Commit
·
5582a81
1
Parent(s):
eb5a86b
Added LP FT model
Browse files- README.md +27 -28
- all_results.json +44 -44
- config.json +1 -1
- pytorch_model.bin +1 -1
- test_results.json +15 -15
- tokenizer_config.json +1 -1
- train_results.json +15 -15
- trainer_state.json +806 -950
- training_args.bin +1 -1
- validation_results.json +15 -15
README.md
CHANGED
|
@@ -5,7 +5,7 @@ tags:
|
|
| 5 |
datasets:
|
| 6 |
- wnut_17
|
| 7 |
model-index:
|
| 8 |
-
- name:
|
| 9 |
results: []
|
| 10 |
---
|
| 11 |
|
|
@@ -17,17 +17,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 17 |
This model is a fine-tuned version of [vinai/bertweet-base](https://huggingface.co/vinai/bertweet-base) on the [wnut_17](https://huggingface.co/datasets/wnut_17) dataset.
|
| 18 |
|
| 19 |
It achieves the following results on the evaluation set:
|
| 20 |
-
- Loss: 0.
|
| 21 |
-
- Overall Precision: 0.
|
| 22 |
-
- Overall Recall: 0.
|
| 23 |
-
- Overall F1: 0.
|
| 24 |
-
- Overall Accuracy: 0.
|
| 25 |
-
- Corporation F1: 0.
|
| 26 |
-
- Creative-work F1: 0.
|
| 27 |
-
- Group F1: 0.
|
| 28 |
-
- Location F1: 0.
|
| 29 |
-
- Person F1: 0.
|
| 30 |
-
- Product F1: 0.
|
| 31 |
|
| 32 |
## Model description
|
| 33 |
|
|
@@ -46,7 +46,7 @@ More information needed
|
|
| 46 |
### Training hyperparameters
|
| 47 |
|
| 48 |
The following hyperparameters were used during training:
|
| 49 |
-
- learning_rate:
|
| 50 |
- train_batch_size: 16
|
| 51 |
- eval_batch_size: 16
|
| 52 |
- seed: 42
|
|
@@ -58,21 +58,20 @@ The following hyperparameters were used during training:
|
|
| 58 |
|
| 59 |
| Training Loss | Epoch | Step | Validation Loss | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy | Corporation F1 | Creative-work F1 | Group F1 | Location F1 | Person F1 | Product F1 |
|
| 60 |
|:-------------:|:-----:|:----:|:---------------:|:-----------------:|:--------------:|:----------:|:----------------:|:--------------:|:----------------:|:--------:|:-----------:|:---------:|:----------:|
|
| 61 |
-
| 0.
|
| 62 |
-
| 0.
|
| 63 |
-
| 0.
|
| 64 |
-
| 0.
|
| 65 |
-
| 0.
|
| 66 |
-
| 0.
|
| 67 |
-
| 0.
|
| 68 |
-
| 0.
|
| 69 |
-
| 0.
|
| 70 |
-
| 0.
|
| 71 |
-
| 0.
|
| 72 |
-
| 0.
|
| 73 |
-
| 0.
|
| 74 |
-
| 0.
|
| 75 |
-
| 0.0075 | 15.0 | 3195 | 0.3239 | 0.6913 | 0.5914 | 0.6374 | 0.9499 | 0.2703 | 0.3636 | 0.4030 | 0.7500 | 0.7733 | 0.4152 |
|
| 76 |
|
| 77 |
|
| 78 |
### Framework versions
|
|
|
|
| 5 |
datasets:
|
| 6 |
- wnut_17
|
| 7 |
model-index:
|
| 8 |
+
- name: fine_tune_bert_output_LP_FP
|
| 9 |
results: []
|
| 10 |
---
|
| 11 |
|
|
|
|
| 17 |
This model is a fine-tuned version of [vinai/bertweet-base](https://huggingface.co/vinai/bertweet-base) on the [wnut_17](https://huggingface.co/datasets/wnut_17) dataset.
|
| 18 |
|
| 19 |
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.3376
|
| 21 |
+
- Overall Precision: 0.6803
|
| 22 |
+
- Overall Recall: 0.6096
|
| 23 |
+
- Overall F1: 0.6430
|
| 24 |
+
- Overall Accuracy: 0.9509
|
| 25 |
+
- Corporation F1: 0.2975
|
| 26 |
+
- Creative-work F1: 0.4436
|
| 27 |
+
- Group F1: 0.3624
|
| 28 |
+
- Location F1: 0.6834
|
| 29 |
+
- Person F1: 0.7902
|
| 30 |
+
- Product F1: 0.3887
|
| 31 |
|
| 32 |
## Model description
|
| 33 |
|
|
|
|
| 46 |
### Training hyperparameters
|
| 47 |
|
| 48 |
The following hyperparameters were used during training:
|
| 49 |
+
- learning_rate: 1e-05
|
| 50 |
- train_batch_size: 16
|
| 51 |
- eval_batch_size: 16
|
| 52 |
- seed: 42
|
|
|
|
| 58 |
|
| 59 |
| Training Loss | Epoch | Step | Validation Loss | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy | Corporation F1 | Creative-work F1 | Group F1 | Location F1 | Person F1 | Product F1 |
|
| 60 |
|:-------------:|:-----:|:----:|:---------------:|:-----------------:|:--------------:|:----------:|:----------------:|:--------------:|:----------------:|:--------:|:-----------:|:---------:|:----------:|
|
| 61 |
+
| 0.0215 | 1.0 | 213 | 0.2913 | 0.7026 | 0.5905 | 0.6417 | 0.9507 | 0.2832 | 0.4444 | 0.2975 | 0.6854 | 0.7788 | 0.4015 |
|
| 62 |
+
| 0.0213 | 2.0 | 426 | 0.3052 | 0.6774 | 0.5772 | 0.6233 | 0.9495 | 0.2830 | 0.3483 | 0.3231 | 0.6857 | 0.7728 | 0.3794 |
|
| 63 |
+
| 0.0288 | 3.0 | 639 | 0.3378 | 0.7061 | 0.5507 | 0.6188 | 0.9467 | 0.3077 | 0.4184 | 0.3529 | 0.6222 | 0.7532 | 0.3910 |
|
| 64 |
+
| 0.0124 | 4.0 | 852 | 0.2712 | 0.6574 | 0.6121 | 0.6340 | 0.9502 | 0.3077 | 0.4842 | 0.3167 | 0.6809 | 0.7735 | 0.3986 |
|
| 65 |
+
| 0.0208 | 5.0 | 1065 | 0.2905 | 0.7108 | 0.6063 | 0.6544 | 0.9518 | 0.3063 | 0.4286 | 0.3419 | 0.7052 | 0.7913 | 0.4223 |
|
| 66 |
+
| 0.0071 | 6.0 | 1278 | 0.3189 | 0.6756 | 0.5847 | 0.6269 | 0.9494 | 0.2759 | 0.4380 | 0.3256 | 0.6744 | 0.7781 | 0.3779 |
|
| 67 |
+
| 0.0073 | 7.0 | 1491 | 0.3593 | 0.7330 | 0.5540 | 0.6310 | 0.9476 | 0.3061 | 0.4388 | 0.3784 | 0.6946 | 0.7631 | 0.3374 |
|
| 68 |
+
| 0.0135 | 8.0 | 1704 | 0.3564 | 0.6875 | 0.5482 | 0.6100 | 0.9471 | 0.34 | 0.4179 | 0.3088 | 0.6632 | 0.7486 | 0.3695 |
|
| 69 |
+
| 0.0097 | 9.0 | 1917 | 0.3085 | 0.6598 | 0.6395 | 0.6495 | 0.9516 | 0.3111 | 0.4609 | 0.3836 | 0.7090 | 0.7906 | 0.4083 |
|
| 70 |
+
| 0.0108 | 10.0 | 2130 | 0.3045 | 0.6605 | 0.6478 | 0.6541 | 0.9509 | 0.3529 | 0.4580 | 0.3649 | 0.6897 | 0.7843 | 0.4387 |
|
| 71 |
+
| 0.013 | 11.0 | 2343 | 0.3383 | 0.6788 | 0.6179 | 0.6470 | 0.9507 | 0.2783 | 0.4248 | 0.3358 | 0.7368 | 0.7958 | 0.3655 |
|
| 72 |
+
| 0.0076 | 12.0 | 2556 | 0.3617 | 0.6920 | 0.5523 | 0.6143 | 0.9474 | 0.2708 | 0.3985 | 0.3333 | 0.6740 | 0.7566 | 0.3525 |
|
| 73 |
+
| 0.0042 | 13.0 | 2769 | 0.3747 | 0.6896 | 0.5664 | 0.6220 | 0.9473 | 0.2478 | 0.3915 | 0.3521 | 0.6561 | 0.7742 | 0.3539 |
|
| 74 |
+
| 0.0049 | 14.0 | 2982 | 0.3376 | 0.6803 | 0.6096 | 0.6430 | 0.9509 | 0.2975 | 0.4436 | 0.3624 | 0.6834 | 0.7902 | 0.3887 |
|
|
|
|
| 75 |
|
| 76 |
|
| 77 |
### Framework versions
|
all_results.json
CHANGED
|
@@ -1,46 +1,46 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"test_corporation_f1": 0.
|
| 4 |
-
"test_creative-work_f1": 0.
|
| 5 |
-
"test_group_f1": 0.
|
| 6 |
-
"test_location_f1": 0.
|
| 7 |
-
"test_loss": 0.
|
| 8 |
-
"test_overall_accuracy": 0.
|
| 9 |
-
"test_overall_f1": 0.
|
| 10 |
-
"test_overall_precision": 0.
|
| 11 |
-
"test_overall_recall": 0.
|
| 12 |
-
"test_person_f1": 0.
|
| 13 |
-
"test_product_f1": 0.
|
| 14 |
-
"test_runtime":
|
| 15 |
-
"test_samples_per_second":
|
| 16 |
-
"test_steps_per_second":
|
| 17 |
-
"total_flos":
|
| 18 |
-
"train_corporation_f1": 0.
|
| 19 |
-
"train_creative-work_f1": 0.
|
| 20 |
-
"train_group_f1": 0.
|
| 21 |
-
"train_location_f1": 0.
|
| 22 |
-
"train_loss": 0.
|
| 23 |
-
"train_overall_accuracy": 0.
|
| 24 |
-
"train_overall_f1": 0.
|
| 25 |
-
"train_overall_precision": 0.
|
| 26 |
-
"train_overall_recall": 0.
|
| 27 |
-
"train_person_f1": 0.
|
| 28 |
-
"train_product_f1": 0.
|
| 29 |
-
"train_runtime":
|
| 30 |
-
"train_samples_per_second":
|
| 31 |
-
"train_steps_per_second":
|
| 32 |
-
"validation_corporation_f1": 0.
|
| 33 |
-
"validation_creative-work_f1": 0.
|
| 34 |
-
"validation_group_f1": 0.
|
| 35 |
-
"validation_location_f1": 0.
|
| 36 |
-
"validation_loss": 0.
|
| 37 |
-
"validation_overall_accuracy": 0.
|
| 38 |
-
"validation_overall_f1": 0.
|
| 39 |
-
"validation_overall_precision": 0.
|
| 40 |
-
"validation_overall_recall": 0.
|
| 41 |
-
"validation_person_f1": 0.
|
| 42 |
-
"validation_product_f1": 0.
