 1997 |
+
"eval_f1_positive": 0.6120720084715848,
|
| 1998 |
+
"eval_loss": 0.6367910504341125,
|
| 1999 |
+
"eval_precision_negative": 0.6816693944353519,
|
| 2000 |
+
"eval_precision_neutral": 0.7766798418972332,
|
| 2001 |
+
"eval_precision_positive": 0.5991706979958535,
|
| 2002 |
+
"eval_recall_negative": 0.7342441604230939,
|
| 2003 |
+
"eval_recall_neutral": 0.7384133611691023,
|
| 2004 |
+
"eval_recall_positive": 0.6255411255411255,
|
| 2005 |
+
"eval_runtime": 13.5341,
|
| 2006 |
+
"eval_samples_per_second": 623.979,
|
| 2007 |
+
"eval_steps_per_second": 3.251,
|
| 2008 |
+
"step": 3200
|
| 2009 |
+
},
|
| 2010 |
+
{
|
| 2011 |
+
"epoch": 2.73,
|
| 2012 |
+
"learning_rate": 1.4545454545454546e-05,
|
| 2013 |
+
"loss": 0.5499,
|
| 2014 |
+
"step": 3240
|
| 2015 |
+
},
|
| 2016 |
+
{
|
| 2017 |
+
"epoch": 2.73,
|
| 2018 |
+
"eval_accuracy": 0.7135583185316755,
|
| 2019 |
+
"eval_f1_macro": 0.6872354106917024,
|
| 2020 |
+
"eval_f1_negative": 0.7027480595762535,
|
| 2021 |
+
"eval_f1_neutral": 0.7509126046811251,
|
| 2022 |
+
"eval_f1_positive": 0.6080455678177287,
|
| 2023 |
+
"eval_loss": 0.6405165791511536,
|
| 2024 |
+
"eval_precision_negative": 0.6705364291433147,
|
| 2025 |
+
"eval_precision_neutral": 0.7729885057471264,
|
| 2026 |
+
"eval_precision_positive": 0.6001405481377372,
|
| 2027 |
+
"eval_recall_negative": 0.7382106654914059,
|
| 2028 |
+
"eval_recall_neutral": 0.7300626304801671,
|
| 2029 |
+
"eval_recall_positive": 0.6161616161616161,
|
| 2030 |
+
"eval_runtime": 13.5535,
|
| 2031 |
+
"eval_samples_per_second": 623.087,
|
| 2032 |
+
"eval_steps_per_second": 3.246,
|
| 2033 |
+
"step": 3240
|
| 2034 |
+
},
|
| 2035 |
+
{
|
| 2036 |
+
"epoch": 2.76,
|
| 2037 |
+
"learning_rate": 1.447811447811448e-05,
|
| 2038 |
+
"loss": 0.5431,
|
| 2039 |
+
"step": 3280
|
| 2040 |
+
},
|
| 2041 |
+
{
|
| 2042 |
+
"epoch": 2.76,
|
| 2043 |
+
"eval_accuracy": 0.7207815275310835,
|
| 2044 |
+
"eval_f1_macro": 0.6892335124121108,
|
| 2045 |
+
"eval_f1_negative": 0.7012817727568977,
|
| 2046 |
+
"eval_f1_neutral": 0.7629961454318157,
|
| 2047 |
+
"eval_f1_positive": 0.6034226190476191,
|
| 2048 |
+
"eval_loss": 0.6356926560401917,
|
| 2049 |
+
"eval_precision_negative": 0.6915167095115681,
|
| 2050 |
+
"eval_precision_neutral": 0.7614888750259929,
|
| 2051 |
+
"eval_precision_positive": 0.6228878648233487,
|
| 2052 |
+
"eval_recall_negative": 0.7113265755839577,
|
| 2053 |
+
"eval_recall_neutral": 0.764509394572025,
|
| 2054 |
+
"eval_recall_positive": 0.5851370851370852,
|
| 2055 |
+
"eval_runtime": 13.6103,
|
| 2056 |
+
"eval_samples_per_second": 620.484,
|
| 2057 |
+
"eval_steps_per_second": 3.233,
|
| 2058 |
+
"step": 3280
|
| 2059 |
+
},
|
| 2060 |
+
{
|
| 2061 |
+
"epoch": 2.79,
|
| 2062 |
+
"learning_rate": 1.4410774410774413e-05,
|
| 2063 |
+
"loss": 0.5388,
|
| 2064 |
+
"step": 3320
|
| 2065 |
+
},
|
| 2066 |
+
{
|
| 2067 |
+
"epoch": 2.79,
|
| 2068 |
+
"eval_accuracy": 0.7103611604499704,
|
| 2069 |
+
"eval_f1_macro": 0.6887331835620346,
|
| 2070 |
+
"eval_f1_negative": 0.7057600332709503,
|
| 2071 |
+
"eval_f1_neutral": 0.7435107787065551,
|
| 2072 |
+
"eval_f1_positive": 0.6169287387085982,
|
| 2073 |
+
"eval_loss": 0.6553226113319397,
|
| 2074 |
+
"eval_precision_negative": 0.6681102362204724,
|
| 2075 |
+
"eval_precision_neutral": 0.7856810785681079,
|
| 2076 |
+
"eval_precision_positive": 0.5751715533374921,
|
| 2077 |
+
"eval_recall_negative": 0.747906566769502,
|
| 2078 |
+
"eval_recall_neutral": 0.7056367432150313,
|
| 2079 |
+
"eval_recall_positive": 0.6652236652236653,
|
| 2080 |
+
"eval_runtime": 13.4151,
|
| 2081 |
+
"eval_samples_per_second": 629.514,
|
| 2082 |
+
"eval_steps_per_second": 3.28,
|
| 2083 |
+
"step": 3320
|
| 2084 |
+
},
|
| 2085 |
+
{
|
| 2086 |
+
"epoch": 2.83,
|
| 2087 |
+
"learning_rate": 1.4343434343434344e-05,
|
| 2088 |
+
"loss": 0.5613,
|
| 2089 |
+
"step": 3360
|
| 2090 |
+
},
|
| 2091 |
+
{
|
| 2092 |
+
"epoch": 2.83,
|
| 2093 |
+
"eval_accuracy": 0.7232682060390764,
|
| 2094 |
+
"eval_f1_macro": 0.6875059651526924,
|
| 2095 |
+
"eval_f1_negative": 0.7051926298157454,
|
| 2096 |
+
"eval_f1_neutral": 0.7659927423535511,
|
| 2097 |
+
"eval_f1_positive": 0.5913325232887808,
|
| 2098 |
+
"eval_loss": 0.6239017844200134,
|
| 2099 |
+
"eval_precision_negative": 0.6717191862784204,
|
| 2100 |
+
"eval_precision_neutral": 0.76086508753862,
|
| 2101 |
+
"eval_precision_positive": 0.6740535549399815,
|
| 2102 |
+
"eval_recall_negative": 0.742177170559718,
|
| 2103 |
+
"eval_recall_neutral": 0.7711899791231733,
|
| 2104 |
+
"eval_recall_positive": 0.5266955266955267,
|
| 2105 |
+
"eval_runtime": 13.4098,
|
| 2106 |
+
"eval_samples_per_second": 629.763,
|
| 2107 |
+
"eval_steps_per_second": 3.281,
|
| 2108 |
+
"step": 3360
|
| 2109 |
+
},
|
| 2110 |
+
{
|
| 2111 |
+
"epoch": 2.86,
|
| 2112 |
+
"learning_rate": 1.4276094276094276e-05,
|
| 2113 |
+
"loss": 0.556,
|
| 2114 |
+
"step": 3400
|
| 2115 |
+
},
|
| 2116 |
+
{
|
| 2117 |
+
"epoch": 2.86,
|
| 2118 |
+
"eval_accuracy": 0.7217288336293665,
|
| 2119 |
+
"eval_f1_macro": 0.6902285547620584,
|
| 2120 |
+
"eval_f1_negative": 0.7011167068097219,
|
| 2121 |
+