|
| 43 |
-
"validation_runtime":
|
| 44 |
-
"validation_samples_per_second":
|
| 45 |
-
"validation_steps_per_second":
|
| 46 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 14.0,
|
| 3 |
+
"test_corporation_f1": 0.3686274509803922,
|
| 4 |
+
"test_creative-work_f1": 0.41527446300715987,
|
| 5 |
+
"test_group_f1": 0.4113475177304965,
|
| 6 |
+
"test_location_f1": 0.6386946386946387,
|
| 7 |
+
"test_loss": 0.2739429473876953,
|
| 8 |
+
"test_overall_accuracy": 0.9499198834668608,
|
| 9 |
+
"test_overall_f1": 0.5508159175493844,
|
| 10 |
+
"test_overall_precision": 0.6154830454254638,
|
| 11 |
+
"test_overall_recall": 0.49844559585492226,
|
| 12 |
+
"test_person_f1": 0.6898263027295285,
|
| 13 |
+
"test_product_f1": 0.27042253521126763,
|
| 14 |
+
"test_runtime": 8.5968,
|
| 15 |
+
"test_samples_per_second": 149.707,
|
| 16 |
+
"test_steps_per_second": 9.422,
|
| 17 |
+
"total_flos": 1093917417133104.0,
|
| 18 |
+
"train_corporation_f1": 0.9368770764119602,
|
| 19 |
+
"train_creative-work_f1": 0.8805620608899297,
|
| 20 |
+
"train_group_f1": 0.9755469755469754,
|
| 21 |
+
"train_location_f1": 0.9788867562380038,
|
| 22 |
+
"train_loss": 0.012029974721372128,
|
| 23 |
+
"train_overall_accuracy": 0.9977088917909593,
|
| 24 |
+
"train_overall_f1": 0.9611130931145201,
|
| 25 |
+
"train_overall_precision": 0.9563365282215123,
|
| 26 |
+
"train_overall_recall": 0.9659376120473288,
|
| 27 |
+
"train_person_f1": 0.9842519685039369,
|
| 28 |
+
"train_product_f1": 0.8932461873638344,
|
| 29 |
+
"train_runtime": 16.2924,
|
| 30 |
+
"train_samples_per_second": 208.318,
|
| 31 |
+
"train_steps_per_second": 13.074,
|
| 32 |
+
"validation_corporation_f1": 0.30769230769230765,
|
| 33 |
+
"validation_creative-work_f1": 0.48421052631578954,
|
| 34 |
+
"validation_group_f1": 0.31666666666666665,
|
| 35 |
+
"validation_location_f1": 0.6808510638297872,
|
| 36 |
+
"validation_loss": 0.27115458250045776,
|
| 37 |
+
"validation_overall_accuracy": 0.9502437284508382,
|
| 38 |
+
"validation_overall_f1": 0.6339784946236559,
|
| 39 |
+
"validation_overall_precision": 0.6574487065120428,
|
| 40 |
+
"validation_overall_recall": 0.6121262458471761,
|
| 41 |
+
"validation_person_f1": 0.7734553775743707,
|
| 42 |
+
"validation_product_f1": 0.3986254295532646,
|
| 43 |
+
"validation_runtime": 5.0688,
|
| 44 |
+
"validation_samples_per_second": 199.06,
|
| 45 |
+
"validation_steps_per_second": 12.626
|
| 46 |
}
|
config.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "
|
| 3 |
"architectures": [
|
| 4 |
"RobertaForTokenClassification"
|
| 5 |
],
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "./bertweet-base_wnut17_ner_LP",
|
| 3 |
"architectures": [
|
| 4 |
"RobertaForTokenClassification"
|
| 5 |
],
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 537360049
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b9c36319539aecd20f2abd9ca4cc349453802d1a83aa9b6cf08ff2cd537a51f
|
| 3 |
size 537360049
|
test_results.json
CHANGED
|
@@ -1,17 +1,17 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"test_corporation_f1": 0.
|
| 4 |
-
"test_creative-work_f1": 0.
|
| 5 |
-
"test_group_f1": 0.
|
| 6 |
-
"test_location_f1": 0.
|
| 7 |
-
"test_loss": 0.
|
| 8 |
-
"test_overall_accuracy": 0.
|
| 9 |
-
"test_overall_f1": 0.
|
| 10 |
-
"test_overall_precision": 0.
|
| 11 |
-
"test_overall_recall": 0.
|
| 12 |
-
"test_person_f1": 0.
|
| 13 |
-
"test_product_f1": 0.
|
| 14 |
-
"test_runtime":
|
| 15 |
-
"test_samples_per_second":
|
| 16 |
-
"test_steps_per_second":
|
| 17 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 14.0,
|
| 3 |
+
"test_corporation_f1": 0.3686274509803922,
|
| 4 |
+
"test_creative-work_f1": 0.41527446300715987,
|
| 5 |
+
"test_group_f1": 0.4113475177304965,
|
| 6 |
+
"test_location_f1": 0.6386946386946387,
|
| 7 |
+
"test_loss": 0.2739429473876953,
|
| 8 |
+
"test_overall_accuracy": 0.9499198834668608,
|
| 9 |
+
"test_overall_f1": 0.5508159175493844,
|
| 10 |
+
"test_overall_precision": 0.6154830454254638,
|
| 11 |
+
"test_overall_recall": 0.49844559585492226,
|
| 12 |
+
"test_person_f1": 0.6898263027295285,
|
| 13 |
+
"test_product_f1": 0.27042253521126763,
|
| 14 |
+
"test_runtime": 8.5968,
|
| 15 |
+
"test_samples_per_second": 149.707,
|
| 16 |
+
"test_steps_per_second": 9.422
|
| 17 |
}
|
tokenizer_config.json
CHANGED
|
@@ -1 +1 @@
|
|
| 1 |
-
{"normalization": false, "bos_token": "<s>", "eos_token": "</s>", "sep_token": "</s>", "cls_token": "<s>", "unk_token": "<unk>", "pad_token": "<pad>", "mask_token": "<mask>", "special_tokens_map_file": null, "name_or_path": "
|
|
|
|
| 1 |
+
{"normalization": false, "bos_token": "<s>", "eos_token": "</s>", "sep_token": "</s>", "cls_token": "<s>", "unk_token": "<unk>", "pad_token": "<pad>", "mask_token": "<mask>", "special_tokens_map_file": null, "name_or_path": "./bertweet-base_wnut17_ner_LP", "tokenizer_class": "BertweetTokenizer"}
|
train_results.json
CHANGED
|
@@ -1,17 +1,17 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_corporation_f1": 0.
|
| 4 |
-
"train_creative-work_f1": 0.
|
| 5 |
-
"train_group_f1": 0.
|
| 6 |
-
"train_location_f1": 0.
|
| 7 |
-
"train_loss": 0.
|
| 8 |
-
"train_overall_accuracy": 0.
|
| 9 |
-
"train_overall_f1": 0.
|
| 10 |
-
"train_overall_precision": 0.
|
| 11 |
-
"train_overall_recall": 0.
|
| 12 |
-
"train_person_f1": 0.
|
| 13 |
-
"train_product_f1": 0.
|
| 14 |
-
"train_runtime":
|
| 15 |
-
"train_samples_per_second":
|
| 16 |
-
"train_steps_per_second":
|
| 17 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 14.0,
|
| 3 |
+
"train_corporation_f1": 0.9368770764119602,
|
| 4 |
+
"train_creative-work_f1": 0.8805620608899297,
|
| 5 |
+
"train_group_f1": 0.9755469755469754,
|
| 6 |
+
"train_location_f1": 0.9788867562380038,
|
| 7 |
+
"train_loss": 0.012029974721372128,
|
| 8 |
+
"train_overall_accuracy": 0.9977088917909593,
|
| 9 |
+
"train_overall_f1": 0.9611130931145201,
|
| 10 |
+
"train_overall_precision": 0.9563365282215123,
|
| 11 |
+
"train_overall_recall": 0.9659376120473288,
|
| 12 |
+
"train_person_f1": 0.9842519685039369,
|
| 13 |
+
"train_product_f1": 0.8932461873638344,
|
| 14 |
+
"train_runtime": 16.2924,
|
| 15 |
+
"train_samples_per_second": 208.318,
|
| 16 |
+
"train_steps_per_second": 13.074
|
| 17 |
}
|
trainer_state.json
CHANGED
|
@@ -1,2215 +1,2071 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
| 9 |
"log_history": [
|
| 10 |
{
|
| 11 |
"epoch": 0.0,
|
| 12 |
-
"learning_rate":
|
| 13 |
-
"loss":
|
| 14 |
"step": 1
|
| 15 |
},
|
| 16 |
{
|
| 17 |
"epoch": 0.05,
|
| 18 |
-
"learning_rate":
|
| 19 |
-
"loss":
|
| 20 |
"step": 10
|
| 21 |
},
|
| 22 |
{
|
| 23 |
"epoch": 0.09,
|
| 24 |
-
"learning_rate":
|
| 25 |
-
"loss": 0.
|
| 26 |
"step": 20
|
| 27 |
},
|
| 28 |
{
|
| 29 |
"epoch": 0.14,
|
| 30 |
-
"learning_rate":
|
| 31 |
-
"loss": 0.
|
| 32 |
"step": 30
|
| 33 |
},
|
| 34 |
{
|
| 35 |
"epoch": 0.19,
|
| 36 |
-
"learning_rate":
|
| 37 |
-
"loss": 0.
|
| 38 |
"step": 40
|
| 39 |
},
|
| 40 |
{
|
| 41 |
"epoch": 0.23,
|
| 42 |
-
"learning_rate":
|
| 43 |
-
"loss": 0.
|
| 44 |
"step": 50
|
| 45 |
},
|
| 46 |
{
|
| 47 |
"epoch": 0.28,
|
| 48 |
-
"learning_rate":
|
| 49 |
-
"loss": 0.
|
| 50 |
"step": 60
|
| 51 |
},
|
| 52 |
{
|
| 53 |
"epoch": 0.33,
|
| 54 |
-
"learning_rate":
|
| 55 |
-
"loss": 0.
|
| 56 |
"step": 70
|
| 57 |
},
|
| 58 |
{
|
| 59 |
"epoch": 0.38,
|
| 60 |
-
"learning_rate":
|
| 61 |
-
"loss": 0.
|
| 62 |
"step": 80
|
| 63 |
},
|
| 64 |
{
|
| 65 |
"epoch": 0.42,
|
| 66 |
-
"learning_rate":
|
| 67 |
-
"loss": 0.
|
| 68 |
"step": 90
|
| 69 |
},
|
| 70 |
{
|
| 71 |
"epoch": 0.47,
|
| 72 |
-
"learning_rate":
|
| 73 |
-
"loss": 0.
|
| 74 |
"step": 100
|
| 75 |
},
|
| 76 |
{
|
| 77 |
"epoch": 0.52,
|
| 78 |
-
"learning_rate":
|
| 79 |
-
"loss": 0.
|
| 80 |
"step": 110
|
| 81 |
},
|
| 82 |
{
|
| 83 |
"epoch": 0.56,
|
| 84 |
-
"learning_rate":
|
| 85 |
-
"loss": 0.
|
| 86 |
"step": 120
|
| 87 |
},
|
| 88 |
{
|
| 89 |
"epoch": 0.61,
|
| 90 |
-
"learning_rate":
|
| 91 |
-
"loss": 0.
|
| 92 |
"step": 130
|
| 93 |
},
|
| 94 |
{
|
| 95 |
"epoch": 0.66,
|
| 96 |
-
"learning_rate":
|
| 97 |
-
"loss": 0.
|
| 98 |
"step": 140
|
| 99 |
},
|
| 100 |
{
|
| 101 |
"epoch": 0.7,
|
| 102 |
-
"learning_rate":
|
| 103 |
-
"loss": 0.
|
| 104 |
"step": 150
|
| 105 |
},
|
| 106 |
{
|
| 107 |
"epoch": 0.75,
|
| 108 |
-
"learning_rate":
|
| 109 |
-
"loss": 0.
|
| 110 |
"step": 160
|
| 111 |
},
|
| 112 |
{
|
| 113 |
"epoch": 0.8,
|
| 114 |
-
"learning_rate":
|
| 115 |
-
"loss": 0.
|
| 116 |
"step": 170
|
| 117 |
},
|
| 118 |
{
|
| 119 |
"epoch": 0.85,
|
| 120 |
-
"learning_rate":
|
| 121 |
-
"loss": 0.
|
| 122 |
"step": 180
|
| 123 |
},
|
| 124 |
{
|
| 125 |
"epoch": 0.89,
|
| 126 |
-
"learning_rate":
|
| 127 |
-
"loss": 0.
|
| 128 |
"step": 190
|
| 129 |
},
|
| 130 |
{
|
| 131 |
"epoch": 0.94,
|
| 132 |
-
"learning_rate":
|
| 133 |
-
"loss": 0.
|
| 134 |
"step": 200
|
| 135 |
},
|
| 136 |
{
|
| 137 |
"epoch": 0.99,
|
| 138 |
-
"learning_rate":
|
| 139 |
-
"loss": 0.
|
| 140 |
"step": 210
|
| 141 |
},
|
| 142 |
{
|
| 143 |
"epoch": 1.0,
|
| 144 |
-
"eval_corporation_f1": 0.
|
| 145 |
-
"eval_creative-work_f1": 0.
|
| 146 |
-
"eval_group_f1": 0.
|
| 147 |
-
"eval_location_f1": 0.
|
| 148 |
-
"eval_loss": 0.
|
| 149 |
-
"eval_overall_accuracy": 0.
|
| 150 |
-
"eval_overall_f1": 0.
|
| 151 |
-
"eval_overall_precision": 0.
|
| 152 |
-
"eval_overall_recall": 0.
|
| 153 |
-
"eval_person_f1": 0.
|
| 154 |
-
"eval_product_f1": 0.
|
| 155 |
-
"eval_runtime":
|
| 156 |
-
"eval_samples_per_second":
|
| 157 |
-
"eval_steps_per_second":
|
| 158 |
"step": 213
|
| 159 |
},
|
| 160 |
{
|
| 161 |
"epoch": 1.03,
|
| 162 |
-
"learning_rate":
|
| 163 |
-
"loss": 0.
|
| 164 |
"step": 220
|
| 165 |
},
|
| 166 |
{
|
| 167 |
"epoch": 1.08,
|
| 168 |
-
"learning_rate":
|
| 169 |
-
"loss": 0.
|
| 170 |
"step": 230
|
| 171 |
},
|
| 172 |
{
|
| 173 |
"epoch": 1.13,
|
| 174 |
-
"learning_rate":
|
| 175 |
-
"loss": 0.
|
| 176 |
"step": 240
|
| 177 |
},
|
| 178 |
{
|
| 179 |
"epoch": 1.17,
|
| 180 |
-
"learning_rate":
|
| 181 |
-
"loss": 0.
|
| 182 |
"step": 250
|
| 183 |
},
|
| 184 |
{
|
| 185 |
"epoch": 1.22,
|
| 186 |
-
"learning_rate":
|
| 187 |
-
"loss": 0.
|
| 188 |
"step": 260
|
| 189 |
},
|
| 190 |
{
|
| 191 |
"epoch": 1.27,
|
| 192 |
-
"learning_rate":
|
| 193 |
-
"loss": 0.
|
| 194 |
"step": 270
|
| 195 |
},
|
| 196 |
{
|
| 197 |
"epoch": 1.31,
|
| 198 |
-
"learning_rate":
|
| 199 |
-
"loss": 0.
|
| 200 |
"step": 280
|
| 201 |
},
|
| 202 |
{
|
| 203 |
"epoch": 1.36,
|
| 204 |
-
"learning_rate":
|
| 205 |
-
"loss": 0.
|
| 206 |
"step": 290
|
| 207 |
},
|
| 208 |
{
|
| 209 |
"epoch": 1.41,
|
| 210 |
-
"learning_rate":
|
| 211 |
-
"loss": 0.
|
| 212 |
"step": 300
|
| 213 |
},
|
| 214 |
{
|
| 215 |
"epoch": 1.46,
|
| 216 |
-
"learning_rate":
|
| 217 |
-
"loss": 0.
|
| 218 |
"step": 310
|
| 219 |
},
|
| 220 |
{
|
| 221 |
"epoch": 1.5,
|
| 222 |
-
"learning_rate":
|
| 223 |
-
"loss": 0.
|
| 224 |
"step": 320
|
| 225 |
},
|
| 226 |
{
|
| 227 |
"epoch": 1.55,
|
| 228 |
-
"learning_rate":
|
| 229 |
-
"loss": 0.
|
| 230 |
"step": 330
|
| 231 |
},
|
| 232 |
{
|
| 233 |
"epoch": 1.6,
|
| 234 |
-
"learning_rate":
|
| 235 |
-
"loss": 0.
|
| 236 |
"step": 340
|
| 237 |
},
|
| 238 |
{
|
| 239 |
"epoch": 1.64,
|
| 240 |
-
"learning_rate":
|
| 241 |
-
"loss": 0.
|
| 242 |
"step": 350
|
| 243 |
},
|
| 244 |
{
|
| 245 |
"epoch": 1.69,
|
| 246 |
-
"learning_rate":
|
| 247 |
-
"loss": 0.
|
| 248 |
"step": 360
|
| 249 |
},
|
| 250 |
{
|
| 251 |
"epoch": 1.74,
|
| 252 |
-
"learning_rate":
|
| 253 |
-
"loss": 0.
|
| 254 |
"step": 370
|
| 255 |
},
|
| 256 |
{
|
| 257 |
"epoch": 1.78,
|
| 258 |
-
"learning_rate":
|
| 259 |
-
"loss": 0.
|
| 260 |
"step": 380
|
| 261 |
},
|
| 262 |
{
|
| 263 |
"epoch": 1.83,
|
| 264 |
-
"learning_rate":
|
| 265 |
-
"loss": 0.
|
| 266 |
"step": 390
|
| 267 |
},
|
| 268 |
{
|
| 269 |
"epoch": 1.88,
|
| 270 |
-
"learning_rate":
|
| 271 |
-
"loss": 0.
|
| 272 |
"step": 400
|
| 273 |
},
|
| 274 |
{
|
| 275 |
"epoch": 1.92,
|
| 276 |
-
"learning_rate":
|
| 277 |
-
"loss": 0.
|
| 278 |
"step": 410
|
| 279 |
},
|
| 280 |
{
|
| 281 |
"epoch": 1.97,
|
| 282 |
-
"learning_rate":
|
| 283 |
-
"loss": 0.
|
| 284 |
"step": 420
|
| 285 |
},
|
| 286 |
{
|
| 287 |
"epoch": 2.0,
|
| 288 |
-
"eval_corporation_f1": 0.
|
| 289 |
-
"eval_creative-work_f1": 0.
|
| 290 |
-
"eval_group_f1": 0.
|
| 291 |
-
"eval_location_f1": 0.
|
| 292 |
-
"eval_loss": 0.
|
| 293 |
-
"eval_overall_accuracy": 0.
|
| 294 |
-
"eval_overall_f1": 0.
|
| 295 |
-
"eval_overall_precision": 0.
|
| 296 |
-
"eval_overall_recall": 0.
|
| 297 |
-
"eval_person_f1": 0.
|
| 298 |
-
"eval_product_f1": 0.
|
| 299 |
-
"eval_runtime":
|
| 300 |
-
"eval_samples_per_second":
|
| 301 |
-
"eval_steps_per_second":
|
| 302 |
"step": 426
|
| 303 |
},
|
| 304 |
{
|
| 305 |
"epoch": 2.02,
|
| 306 |
-
"learning_rate":
|
| 307 |
-
"loss": 0.
|
| 308 |
"step": 430
|
| 309 |
},
|
| 310 |
{
|
| 311 |
"epoch": 2.07,
|
| 312 |
-
"learning_rate":
|
| 313 |
-
"loss": 0.
|
| 314 |
"step": 440
|
| 315 |
},
|
| 316 |
{
|
| 317 |
"epoch": 2.11,
|
| 318 |
-
"learning_rate":
|
| 319 |
-
"loss": 0.
|
| 320 |
"step": 450
|
| 321 |
},
|
| 322 |
{
|
| 323 |
"epoch": 2.16,
|
| 324 |
-
"learning_rate":
|
| 325 |
-
"loss": 0.
|
| 326 |
"step": 460
|
| 327 |
},
|
| 328 |
{
|
| 329 |
"epoch": 2.21,
|
| 330 |
-
"learning_rate":
|
| 331 |
-
"loss": 0.
|
| 332 |
"step": 470
|
| 333 |
},
|
| 334 |
{
|
| 335 |
"epoch": 2.25,
|
| 336 |
-
"learning_rate":
|
| 337 |
-
"loss": 0.
|
| 338 |
"step": 480
|
| 339 |
},
|
| 340 |
{
|
| 341 |
"epoch": 2.3,
|
| 342 |
-
"learning_rate":
|
| 343 |
-
"loss": 0.
|
| 344 |
"step": 490
|
| 345 |
},
|
| 346 |
{
|
| 347 |
"epoch": 2.35,
|
| 348 |
-
"learning_rate":
|
| 349 |
-
"loss": 0.
|
| 350 |
"step": 500
|
| 351 |
},
|
| 352 |
{
|
| 353 |
"epoch": 2.39,
|
| 354 |
-
"learning_rate":
|
| 355 |
-
"loss": 0.
|
| 356 |
"step": 510
|
| 357 |
},
|
| 358 |
{
|
| 359 |
"epoch": 2.44,
|
| 360 |
-
"learning_rate":
|
| 361 |
-
"loss": 0.
|
| 362 |
"step": 520
|
| 363 |
},
|
| 364 |
{
|
| 365 |
"epoch": 2.49,
|
| 366 |
-
"learning_rate":
|
| 367 |
-
"loss": 0.
|
| 368 |
"step": 530
|
| 369 |
},
|
| 370 |
{
|
| 371 |
"epoch": 2.54,
|
| 372 |
-
"learning_rate":
|
| 373 |
-
"loss": 0.
|
| 374 |
"step": 540
|
| 375 |
},
|
| 376 |
{
|
| 377 |
"epoch": 2.58,
|
| 378 |
-
"learning_rate":
|
| 379 |
-
"loss": 0.
|
| 380 |
"step": 550
|
| 381 |
},
|
| 382 |
{
|
| 383 |
"epoch": 2.63,
|
| 384 |
-
"learning_rate":
|
| 385 |
-
"loss": 0.
|
| 386 |
"step": 560
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 2.68,
|
| 390 |
-
"learning_rate":
|
| 391 |
-
"loss": 0.
|
| 392 |
"step": 570
|
| 393 |
},
|
| 394 |
{
|
| 395 |
"epoch": 2.72,
|
| 396 |
-
"learning_rate":
|
| 397 |
-
"loss": 0.
|
| 398 |
"step": 580
|
| 399 |
},
|
| 400 |
{
|
| 401 |
"epoch": 2.77,
|
| 402 |
-
"learning_rate":
|
| 403 |
-
"loss": 0.
|
| 404 |
"step": 590
|
| 405 |
},
|
| 406 |
{
|
| 407 |
"epoch": 2.82,
|
| 408 |
-
"learning_rate":
|
| 409 |
-
"loss": 0.
|
| 410 |
"step": 600
|
| 411 |
},
|
| 412 |
{
|
| 413 |
"epoch": 2.86,
|
| 414 |
-
"learning_rate":
|
| 415 |
-
"loss": 0.
|
| 416 |
"step": 610
|
| 417 |
},
|
| 418 |
{
|
| 419 |
"epoch": 2.91,
|
| 420 |
-
"learning_rate":
|
| 421 |
-
"loss": 0.
|
| 422 |
"step": 620
|
| 423 |
},
|
| 424 |
{
|
| 425 |
"epoch": 2.96,
|
| 426 |
-
"learning_rate":
|
| 427 |
-
"loss": 0.
|
| 428 |
"step": 630
|
| 429 |
},
|
| 430 |
{
|
| 431 |
"epoch": 3.0,
|
| 432 |
-
"eval_corporation_f1": 0.
|
| 433 |
-
"eval_creative-work_f1": 0.
|
| 434 |
-
"eval_group_f1": 0.
|
| 435 |
-
"eval_location_f1": 0.
|
| 436 |
-
"eval_loss": 0.
|
| 437 |
-
"eval_overall_accuracy": 0.
|
| 438 |
-
"eval_overall_f1": 0.
|
| 439 |
-
"eval_overall_precision": 0.
|
| 440 |
-
"eval_overall_recall": 0.
|
| 441 |
-
"eval_person_f1": 0.
|
| 442 |
-
"eval_product_f1": 0.
|
| 443 |
-
"eval_runtime":
|
| 444 |
-
"eval_samples_per_second":
|
| 445 |
-
"eval_steps_per_second":
|
| 446 |
"step": 639
|
| 447 |
},
|
| 448 |
{
|
| 449 |
"epoch": 3.0,
|
| 450 |
-
"learning_rate":
|
| 451 |
-
"loss": 0.