"eval_f1_neutral": 0.7648837451777708,
|
| 2122 |
+
"eval_f1_positive": 0.6046852122986822,
|
| 2123 |
+
"eval_loss": 0.6339342594146729,
|
| 2124 |
+
"eval_precision_negative": 0.696692776327241,
|
| 2125 |
+
"eval_precision_neutral": 0.7640074984378254,
|
| 2126 |
+
"eval_precision_positive": 0.6136701337295691,
|
| 2127 |
+
"eval_recall_negative": 0.7055971793741737,
|
| 2128 |
+
"eval_recall_neutral": 0.7657620041753653,
|
| 2129 |
+
"eval_recall_positive": 0.5959595959595959,
|
| 2130 |
+
"eval_runtime": 13.436,
|
| 2131 |
+
"eval_samples_per_second": 628.537,
|
| 2132 |
+
"eval_steps_per_second": 3.275,
|
| 2133 |
+
"step": 3400
|
| 2134 |
+
},
|
| 2135 |
+
{
|
| 2136 |
+
"epoch": 2.9,
|
| 2137 |
+
"learning_rate": 1.420875420875421e-05,
|
| 2138 |
+
"loss": 0.5536,
|
| 2139 |
+
"step": 3440
|
| 2140 |
+
},
|
| 2141 |
+
{
|
| 2142 |
+
"epoch": 2.9,
|
| 2143 |
+
"eval_accuracy": 0.7108348134991119,
|
| 2144 |
+
"eval_f1_macro": 0.6898151957724491,
|
| 2145 |
+
"eval_f1_negative": 0.7129761662253005,
|
| 2146 |
+
"eval_f1_neutral": 0.7415480427046264,
|
| 2147 |
+
"eval_f1_positive": 0.6149213783874204,
|
| 2148 |
+
"eval_loss": 0.659329891204834,
|
| 2149 |
+
"eval_precision_negative": 0.6628787878787878,
|
| 2150 |
+
"eval_precision_neutral": 0.7934316991908615,
|
| 2151 |
+
"eval_precision_positive": 0.5733000623830318,
|
| 2152 |
+
"eval_recall_negative": 0.7712648743940062,
|
| 2153 |
+
"eval_recall_neutral": 0.6960334029227557,
|
| 2154 |
+
"eval_recall_positive": 0.663059163059163,
|
| 2155 |
+
"eval_runtime": 13.4431,
|
| 2156 |
+
"eval_samples_per_second": 628.201,
|
| 2157 |
+
"eval_steps_per_second": 3.273,
|
| 2158 |
+
"step": 3440
|
| 2159 |
+
},
|
| 2160 |
+
{
|
| 2161 |
+
"epoch": 2.93,
|
| 2162 |
+
"learning_rate": 1.4141414141414143e-05,
|
| 2163 |
+
"loss": 0.5663,
|
| 2164 |
+
"step": 3480
|
| 2165 |
+
},
|
| 2166 |
+
{
|
| 2167 |
+
"epoch": 2.93,
|
| 2168 |
+
"eval_accuracy": 0.7154529307282416,
|
| 2169 |
+
"eval_f1_macro": 0.6878232234805552,
|
| 2170 |
+
"eval_f1_negative": 0.6974677920924034,
|
| 2171 |
+
"eval_f1_neutral": 0.7573062261753494,
|
| 2172 |
+
"eval_f1_positive": 0.6086956521739131,
|
| 2173 |
+
"eval_loss": 0.6353215575218201,
|
| 2174 |
+
"eval_precision_negative": 0.703090013434841,
|
| 2175 |
+
"eval_precision_neutral": 0.7683712935109583,
|
| 2176 |
+
"eval_precision_positive": 0.5750962772785623,
|
| 2177 |
+
"eval_recall_negative": 0.6919347730277655,
|
| 2178 |
+
"eval_recall_neutral": 0.7465553235908142,
|
| 2179 |
+
"eval_recall_positive": 0.6464646464646465,
|
| 2180 |
+