|
| 452 |
"step": 640
|
| 453 |
},
|
| 454 |
{
|
| 455 |
"epoch": 3.05,
|
| 456 |
-
"learning_rate":
|
| 457 |
-
"loss": 0.
|
| 458 |
"step": 650
|
| 459 |
},
|
| 460 |
{
|
| 461 |
"epoch": 3.1,
|
| 462 |
-
"learning_rate":
|
| 463 |
-
"loss": 0.
|
| 464 |
"step": 660
|
| 465 |
},
|
| 466 |
{
|
| 467 |
"epoch": 3.15,
|
| 468 |
-
"learning_rate":
|
| 469 |
-
"loss": 0.
|
| 470 |
"step": 670
|
| 471 |
},
|
| 472 |
{
|
| 473 |
"epoch": 3.19,
|
| 474 |
-
"learning_rate":
|
| 475 |
-
"loss": 0.
|
| 476 |
"step": 680
|
| 477 |
},
|
| 478 |
{
|
| 479 |
"epoch": 3.24,
|
| 480 |
-
"learning_rate":
|
| 481 |
-
"loss": 0.
|
| 482 |
"step": 690
|
| 483 |
},
|
| 484 |
{
|
| 485 |
"epoch": 3.29,
|
| 486 |
-
"learning_rate":
|
| 487 |
-
"loss": 0.
|
| 488 |
"step": 700
|
| 489 |
},
|
| 490 |
{
|
| 491 |
"epoch": 3.33,
|
| 492 |
-
"learning_rate":
|
| 493 |
-
"loss": 0.
|
| 494 |
"step": 710
|
| 495 |
},
|
| 496 |
{
|
| 497 |
"epoch": 3.38,
|
| 498 |
-
"learning_rate":
|
| 499 |
-
"loss": 0.
|
| 500 |
"step": 720
|
| 501 |
},
|
| 502 |
{
|
| 503 |
"epoch": 3.43,
|
| 504 |
-
"learning_rate":
|
| 505 |
-
"loss": 0.
|
| 506 |
"step": 730
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 3.47,
|
| 510 |
-
"learning_rate":
|
| 511 |
-
"loss": 0.
|
| 512 |
"step": 740
|
| 513 |
},
|
| 514 |
{
|
| 515 |
"epoch": 3.52,
|
| 516 |
-
"learning_rate":
|
| 517 |
-
"loss": 0.
|
| 518 |
"step": 750
|
| 519 |
},
|
| 520 |
{
|
| 521 |
"epoch": 3.57,
|
| 522 |
-
"learning_rate":
|
| 523 |
-
"loss": 0.
|
| 524 |
"step": 760
|
| 525 |
},
|
| 526 |
{
|
| 527 |
"epoch": 3.62,
|
| 528 |
-
"learning_rate":
|
| 529 |
-
"loss": 0.
|
| 530 |
"step": 770
|
| 531 |
},
|
| 532 |
{
|
| 533 |
"epoch": 3.66,
|
| 534 |
-
"learning_rate":
|
| 535 |
-
"loss": 0.
|
| 536 |
"step": 780
|
| 537 |
},
|
| 538 |
{
|
| 539 |
"epoch": 3.71,
|
| 540 |
-
"learning_rate":
|
| 541 |
-
"loss": 0.
|
| 542 |
"step": 790
|
| 543 |
},
|
| 544 |
{
|
| 545 |
"epoch": 3.76,
|
| 546 |
-
"learning_rate":
|
| 547 |
-
"loss": 0.
|
| 548 |
"step": 800
|
| 549 |
},
|
| 550 |
{
|
| 551 |
"epoch": 3.8,
|
| 552 |
-
"learning_rate":
|
| 553 |
-
"loss": 0.
|
| 554 |
"step": 810
|
| 555 |
},
|
| 556 |
{
|
| 557 |
"epoch": 3.85,
|
| 558 |
-
"learning_rate":
|
| 559 |
-
"loss": 0.
|
| 560 |
"step": 820
|
| 561 |
},
|
| 562 |
{
|
| 563 |
"epoch": 3.9,
|
| 564 |
-
"learning_rate":
|
| 565 |
-
"loss": 0.
|
| 566 |
"step": 830
|
| 567 |
},
|
| 568 |
{
|
| 569 |
"epoch": 3.94,
|
| 570 |
-
"learning_rate":
|
| 571 |
-
"loss": 0.
|
| 572 |
"step": 840
|
| 573 |
},
|
| 574 |
{
|
| 575 |
"epoch": 3.99,
|
| 576 |
-
"learning_rate":
|
| 577 |
-
"loss": 0.
|
| 578 |
"step": 850
|
| 579 |
},
|
| 580 |
{
|
| 581 |
"epoch": 4.0,
|
| 582 |
-
"eval_corporation_f1": 0.
|
| 583 |
-
"eval_creative-work_f1": 0.
|
| 584 |
-
"eval_group_f1": 0.
|
| 585 |
-
"eval_location_f1": 0.
|
| 586 |
-
"eval_loss": 0.
|
| 587 |
-
"eval_overall_accuracy": 0.
|
| 588 |
-
"eval_overall_f1": 0.
|
| 589 |
-
"eval_overall_precision": 0.
|
| 590 |
-
"eval_overall_recall": 0.
|
| 591 |
-
"eval_person_f1": 0.
|
| 592 |
-
"eval_product_f1": 0.
|
| 593 |
-
"eval_runtime": 5.
|
| 594 |
-
"eval_samples_per_second":
|
| 595 |
-
"eval_steps_per_second": 12.
|
| 596 |
"step": 852
|
| 597 |
},
|
| 598 |
{
|
| 599 |
"epoch": 4.04,
|
| 600 |
-
"learning_rate":
|
| 601 |
-
"loss": 0.
|
| 602 |
"step": 860
|
| 603 |
},
|
| 604 |
{
|
| 605 |
"epoch": 4.08,
|
| 606 |
-
"learning_rate":
|
| 607 |
-
"loss": 0.
|
| 608 |
"step": 870
|
| 609 |
},
|
| 610 |
{
|
| 611 |
"epoch": 4.13,
|
| 612 |
-
"learning_rate":
|
| 613 |
-
"loss": 0.
|
| 614 |
"step": 880
|
| 615 |
},
|
| 616 |
{
|
| 617 |
"epoch": 4.18,
|
| 618 |
-
"learning_rate":
|
| 619 |
-
"loss": 0.
|
| 620 |
"step": 890
|
| 621 |
},
|
| 622 |
{
|
| 623 |
"epoch": 4.23,
|
| 624 |
-
"learning_rate":
|
| 625 |
-
"loss": 0.
|
| 626 |
"step": 900
|
| 627 |
},
|
| 628 |
{
|
| 629 |
"epoch": 4.27,
|
| 630 |
-
"learning_rate":
|
| 631 |
-
"loss": 0.
|
| 632 |
"step": 910
|
| 633 |
},
|
| 634 |
{
|
| 635 |
"epoch": 4.32,
|
| 636 |
-
"learning_rate":
|
| 637 |
-
"loss": 0.
|
| 638 |
"step": 920
|
| 639 |
},
|
| 640 |
{
|
| 641 |
"epoch": 4.37,
|
| 642 |
-
"learning_rate":
|
| 643 |
-
"loss": 0.
|
| 644 |
"step": 930
|
| 645 |
},
|
| 646 |
{
|
| 647 |
"epoch": 4.41,
|
| 648 |
-
"learning_rate":
|
| 649 |
-
"loss": 0.
|
| 650 |
"step": 940
|
| 651 |
},
|
| 652 |
{
|
| 653 |
"epoch": 4.46,
|
| 654 |
-
"learning_rate":
|
| 655 |
-
"loss": 0.
|
| 656 |
"step": 950
|
| 657 |
},
|
| 658 |
{
|
| 659 |
"epoch": 4.51,
|
| 660 |
-
"learning_rate":
|
| 661 |
-
"loss": 0.
|
| 662 |
"step": 960
|
| 663 |
},
|
| 664 |
{
|
| 665 |
"epoch": 4.55,
|
| 666 |
-
"learning_rate":
|
| 667 |
-
"loss": 0.
|
| 668 |
"step": 970
|
| 669 |
},
|
| 670 |
{
|
| 671 |
"epoch": 4.6,
|
| 672 |
-
"learning_rate":
|
| 673 |
-
"loss": 0.
|
| 674 |
"step": 980
|
| 675 |
},
|
| 676 |
{
|
| 677 |
"epoch": 4.65,
|
| 678 |
-
"learning_rate":
|
| 679 |
-
"loss": 0.
|
| 680 |
"step": 990
|
| 681 |
},
|
| 682 |
{
|
| 683 |
"epoch": 4.69,
|
| 684 |
-
"learning_rate":
|
| 685 |
-
"loss": 0.
|
| 686 |
"step": 1000
|
| 687 |
},
|
| 688 |
{
|
| 689 |
"epoch": 4.74,
|
| 690 |
-
"learning_rate":
|
| 691 |
-
"loss": 0.
|
| 692 |
"step": 1010
|
| 693 |
},
|
| 694 |
{
|
| 695 |
"epoch": 4.79,
|
| 696 |
-
"learning_rate":
|
| 697 |
-
"loss": 0.
|
| 698 |
"step": 1020
|
| 699 |
},
|
| 700 |
{
|
| 701 |
"epoch": 4.84,
|
| 702 |
-
"learning_rate":
|
| 703 |
-
"loss": 0.
|
| 704 |
"step": 1030
|
| 705 |
},
|
| 706 |
{
|
| 707 |
"epoch": 4.88,
|
| 708 |
-
"learning_rate":
|
| 709 |
-
"loss": 0.
|
| 710 |
"step": 1040
|
| 711 |
},
|
| 712 |
{
|
| 713 |
"epoch": 4.93,
|
| 714 |
-
"learning_rate":
|
| 715 |
-
"loss": 0.
|
| 716 |
"step": 1050
|
| 717 |
},
|
| 718 |
{
|
| 719 |
"epoch": 4.98,
|
| 720 |
-
"learning_rate":
|
| 721 |
-
"loss": 0.
|
| 722 |
"step": 1060
|
| 723 |
},
|
| 724 |
{
|
| 725 |
"epoch": 5.0,
|
| 726 |
-
"eval_corporation_f1": 0.
|
| 727 |
-
"eval_creative-work_f1": 0.
|
| 728 |
-
"eval_group_f1": 0.
|
| 729 |
-
"eval_location_f1": 0.
|
| 730 |
-
"eval_loss": 0.
|
| 731 |
-
"eval_overall_accuracy": 0.
|
| 732 |
-
"eval_overall_f1": 0.
|
| 733 |
-
"eval_overall_precision": 0.
|
| 734 |
-
"eval_overall_recall": 0.
|
| 735 |
-
"eval_person_f1": 0.
|
| 736 |
-
"eval_product_f1": 0.
|
| 737 |
-
"eval_runtime":
|
| 738 |
-
"eval_samples_per_second":
|
| 739 |
-
"eval_steps_per_second":
|
| 740 |
"step": 1065
|
| 741 |
},
|
| 742 |
{
|
| 743 |
"epoch": 5.02,
|
| 744 |
-
"learning_rate":
|
| 745 |
-
"loss": 0.
|
| 746 |
"step": 1070
|
| 747 |
},
|
| 748 |
{
|
| 749 |
"epoch": 5.07,
|
| 750 |
-
"learning_rate":
|
| 751 |
-
"loss": 0.
|
| 752 |
"step": 1080
|
| 753 |
},
|
| 754 |
{
|
| 755 |
"epoch": 5.12,
|
| 756 |
-
"learning_rate":
|
| 757 |
-
"loss": 0.
|
| 758 |
"step": 1090
|
| 759 |
},
|
| 760 |
{
|
| 761 |
"epoch": 5.16,
|
| 762 |
-
"learning_rate":
|
| 763 |
-
"loss": 0.
|
| 764 |
"step": 1100
|
| 765 |
},
|
| 766 |
{
|
| 767 |
"epoch": 5.21,
|
| 768 |
-
"learning_rate":
|
| 769 |
-
"loss": 0.
|
| 770 |
"step": 1110
|
| 771 |
},
|
| 772 |
{
|
| 773 |
"epoch": 5.26,
|
| 774 |
-
"learning_rate":
|
| 775 |
-
"loss": 0.
|
| 776 |
"step": 1120
|
| 777 |
},
|
| 778 |
{
|
| 779 |
"epoch": 5.31,
|
| 780 |
-
"learning_rate":
|
| 781 |
-
"loss": 0.
|
| 782 |
"step": 1130
|
| 783 |
},
|
| 784 |
{
|
| 785 |
"epoch": 5.35,
|
| 786 |
-
"learning_rate":
|
| 787 |
-
"loss": 0.
|
| 788 |
"step": 1140
|
| 789 |
},
|
| 790 |
{
|
| 791 |
"epoch": 5.4,
|
| 792 |
-
"learning_rate":
|
| 793 |
-
"loss": 0.
|
| 794 |
"step": 1150
|
| 795 |
},
|
| 796 |
{
|
| 797 |
"epoch": 5.45,
|
| 798 |
-
"learning_rate":
|
| 799 |
-
"loss": 0.
|
| 800 |
"step": 1160
|
| 801 |
},
|
| 802 |
{
|
| 803 |
"epoch": 5.49,
|
| 804 |
-
"learning_rate":
|
| 805 |
-
"loss": 0.
|
| 806 |
"step": 1170
|
| 807 |
},
|
| 808 |
{
|
| 809 |
"epoch": 5.54,
|
| 810 |
-
"learning_rate":
|
| 811 |
-
"loss": 0.
|
| 812 |
"step": 1180
|
| 813 |
},
|
| 814 |
{
|
| 815 |
"epoch": 5.59,
|
| 816 |
-
"learning_rate":
|
| 817 |
-
"loss": 0.
|
| 818 |
"step": 1190
|
| 819 |
},
|
| 820 |
{
|
| 821 |
"epoch": 5.63,
|
| 822 |
-
"learning_rate":
|
| 823 |
-
"loss": 0.
|
| 824 |
"step": 1200
|
| 825 |
},
|
| 826 |
{
|
| 827 |
"epoch": 5.68,
|
| 828 |
-
"learning_rate":
|
| 829 |
-
"loss": 0.
|
| 830 |
"step": 1210
|
| 831 |
},
|
| 832 |
{
|
| 833 |
"epoch": 5.73,
|
| 834 |
-
"learning_rate":
|
| 835 |
-
"loss": 0.
|
| 836 |
"step": 1220
|
| 837 |
},
|
| 838 |
{
|
| 839 |
"epoch": 5.77,
|
| 840 |
-
"learning_rate":
|
| 841 |
-
"loss": 0.
|
| 842 |
"step": 1230
|
| 843 |
},
|
| 844 |
{
|
| 845 |
"epoch": 5.82,
|
| 846 |
-
"learning_rate":
|
| 847 |
-
"loss": 0.
|
| 848 |
"step": 1240
|
| 849 |
},
|
| 850 |
{
|
| 851 |
"epoch": 5.87,
|
| 852 |
-
"learning_rate":
|
| 853 |
-
"loss": 0.
|
| 854 |
"step": 1250
|
| 855 |
},
|
| 856 |
{
|
| 857 |
"epoch": 5.92,
|
| 858 |
-
"learning_rate":
|
| 859 |
-
"loss": 0.
|
| 860 |
"step": 1260
|
| 861 |
},
|
| 862 |
{
|
| 863 |
"epoch": 5.96,
|
| 864 |
-
"learning_rate":
|
| 865 |
-
"loss": 0.
|
| 866 |
"step": 1270
|
| 867 |
},
|
| 868 |
{
|
| 869 |
"epoch": 6.0,
|
| 870 |
-
"eval_corporation_f1": 0.
|
| 871 |
-
"eval_creative-work_f1": 0.
|
| 872 |
-
"eval_group_f1": 0.
|
| 873 |
-
"eval_location_f1": 0.
|
| 874 |
-
"eval_loss": 0.
|
| 875 |
-
"eval_overall_accuracy": 0.
|
| 876 |
-
"eval_overall_f1": 0.
|
| 877 |
-
"eval_overall_precision": 0.
|
| 878 |
-
"eval_overall_recall": 0.
|
| 879 |
-
"eval_person_f1": 0.
|
| 880 |
-
"eval_product_f1": 0.
|
| 881 |
-
"eval_runtime":
|
| 882 |
-
"eval_samples_per_second":
|
| 883 |
-
"eval_steps_per_second":
|
| 884 |
"step": 1278
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 6.01,
|
| 888 |
-
"learning_rate":
|
| 889 |
-
"loss": 0.
|
| 890 |
"step": 1280
|
| 891 |
},
|
| 892 |
{
|
| 893 |
"epoch": 6.06,
|
| 894 |
-
"learning_rate":
|
| 895 |
-
"loss": 0.
|
| 896 |
"step": 1290
|
| 897 |
},
|
| 898 |
{
|
| 899 |
"epoch": 6.1,
|
| 900 |
-
"learning_rate":
|
| 901 |
-
"loss": 0.
|
| 902 |
"step": 1300
|
| 903 |
},
|
| 904 |
{
|
| 905 |
"epoch": 6.15,
|
| 906 |
-
"learning_rate":
|
| 907 |
-
"loss": 0.
|
| 908 |
"step": 1310
|
| 909 |
},
|
| 910 |
{
|
| 911 |
"epoch": 6.2,
|
| 912 |
-
"learning_rate":
|
| 913 |
-
"loss": 0.
|
| 914 |
"step": 1320
|
| 915 |
},
|
| 916 |
{
|
| 917 |
"epoch": 6.24,
|
| 918 |
-
"learning_rate":
|
| 919 |
-
"loss": 0.
|
| 920 |
"step": 1330
|
| 921 |
},
|
| 922 |
{
|
| 923 |
"epoch": 6.29,
|
| 924 |
-
"learning_rate":
|
| 925 |
-
"loss": 0.
|
| 926 |
"step": 1340
|
| 927 |
},
|
| 928 |
{
|
| 929 |
"epoch": 6.34,
|
| 930 |
-
"learning_rate":
|
| 931 |
-
"loss": 0.
|
| 932 |
"step": 1350
|
| 933 |
},
|
| 934 |
{
|
| 935 |
"epoch": 6.38,
|
| 936 |
-
"learning_rate":
|
| 937 |
-
"loss": 0.
|
| 938 |
"step": 1360
|
| 939 |
},
|
| 940 |
{
|
| 941 |
"epoch": 6.43,
|
| 942 |
-
"learning_rate":
|
| 943 |
-
"loss": 0.
|
| 944 |
"step": 1370
|
| 945 |
},
|
| 946 |
{
|
| 947 |
"epoch": 6.48,
|
| 948 |
-
"learning_rate":
|
| 949 |
-
"loss": 0.
|
| 950 |
"step": 1380
|
| 951 |
},
|
| 952 |
{
|
| 953 |
"epoch": 6.53,
|
| 954 |
-
"learning_rate":
|
| 955 |
-
"loss": 0.
|
| 956 |
"step": 1390
|
| 957 |
},
|
| 958 |
{
|
| 959 |
"epoch": 6.57,
|
| 960 |
-
"learning_rate":
|
| 961 |
-
"loss": 0.
|
| 962 |
"step": 1400
|
| 963 |
},
|
| 964 |
{
|
| 965 |
"epoch": 6.62,
|
| 966 |
-
"learning_rate":
|
| 967 |
-
"loss": 0.
|
| 968 |
"step": 1410
|
| 969 |
},
|
| 970 |
{
|
| 971 |
"epoch": 6.67,
|
| 972 |
-
"learning_rate":
|
| 973 |
-
"loss": 0.
|
| 974 |
"step": 1420
|
| 975 |
},
|
| 976 |
{
|
| 977 |
"epoch": 6.71,
|
| 978 |
-
"learning_rate":
|
| 979 |
-
"loss": 0.
|
| 980 |
"step": 1430
|
| 981 |
},
|
| 982 |
{
|
| 983 |
"epoch": 6.76,
|
| 984 |
-
"learning_rate":
|
| 985 |
-
"loss": 0.
|
| 986 |
"step": 1440
|
| 987 |
},
|
| 988 |
{
|
| 989 |
"epoch": 6.81,
|
| 990 |
-
"learning_rate":
|
| 991 |
-
"loss": 0.
|
| 992 |
"step": 1450
|
| 993 |
},
|
| 994 |
{
|
| 995 |
"epoch": 6.85,
|
| 996 |
-
"learning_rate":
|
| 997 |
-
"loss": 0.
|
| 998 |
"step": 1460
|
| 999 |
},
|
| 1000 |
{
|
| 1001 |
"epoch": 6.9,
|
| 1002 |
-
"learning_rate":
|
| 1003 |
-
"loss": 0.
|
| 1004 |
"step": 1470
|
| 1005 |
},
|
| 1006 |
{
|
| 1007 |
"epoch": 6.95,
|
| 1008 |
-
"learning_rate":
|
| 1009 |
-
"loss": 0.
|
| 1010 |
"step": 1480
|
| 1011 |
},
|
| 1012 |
{
|
| 1013 |
"epoch": 7.0,
|
| 1014 |
-
"learning_rate":
|
| 1015 |
-
"loss": 0.
|
| 1016 |
"step": 1490
|
| 1017 |
},
|
| 1018 |
{
|
| 1019 |
"epoch": 7.0,
|
| 1020 |
-
"eval_corporation_f1": 0.
|
| 1021 |
-
"eval_creative-work_f1": 0.
|
| 1022 |
-
"eval_group_f1": 0.
|
| 1023 |
-
"eval_location_f1": 0.
|
| 1024 |
-
"eval_loss": 0.
|
| 1025 |
-
"eval_overall_accuracy": 0.
|
| 1026 |
-
"eval_overall_f1": 0.
|
| 1027 |
-
"eval_overall_precision": 0.
|
| 1028 |
-
"eval_overall_recall": 0.
|
| 1029 |
-
"eval_person_f1": 0.
|
| 1030 |
-
"eval_product_f1": 0.
|
| 1031 |
-
"eval_runtime":
|
| 1032 |
-
"eval_samples_per_second":
|
| 1033 |
-
"eval_steps_per_second":
|
| 1034 |
"step": 1491
|
| 1035 |
},
|
| 1036 |
{
|
| 1037 |
"epoch": 7.04,
|
| 1038 |
-
"learning_rate":
|
| 1039 |
-
"loss": 0.
|
| 1040 |
"step": 1500
|
| 1041 |
},
|
| 1042 |
{
|
| 1043 |
"epoch": 7.09,
|
| 1044 |
-
"learning_rate":
|
| 1045 |
-
"loss": 0.
|
| 1046 |
"step": 1510
|
| 1047 |
},
|
| 1048 |
{
|
| 1049 |
"epoch": 7.14,
|
| 1050 |
-
"learning_rate":
|
| 1051 |
-
"loss": 0.
|
| 1052 |
"step": 1520
|
| 1053 |
},
|
| 1054 |
{
|
| 1055 |
"epoch": 7.18,
|
| 1056 |
-
"learning_rate":
|
| 1057 |
-
"loss": 0.
|
| 1058 |
"step": 1530
|
| 1059 |
},
|
| 1060 |
{
|
| 1061 |
"epoch": 7.23,
|
| 1062 |
-
"learning_rate":
|
| 1063 |
-
"loss": 0.
|
| 1064 |
"step": 1540
|
| 1065 |
},
|
| 1066 |
{
|
| 1067 |
"epoch": 7.28,
|
| 1068 |
-
"learning_rate":
|
| 1069 |
-
"loss": 0.
|
| 1070 |
"step": 1550
|
| 1071 |
},
|
| 1072 |
{
|
| 1073 |
"epoch": 7.32,
|
| 1074 |
-
"learning_rate":
|
| 1075 |
-
"loss": 0.
|
| 1076 |
"step": 1560
|
| 1077 |
},
|
| 1078 |
{
|
| 1079 |
"epoch": 7.37,
|
| 1080 |
-
"learning_rate":
|
| 1081 |
-
"loss": 0.