"eval_runtime": 13.4095,
|
| 2181 |
+
"eval_samples_per_second": 629.78,
|
| 2182 |
+
"eval_steps_per_second": 3.281,
|
| 2183 |
+
"step": 3480
|
| 2184 |
+
},
|
| 2185 |
+
{
|
| 2186 |
+
"epoch": 2.96,
|
| 2187 |
+
"learning_rate": 1.4074074074074075e-05,
|
| 2188 |
+
"loss": 0.5499,
|
| 2189 |
+
"step": 3520
|
| 2190 |
+
},
|
| 2191 |
+
{
|
| 2192 |
+
"epoch": 2.96,
|
| 2193 |
+
"eval_accuracy": 0.7143872113676731,
|
| 2194 |
+
"eval_f1_macro": 0.6899472024314622,
|
| 2195 |
+
"eval_f1_negative": 0.7078491335372069,
|
| 2196 |
+
"eval_f1_neutral": 0.7488017429193901,
|
| 2197 |
+
"eval_f1_positive": 0.6131907308377897,
|
| 2198 |
+
"eval_loss": 0.6429401636123657,
|
| 2199 |
+
"eval_precision_negative": 0.6585735963581184,
|
| 2200 |
+
"eval_precision_neutral": 0.7829157175398633,
|
| 2201 |
+
"eval_precision_positive": 0.6060606060606061,
|
| 2202 |
+
"eval_recall_negative": 0.7650947553988541,
|
| 2203 |
+
"eval_recall_neutral": 0.7175365344467641,
|
| 2204 |
+
"eval_recall_positive": 0.6204906204906205,
|
| 2205 |
+
"eval_runtime": 13.4298,
|
| 2206 |
+
"eval_samples_per_second": 628.824,
|
| 2207 |
+
"eval_steps_per_second": 3.276,
|
| 2208 |
+
"step": 3520
|
| 2209 |
+
},
|
| 2210 |
+
{
|
| 2211 |
+
"epoch": 3.0,
|
| 2212 |
+
"learning_rate": 1.4006734006734009e-05,
|
| 2213 |
+
"loss": 0.5619,
|
| 2214 |
+
"step": 3560
|
| 2215 |
+
},
|
| 2216 |
+
{
|
| 2217 |
+
"epoch": 3.0,
|
| 2218 |
+
"eval_accuracy": 0.7207815275310835,
|
| 2219 |
+
"eval_f1_macro": 0.6934678744913757,
|
| 2220 |
+
"eval_f1_negative": 0.7042136003337507,
|
| 2221 |
+
"eval_f1_neutral": 0.759215853398679,
|
| 2222 |
+
"eval_f1_positive": 0.6169741697416974,
|
| 2223 |
+
"eval_loss": 0.6337869167327881,
|
| 2224 |
+
"eval_precision_negative": 0.6685148514851486,
|
| 2225 |
+
"eval_precision_neutral": 0.7752393385552655,
|
| 2226 |
+
"eval_precision_positive": 0.6314199395770392,
|
| 2227 |
+
"eval_recall_negative": 0.74394006170119,
|
| 2228 |
+
"eval_recall_neutral": 0.7438413361169103,
|
| 2229 |
+
"eval_recall_positive": 0.6031746031746031,
|
| 2230 |
+
"eval_runtime": 13.4185,
|
| 2231 |
+
"eval_samples_per_second": 629.354,
|
| 2232 |
+
"eval_steps_per_second": 3.279,
|
| 2233 |
+
"step": 3560
|
| 2234 |
+
}
|
| 2235 |
+
],
|
| 2236 |
+
"max_steps": 11880,
|
| 2237 |
+
"num_train_epochs": 10,
|
| 2238 |
+
"total_flos": 1.849503866686182e+16,
|
| 2239 |
+
"trial_name": null,
|
| 2240 |
+
"trial_params": null
|
| 2241 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:484a72da6829ad6a4c10c064f5e6c35d5339581c032251ac7757b97812f36562
|
| 3 |
+
size 3439
|