|
| 1082 |
"step": 1570
|
| 1083 |
},
|
| 1084 |
{
|
| 1085 |
"epoch": 7.42,
|
| 1086 |
-
"learning_rate":
|
| 1087 |
-
"loss": 0.
|
| 1088 |
"step": 1580
|
| 1089 |
},
|
| 1090 |
{
|
| 1091 |
"epoch": 7.46,
|
| 1092 |
-
"learning_rate":
|
| 1093 |
-
"loss": 0.
|
| 1094 |
"step": 1590
|
| 1095 |
},
|
| 1096 |
{
|
| 1097 |
"epoch": 7.51,
|
| 1098 |
-
"learning_rate":
|
| 1099 |
-
"loss": 0.
|
| 1100 |
"step": 1600
|
| 1101 |
},
|
| 1102 |
{
|
| 1103 |
"epoch": 7.56,
|
| 1104 |
-
"learning_rate":
|
| 1105 |
-
"loss": 0.
|
| 1106 |
"step": 1610
|
| 1107 |
},
|
| 1108 |
{
|
| 1109 |
"epoch": 7.61,
|
| 1110 |
-
"learning_rate":
|
| 1111 |
-
"loss": 0.
|
| 1112 |
"step": 1620
|
| 1113 |
},
|
| 1114 |
{
|
| 1115 |
"epoch": 7.65,
|
| 1116 |
-
"learning_rate":
|
| 1117 |
-
"loss": 0.
|
| 1118 |
"step": 1630
|
| 1119 |
},
|
| 1120 |
{
|
| 1121 |
"epoch": 7.7,
|
| 1122 |
-
"learning_rate":
|
| 1123 |
-
"loss": 0.
|
| 1124 |
"step": 1640
|
| 1125 |
},
|
| 1126 |
{
|
| 1127 |
"epoch": 7.75,
|
| 1128 |
-
"learning_rate":
|
| 1129 |
-
"loss": 0.
|
| 1130 |
"step": 1650
|
| 1131 |
},
|
| 1132 |
{
|
| 1133 |
"epoch": 7.79,
|
| 1134 |
-
"learning_rate":
|
| 1135 |
-
"loss": 0.
|
| 1136 |
"step": 1660
|
| 1137 |
},
|
| 1138 |
{
|
| 1139 |
"epoch": 7.84,
|
| 1140 |
-
"learning_rate":
|
| 1141 |
-
"loss": 0.
|
| 1142 |
"step": 1670
|
| 1143 |
},
|
| 1144 |
{
|
| 1145 |
"epoch": 7.89,
|
| 1146 |
-
"learning_rate":
|
| 1147 |
-
"loss": 0.
|
| 1148 |
"step": 1680
|
| 1149 |
},
|
| 1150 |
{
|
| 1151 |
"epoch": 7.93,
|
| 1152 |
-
"learning_rate":
|
| 1153 |
-
"loss": 0.
|
| 1154 |
"step": 1690
|
| 1155 |
},
|
| 1156 |
{
|
| 1157 |
"epoch": 7.98,
|
| 1158 |
-
"learning_rate":
|
| 1159 |
-
"loss": 0.
|
| 1160 |
"step": 1700
|
| 1161 |
},
|
| 1162 |
{
|
| 1163 |
"epoch": 8.0,
|
| 1164 |
-
"eval_corporation_f1": 0.
|
| 1165 |
-
"eval_creative-work_f1": 0.
|
| 1166 |
-
"eval_group_f1": 0.
|
| 1167 |
-
"eval_location_f1": 0.
|
| 1168 |
-
"eval_loss": 0.
|
| 1169 |
-
"eval_overall_accuracy": 0.
|
| 1170 |
-
"eval_overall_f1": 0.
|
| 1171 |
-
"eval_overall_precision": 0.
|
| 1172 |
-
"eval_overall_recall": 0.
|
| 1173 |
-
"eval_person_f1": 0.
|
| 1174 |
-
"eval_product_f1": 0.
|
| 1175 |
-
"eval_runtime": 6.
|
| 1176 |
-
"eval_samples_per_second":
|
| 1177 |
-
"eval_steps_per_second":
|
| 1178 |
"step": 1704
|
| 1179 |
},
|
| 1180 |
{
|
| 1181 |
"epoch": 8.03,
|
| 1182 |
-
"learning_rate":
|
| 1183 |
-
"loss": 0.
|
| 1184 |
"step": 1710
|
| 1185 |
},
|
| 1186 |
{
|
| 1187 |
"epoch": 8.08,
|
| 1188 |
-
"learning_rate":
|
| 1189 |
-
"loss": 0.
|
| 1190 |
"step": 1720
|
| 1191 |
},
|
| 1192 |
{
|
| 1193 |
"epoch": 8.12,
|
| 1194 |
-
"learning_rate":
|
| 1195 |
-
"loss": 0.
|
| 1196 |
"step": 1730
|
| 1197 |
},
|
| 1198 |
{
|
| 1199 |
"epoch": 8.17,
|
| 1200 |
-
"learning_rate":
|
| 1201 |
-
"loss": 0.
|
| 1202 |
"step": 1740
|
| 1203 |
},
|
| 1204 |
{
|
| 1205 |
"epoch": 8.22,
|
| 1206 |
-
"learning_rate":
|
| 1207 |
-
"loss": 0.
|
| 1208 |
"step": 1750
|
| 1209 |
},
|
| 1210 |
{
|
| 1211 |
"epoch": 8.26,
|
| 1212 |
-
"learning_rate":
|
| 1213 |
-
"loss": 0.
|
| 1214 |
"step": 1760
|
| 1215 |
},
|
| 1216 |
{
|
| 1217 |
"epoch": 8.31,
|
| 1218 |
-
"learning_rate":
|
| 1219 |
-
"loss": 0.
|
| 1220 |
"step": 1770
|
| 1221 |
},
|
| 1222 |
{
|
| 1223 |
"epoch": 8.36,
|
| 1224 |
-
"learning_rate":
|
| 1225 |
-
"loss": 0.
|
| 1226 |
"step": 1780
|
| 1227 |
},
|
| 1228 |
{
|
| 1229 |
"epoch": 8.4,
|
| 1230 |
-
"learning_rate":
|
| 1231 |
-
"loss": 0.
|
| 1232 |
"step": 1790
|
| 1233 |
},
|
| 1234 |
{
|
| 1235 |
"epoch": 8.45,
|
| 1236 |
-
"learning_rate":
|
| 1237 |
-
"loss": 0.
|
| 1238 |
"step": 1800
|
| 1239 |
},
|
| 1240 |
{
|
| 1241 |
"epoch": 8.5,
|
| 1242 |
-
"learning_rate":
|
| 1243 |
-
"loss": 0.
|
| 1244 |
"step": 1810
|
| 1245 |
},
|
| 1246 |
{
|
| 1247 |
"epoch": 8.54,
|
| 1248 |
-
"learning_rate":
|
| 1249 |
-
"loss": 0.
|
| 1250 |
"step": 1820
|
| 1251 |
},
|
| 1252 |
{
|
| 1253 |
"epoch": 8.59,
|
| 1254 |
-
"learning_rate":
|
| 1255 |
-
"loss": 0.
|
| 1256 |
"step": 1830
|
| 1257 |
},
|
| 1258 |
{
|
| 1259 |
"epoch": 8.64,
|
| 1260 |
-
"learning_rate":
|
| 1261 |
-
"loss": 0.
|
| 1262 |
"step": 1840
|
| 1263 |
},
|
| 1264 |
{
|
| 1265 |
"epoch": 8.69,
|
| 1266 |
-
"learning_rate":
|
| 1267 |
-
"loss": 0.
|
| 1268 |
"step": 1850
|
| 1269 |
},
|
| 1270 |
{
|
| 1271 |
"epoch": 8.73,
|
| 1272 |
-
"learning_rate":
|
| 1273 |
-
"loss": 0.
|
| 1274 |
"step": 1860
|
| 1275 |
},
|
| 1276 |
{
|
| 1277 |
"epoch": 8.78,
|
| 1278 |
-
"learning_rate":
|
| 1279 |
-
"loss": 0.
|
| 1280 |
"step": 1870
|
| 1281 |
},
|
| 1282 |
{
|
| 1283 |
"epoch": 8.83,
|
| 1284 |
-
"learning_rate":
|
| 1285 |
-
"loss": 0.
|
| 1286 |
"step": 1880
|
| 1287 |
},
|
| 1288 |
{
|
| 1289 |
"epoch": 8.87,
|
| 1290 |
-
"learning_rate":
|
| 1291 |
-
"loss": 0.
|
| 1292 |
"step": 1890
|
| 1293 |
},
|
| 1294 |
{
|
| 1295 |
"epoch": 8.92,
|
| 1296 |
-
"learning_rate":
|
| 1297 |
-
"loss": 0.
|
| 1298 |
"step": 1900
|
| 1299 |
},
|
| 1300 |
{
|
| 1301 |
"epoch": 8.97,
|
| 1302 |
-
"learning_rate":
|
| 1303 |
-
"loss": 0.
|
| 1304 |
"step": 1910
|
| 1305 |
},
|
| 1306 |
{
|
| 1307 |
"epoch": 9.0,
|
| 1308 |
-
"eval_corporation_f1": 0.
|
| 1309 |
-
"eval_creative-work_f1": 0.
|
| 1310 |
-
"eval_group_f1": 0.
|
| 1311 |
-
"eval_location_f1": 0.
|
| 1312 |
-
"eval_loss": 0.
|
| 1313 |
-
"eval_overall_accuracy": 0.
|
| 1314 |
-
"eval_overall_f1": 0.
|
| 1315 |
-
"eval_overall_precision": 0.
|
| 1316 |
-
"eval_overall_recall": 0.
|
| 1317 |
-
"eval_person_f1": 0.
|
| 1318 |
-
"eval_product_f1": 0.
|
| 1319 |
-
"eval_runtime":
|
| 1320 |
-
"eval_samples_per_second":
|
| 1321 |
-
"eval_steps_per_second":
|
| 1322 |
"step": 1917
|
| 1323 |
},
|
| 1324 |
{
|
| 1325 |
"epoch": 9.01,
|
| 1326 |
-
"learning_rate":
|
| 1327 |
-
"loss": 0.
|
| 1328 |
"step": 1920
|
| 1329 |
},
|
| 1330 |
{
|
| 1331 |
"epoch": 9.06,
|
| 1332 |
-
"learning_rate":
|
| 1333 |
-
"loss": 0.
|
| 1334 |
"step": 1930
|
| 1335 |
},
|
| 1336 |
{
|
| 1337 |
"epoch": 9.11,
|
| 1338 |
-
"learning_rate":
|
| 1339 |
-
"loss": 0.
|
| 1340 |
"step": 1940
|
| 1341 |
},
|
| 1342 |
{
|
| 1343 |
"epoch": 9.15,
|
| 1344 |
-
"learning_rate":
|
| 1345 |
-
"loss": 0.
|
| 1346 |
"step": 1950
|
| 1347 |
},
|
| 1348 |
{
|
| 1349 |
"epoch": 9.2,
|
| 1350 |
-
"learning_rate":
|
| 1351 |
-
"loss": 0.
|
| 1352 |
"step": 1960
|
| 1353 |
},
|
| 1354 |
{
|
| 1355 |
"epoch": 9.25,
|
| 1356 |
-
"learning_rate":
|
| 1357 |
-
"loss": 0.
|
| 1358 |
"step": 1970
|
| 1359 |
},
|
| 1360 |
{
|
| 1361 |
"epoch": 9.3,
|
| 1362 |
-
"learning_rate":
|
| 1363 |
-
"loss": 0.
|
| 1364 |
"step": 1980
|
| 1365 |
},
|
| 1366 |
{
|
| 1367 |
"epoch": 9.34,
|
| 1368 |
-
"learning_rate":
|
| 1369 |
-
"loss": 0.
|
| 1370 |
"step": 1990
|
| 1371 |
},
|
| 1372 |
{
|
| 1373 |
"epoch": 9.39,
|
| 1374 |
-
"learning_rate":
|
| 1375 |
-
"loss": 0.
|
| 1376 |
"step": 2000
|
| 1377 |
},
|
| 1378 |
{
|
| 1379 |
"epoch": 9.44,
|
| 1380 |
-
"learning_rate":
|
| 1381 |
-
"loss": 0.
|
| 1382 |
"step": 2010
|
| 1383 |
},
|
| 1384 |
{
|
| 1385 |
"epoch": 9.48,
|
| 1386 |
-
"learning_rate":
|
| 1387 |
-
"loss": 0.
|
| 1388 |
"step": 2020
|
| 1389 |
},
|
| 1390 |
{
|
| 1391 |
"epoch": 9.53,
|
| 1392 |
-
"learning_rate":
|
| 1393 |
-
"loss": 0.
|
| 1394 |
"step": 2030
|
| 1395 |
},
|
| 1396 |
{
|
| 1397 |
"epoch": 9.58,
|
| 1398 |
-
"learning_rate":
|
| 1399 |
-
"loss": 0.
|
| 1400 |
"step": 2040
|
| 1401 |
},
|
| 1402 |
{
|
| 1403 |
"epoch": 9.62,
|
| 1404 |
-
"learning_rate":
|
| 1405 |
-
"loss": 0.
|
| 1406 |
"step": 2050
|
| 1407 |
},
|
| 1408 |
{
|
| 1409 |
"epoch": 9.67,
|
| 1410 |
-
"learning_rate":
|
| 1411 |
-
"loss": 0.
|
| 1412 |
"step": 2060
|
| 1413 |
},
|
| 1414 |
{
|
| 1415 |
"epoch": 9.72,
|
| 1416 |
-
"learning_rate":
|
| 1417 |
-
"loss": 0.
|
| 1418 |
"step": 2070
|
| 1419 |
},
|
| 1420 |
{
|
| 1421 |
"epoch": 9.77,
|
| 1422 |
-
"learning_rate":
|
| 1423 |
-
"loss": 0.
|
| 1424 |
"step": 2080
|
| 1425 |
},
|
| 1426 |
{
|
| 1427 |
"epoch": 9.81,
|
| 1428 |
-
"learning_rate":
|
| 1429 |
-
"loss": 0.
|
| 1430 |
"step": 2090
|
| 1431 |
},
|
| 1432 |
{
|
| 1433 |
"epoch": 9.86,
|
| 1434 |
-
"learning_rate":
|
| 1435 |
-
"loss": 0.
|
| 1436 |
"step": 2100
|
| 1437 |
},
|
| 1438 |
{
|
| 1439 |
"epoch": 9.91,
|
| 1440 |
-
"learning_rate":
|
| 1441 |
-
"loss": 0.
|
| 1442 |
"step": 2110
|
| 1443 |
},
|
| 1444 |
{
|
| 1445 |
"epoch": 9.95,
|
| 1446 |
-
"learning_rate":
|
| 1447 |
-
"loss": 0.
|
| 1448 |
"step": 2120
|
| 1449 |
},
|
| 1450 |
{
|
| 1451 |
"epoch": 10.0,
|
| 1452 |
-
"learning_rate":
|
| 1453 |
-
"loss": 0.
|
| 1454 |
"step": 2130
|
| 1455 |
},
|
| 1456 |
{
|
| 1457 |
"epoch": 10.0,
|
| 1458 |
-
"eval_corporation_f1": 0.
|
| 1459 |
-
"eval_creative-work_f1": 0.
|
| 1460 |
-
"eval_group_f1": 0.
|
| 1461 |
-
"eval_location_f1": 0.
|
| 1462 |
-
"eval_loss": 0.
|
| 1463 |
-
"eval_overall_accuracy": 0.
|
| 1464 |
-
"eval_overall_f1": 0.
|
| 1465 |
-
"eval_overall_precision": 0.
|
| 1466 |
-
"eval_overall_recall": 0.
|
| 1467 |
-
"eval_person_f1": 0.
|
| 1468 |
-
"eval_product_f1": 0.
|
| 1469 |
-
"eval_runtime":
|
| 1470 |
-
"eval_samples_per_second":
|
| 1471 |
-
"eval_steps_per_second":
|
| 1472 |
"step": 2130
|
| 1473 |
},
|
| 1474 |
{
|
| 1475 |
"epoch": 10.05,
|
| 1476 |
-
"learning_rate":
|
| 1477 |
-
"loss": 0.
|
| 1478 |
"step": 2140
|
| 1479 |
},
|
| 1480 |
{
|
| 1481 |
"epoch": 10.09,
|
| 1482 |
-
"learning_rate":
|
| 1483 |
-
"loss": 0.
|
| 1484 |
"step": 2150
|
| 1485 |
},
|
| 1486 |
{
|
| 1487 |
"epoch": 10.14,
|
| 1488 |
-
"learning_rate":
|
| 1489 |
-
"loss": 0.
|
| 1490 |
"step": 2160
|
| 1491 |
},
|
| 1492 |
{
|
| 1493 |
"epoch": 10.19,
|
| 1494 |
-
"learning_rate":
|
| 1495 |
-
"loss": 0.
|
| 1496 |
"step": 2170
|
| 1497 |
},
|
| 1498 |
{
|
| 1499 |
"epoch": 10.23,
|
| 1500 |
-
"learning_rate":
|
| 1501 |
-
"loss": 0.
|
| 1502 |
"step": 2180
|
| 1503 |
},
|
| 1504 |
{
|
| 1505 |
"epoch": 10.28,
|
| 1506 |
-
"learning_rate":
|
| 1507 |
-
"loss": 0.
|
| 1508 |
"step": 2190
|
| 1509 |
},
|
| 1510 |
{
|
| 1511 |
"epoch": 10.33,
|
| 1512 |
-
"learning_rate":
|
| 1513 |
-
"loss": 0.
|
| 1514 |
"step": 2200
|
| 1515 |
},
|
| 1516 |
{
|
| 1517 |
"epoch": 10.38,
|
| 1518 |
-
"learning_rate":
|
| 1519 |
-
"loss": 0.
|
| 1520 |
"step": 2210
|
| 1521 |
},
|
| 1522 |
{
|
| 1523 |
"epoch": 10.42,
|
| 1524 |
-
"learning_rate":
|
| 1525 |
-
"loss": 0.
|
| 1526 |
"step": 2220
|
| 1527 |
},
|
| 1528 |
{
|
| 1529 |
"epoch": 10.47,
|
| 1530 |
-
"learning_rate":
|
| 1531 |
-
"loss": 0.
|
| 1532 |
"step": 2230
|
| 1533 |
},
|
| 1534 |
{
|
| 1535 |
"epoch": 10.52,
|
| 1536 |
-
"learning_rate":
|
| 1537 |
-
"loss": 0.
|
| 1538 |
"step": 2240
|
| 1539 |
},
|
| 1540 |
{
|
| 1541 |
"epoch": 10.56,
|
| 1542 |
-
"learning_rate":
|
| 1543 |
-
"loss": 0.
|
| 1544 |
"step": 2250
|
| 1545 |
},
|
| 1546 |
{
|
| 1547 |
"epoch": 10.61,
|
| 1548 |
-
"learning_rate":
|
| 1549 |
-
"loss": 0.
|
| 1550 |
"step": 2260
|
| 1551 |
},
|
| 1552 |
{
|
| 1553 |
"epoch": 10.66,
|
| 1554 |
-
"learning_rate":
|
| 1555 |
-
"loss": 0.
|
| 1556 |
"step": 2270
|
| 1557 |
},
|
| 1558 |
{
|
| 1559 |
"epoch": 10.7,
|
| 1560 |
-
"learning_rate":
|
| 1561 |
-
"loss": 0.
|
| 1562 |
"step": 2280
|
| 1563 |
},
|
| 1564 |
{
|
| 1565 |
"epoch": 10.75,
|
| 1566 |
-
"learning_rate":
|
| 1567 |
-
"loss": 0.
|
| 1568 |
"step": 2290
|
| 1569 |
},
|
| 1570 |
{
|
| 1571 |
"epoch": 10.8,
|
| 1572 |
-
"learning_rate":
|
| 1573 |
-
"loss": 0.
|
| 1574 |
"step": 2300
|
| 1575 |
},
|
| 1576 |
{
|
| 1577 |
"epoch": 10.85,
|
| 1578 |
-
"learning_rate":
|
| 1579 |
-
"loss": 0.
|
| 1580 |
"step": 2310
|
| 1581 |
},
|
| 1582 |
{
|
| 1583 |
"epoch": 10.89,
|
| 1584 |
-
"learning_rate":
|
| 1585 |
-
"loss": 0.
|
| 1586 |
"step": 2320
|
| 1587 |
},
|
| 1588 |
{
|
| 1589 |
"epoch": 10.94,
|
| 1590 |
-
"learning_rate":
|
| 1591 |
-
"loss": 0.
|
| 1592 |
"step": 2330
|
| 1593 |
},
|
| 1594 |
{
|
| 1595 |
"epoch": 10.99,
|
| 1596 |
-
"learning_rate":
|
| 1597 |
-
"loss": 0.
|
| 1598 |
"step": 2340
|
| 1599 |
},
|
| 1600 |
{
|
| 1601 |
"epoch": 11.0,
|
| 1602 |
-
"eval_corporation_f1": 0.
|
| 1603 |
-
"eval_creative-work_f1": 0.
|
| 1604 |
-
"eval_group_f1": 0.
|
| 1605 |
-
"eval_location_f1": 0.
|
| 1606 |
-
"eval_loss": 0.
|
| 1607 |
-
"eval_overall_accuracy": 0.
|
| 1608 |
-
"eval_overall_f1": 0.
|
| 1609 |
-
"eval_overall_precision": 0.
|
| 1610 |
-
"eval_overall_recall": 0.
|
| 1611 |
-
"eval_person_f1": 0.
|
| 1612 |
-
"eval_product_f1": 0.
|
| 1613 |
-
"eval_runtime":
|
| 1614 |
-
"eval_samples_per_second":
|
| 1615 |
-
"eval_steps_per_second":
|
| 1616 |
"step": 2343
|
| 1617 |
},
|
| 1618 |
{
|
| 1619 |
"epoch": 11.03,
|
| 1620 |
-
"learning_rate":
|
| 1621 |
-
"loss": 0.
|
| 1622 |
"step": 2350
|
| 1623 |
},
|
| 1624 |
{
|
| 1625 |
"epoch": 11.08,
|
| 1626 |
-
"learning_rate":
|
| 1627 |
-
"loss": 0.
|
| 1628 |
"step": 2360
|
| 1629 |
},
|
| 1630 |
{
|
| 1631 |
"epoch": 11.13,
|
| 1632 |
-
"learning_rate":
|
| 1633 |
-
"loss": 0.
|
| 1634 |
"step": 2370
|
| 1635 |
},
|
| 1636 |
{
|
| 1637 |
"epoch": 11.17,
|
| 1638 |
-
"learning_rate":
|
| 1639 |
-
"loss": 0.
|
| 1640 |
"step": 2380
|
| 1641 |
},
|
| 1642 |
{
|
| 1643 |
"epoch": 11.22,
|
| 1644 |
-
"learning_rate":
|
| 1645 |
-
"loss": 0.
|
| 1646 |
"step": 2390
|
| 1647 |
},
|
| 1648 |
{
|
| 1649 |
"epoch": 11.27,
|
| 1650 |
-
"learning_rate":
|
| 1651 |
-
"loss": 0.
|
| 1652 |
"step": 2400
|
| 1653 |
},
|
| 1654 |
{
|
| 1655 |
"epoch": 11.31,
|
| 1656 |
-
"learning_rate":
|
| 1657 |
-
"loss": 0.
|
| 1658 |
"step": 2410
|
| 1659 |
},
|
| 1660 |
{
|
| 1661 |
"epoch": 11.36,
|
| 1662 |
-
"learning_rate":
|
| 1663 |
-
"loss": 0.
|
| 1664 |
"step": 2420
|
| 1665 |
},
|
| 1666 |
{
|
| 1667 |
"epoch": 11.41,
|
| 1668 |
-
"learning_rate":
|
| 1669 |
-
"loss": 0.
|
| 1670 |
"step": 2430
|
| 1671 |
},
|
| 1672 |
{
|
| 1673 |
"epoch": 11.46,
|
| 1674 |
-
"learning_rate":
|
| 1675 |
-
"loss": 0.
|
| 1676 |
"step": 2440
|
| 1677 |
},
|
| 1678 |
{
|
| 1679 |
"epoch": 11.5,
|
| 1680 |
-
"learning_rate":
|
| 1681 |
-
"loss": 0.
|
| 1682 |
"step": 2450
|
| 1683 |
},
|
| 1684 |
{
|
| 1685 |
"epoch": 11.55,
|
| 1686 |
-
"learning_rate":
|
| 1687 |
-
"loss": 0.
|
| 1688 |
"step": 2460
|
| 1689 |
},
|
| 1690 |
{
|
| 1691 |
"epoch": 11.6,
|
| 1692 |
-
"learning_rate":
|
| 1693 |
-
"loss": 0.
|
| 1694 |
"step": 2470
|
| 1695 |
},
|
| 1696 |
{
|
| 1697 |
"epoch": 11.64,
|
| 1698 |
-
"learning_rate":
|
| 1699 |
-
"loss": 0.
|
| 1700 |
"step": 2480
|
| 1701 |
},
|
| 1702 |
{
|
| 1703 |
"epoch": 11.69,
|
| 1704 |
-
"learning_rate":
|
| 1705 |
-
"loss": 0.
|
| 1706 |
"step": 2490
|
| 1707 |
},
|
| 1708 |
{
|
| 1709 |
"epoch": 11.74,
|
| 1710 |
-
"learning_rate":
|
| 1711 |
-
"loss": 0.
|
| 1712 |
"step": 2500
|
| 1713 |
},
|
| 1714 |
{
|
| 1715 |
"epoch": 11.78,
|
| 1716 |
-
"learning_rate":
|
| 1717 |
-
"loss": 0.
|
| 1718 |
"step": 2510
|
| 1719 |
},
|
| 1720 |
{
|
| 1721 |
"epoch": 11.83,
|
| 1722 |
-
"learning_rate":
|
| 1723 |
-
"loss": 0.
|
| 1724 |
"step": 2520
|
| 1725 |
},
|
| 1726 |
{
|
| 1727 |
"epoch": 11.88,
|
| 1728 |
-
"learning_rate":
|
| 1729 |
-
"loss": 0.
|
| 1730 |
"step": 2530
|
| 1731 |
},
|
| 1732 |
{
|
| 1733 |
"epoch": 11.92,
|
| 1734 |
-
"learning_rate":
|
| 1735 |
-
"loss": 0.
|
| 1736 |
"step": 2540
|
| 1737 |
},
|
| 1738 |
{
|
| 1739 |
"epoch": 11.97,
|
| 1740 |
-
"learning_rate":
|
| 1741 |
-
"loss": 0.
|
| 1742 |
"step": 2550
|
| 1743 |
},
|
| 1744 |
{
|
| 1745 |
"epoch": 12.0,
|
| 1746 |
-
"eval_corporation_f1": 0.
|
| 1747 |
-
"eval_creative-work_f1": 0.
|
| 1748 |
-
"eval_group_f1": 0.
|
| 1749 |
-
"eval_location_f1": 0.
|
| 1750 |
-
"eval_loss": 0.
|
| 1751 |
-
"eval_overall_accuracy": 0.
|
| 1752 |
-
"eval_overall_f1": 0.
|
| 1753 |
-
"eval_overall_precision": 0.
|
| 1754 |
-
"eval_overall_recall": 0.
|
| 1755 |
-
"eval_person_f1": 0.
|
| 1756 |
-
"eval_product_f1": 0.
|
| 1757 |
-
"eval_runtime":
|
| 1758 |
-
"eval_samples_per_second":
|
| 1759 |
-
"eval_steps_per_second":
|
| 1760 |
"step": 2556
|
| 1761 |
},
|
| 1762 |
{
|
| 1763 |
"epoch": 12.02,
|
| 1764 |
-
"learning_rate":
|
| 1765 |
-
"loss": 0.
|
| 1766 |
"step": 2560
|
| 1767 |
},
|
| 1768 |
{
|
| 1769 |
"epoch": 12.07,
|
| 1770 |
-
"learning_rate":
|
| 1771 |
-
"loss": 0.
|
| 1772 |
"step": 2570
|
| 1773 |
},
|
| 1774 |
{
|
| 1775 |
"epoch": 12.11,
|
| 1776 |
-
"learning_rate":
|
| 1777 |
-
"loss": 0.
|
| 1778 |
"step": 2580
|
| 1779 |
},
|
| 1780 |
{
|
| 1781 |
"epoch": 12.16,
|
| 1782 |
-
"learning_rate":
|
| 1783 |
-
"loss": 0.
|
| 1784 |
"step": 2590
|
| 1785 |
},
|
| 1786 |
{
|
| 1787 |
"epoch": 12.21,
|
| 1788 |
-
"learning_rate":
|
| 1789 |
-
"loss": 0.
|
| 1790 |
"step": 2600
|
| 1791 |
},
|
| 1792 |
{
|
| 1793 |
"epoch": 12.25,
|
| 1794 |
-
"learning_rate":
|
| 1795 |
-
"loss": 0.
|
| 1796 |
"step": 2610
|
| 1797 |
},
|
| 1798 |
{
|
| 1799 |
"epoch": 12.3,
|
| 1800 |
-
"learning_rate":
|
| 1801 |
-
"loss": 0.
|
| 1802 |
"step": 2620
|
| 1803 |
},
|
| 1804 |
{
|
| 1805 |
"epoch": 12.35,
|
| 1806 |
-
"learning_rate":
|
| 1807 |
-
"loss": 0.
|
| 1808 |
"step": 2630
|
| 1809 |
},
|
| 1810 |
{
|
| 1811 |
"epoch": 12.39,
|
| 1812 |
-
"learning_rate":
|
| 1813 |
-
"loss": 0.
|
| 1814 |
"step": 2640
|
| 1815 |
},
|
| 1816 |
{
|
| 1817 |
"epoch": 12.44,
|
| 1818 |
-
"learning_rate":
|
| 1819 |
-
"loss": 0.
|
| 1820 |
"step": 2650
|
| 1821 |
},
|
| 1822 |
{
|
| 1823 |
"epoch": 12.49,
|
| 1824 |
-
"learning_rate":
|
| 1825 |
-
"loss": 0.
|
| 1826 |
"step": 2660
|
| 1827 |
},
|
| 1828 |
{
|
| 1829 |
"epoch": 12.54,
|
| 1830 |
-
"learning_rate":
|
| 1831 |
-
"loss": 0.
|
| 1832 |
"step": 2670
|
| 1833 |
},
|
| 1834 |
{
|
| 1835 |
"epoch": 12.58,
|
| 1836 |
-
"learning_rate":
|
| 1837 |
-
"loss": 0.
|
| 1838 |
"step": 2680
|
| 1839 |
},
|
| 1840 |
{
|
| 1841 |
"epoch": 12.63,
|
| 1842 |
-
"learning_rate":
|
| 1843 |
-
"loss": 0.
|
| 1844 |
"step": 2690
|
| 1845 |
},
|
| 1846 |
{
|
| 1847 |
"epoch": 12.68,
|
| 1848 |
-
"learning_rate":
|
| 1849 |
-
"loss": 0.
|
| 1850 |
"step": 2700
|
| 1851 |
},
|
| 1852 |
{
|
| 1853 |
"epoch": 12.72,
|
| 1854 |
-
"learning_rate":
|
| 1855 |
-
"loss": 0.
|
| 1856 |
"step": 2710
|
| 1857 |
},
|
| 1858 |
{
|
| 1859 |
"epoch": 12.77,
|
| 1860 |
-
"learning_rate":
|
| 1861 |
-
"loss": 0.
|
| 1862 |
"step": 2720
|
| 1863 |
},
|
| 1864 |
{
|
| 1865 |
"epoch": 12.82,
|
| 1866 |
-
"learning_rate":
|
| 1867 |
-
"loss": 0.
|
| 1868 |
"step": 2730
|
| 1869 |
},
|
| 1870 |
{
|
| 1871 |
"epoch": 12.86,
|
| 1872 |
-
"learning_rate":
|
| 1873 |
-
"loss": 0.
|
| 1874 |
"step": 2740
|
| 1875 |
},
|
| 1876 |
{
|
| 1877 |
"epoch": 12.91,
|
| 1878 |
-
"learning_rate":
|
| 1879 |
-
"loss": 0.
|
| 1880 |
"step": 2750
|
| 1881 |
},
|
| 1882 |
{
|
| 1883 |
"epoch": 12.96,
|
| 1884 |
-
"learning_rate":
|
| 1885 |
-
"loss": 0.
|
| 1886 |
"step": 2760
|
| 1887 |
},
|
| 1888 |
{
|
| 1889 |
"epoch": 13.0,
|
| 1890 |
-
"eval_corporation_f1": 0.
|
| 1891 |
-
"eval_creative-work_f1": 0.
|
| 1892 |
-
"eval_group_f1": 0.
|
| 1893 |
-
"eval_location_f1": 0.
|
| 1894 |
-
"eval_loss": 0.
|
| 1895 |
-
"eval_overall_accuracy": 0.
|
| 1896 |
-
"eval_overall_f1": 0.
|
| 1897 |
-
"eval_overall_precision": 0.
|
| 1898 |
-
"eval_overall_recall": 0.
|
| 1899 |
-
"eval_person_f1": 0.
|
| 1900 |
-
"eval_product_f1": 0.
|
| 1901 |
-
"eval_runtime":
|
| 1902 |
-
"eval_samples_per_second":
|
| 1903 |
-
"eval_steps_per_second":
|
| 1904 |
"step": 2769
|
| 1905 |
},
|
| 1906 |
{
|
| 1907 |
"epoch": 13.0,
|
| 1908 |
-
"learning_rate":
|
| 1909 |
-
"loss": 0.
|
| 1910 |
"step": 2770
|
| 1911 |
},
|
| 1912 |
{
|
| 1913 |
"epoch": 13.05,
|
| 1914 |
-
"learning_rate":
|
| 1915 |
-
"loss": 0.
|
| 1916 |
"step": 2780
|
| 1917 |
},
|
| 1918 |
{
|
| 1919 |
"epoch": 13.1,
|
| 1920 |
-
"learning_rate":
|
| 1921 |
-
"loss": 0.
|
| 1922 |
"step": 2790
|
| 1923 |
},
|
| 1924 |
{
|
| 1925 |
"epoch": 13.15,
|
| 1926 |
-
"learning_rate":
|
| 1927 |
-
"loss": 0.
|
| 1928 |
"step": 2800
|
| 1929 |
},
|
| 1930 |
{
|
| 1931 |
"epoch": 13.19,
|
| 1932 |
-
"learning_rate":
|
| 1933 |
-
"loss": 0.
|
| 1934 |
"step": 2810
|
| 1935 |
},
|
| 1936 |
{
|
| 1937 |
"epoch": 13.24,
|
| 1938 |
-
"learning_rate":
|
| 1939 |
-
"loss": 0.
|
| 1940 |
"step": 2820
|
| 1941 |
},
|
| 1942 |
{
|
| 1943 |
"epoch": 13.29,
|
| 1944 |
-
"learning_rate":
|
| 1945 |
-
"loss": 0.
|
| 1946 |
"step": 2830
|
| 1947 |
},
|
| 1948 |
{
|
| 1949 |
"epoch": 13.33,
|
| 1950 |
-
"learning_rate":
|
| 1951 |
-
"loss": 0.
|
| 1952 |
"step": 2840
|
| 1953 |
},
|
| 1954 |
{
|
| 1955 |
"epoch": 13.38,
|
| 1956 |
-
"learning_rate":
|
| 1957 |
-
"loss": 0.
|
| 1958 |
"step": 2850
|
| 1959 |
},
|
| 1960 |
{
|
| 1961 |
"epoch": 13.43,
|
| 1962 |
-
"learning_rate":
|
| 1963 |
-
"loss": 0.
|
| 1964 |
"step": 2860
|
| 1965 |
},
|
| 1966 |
{
|
| 1967 |
"epoch": 13.47,
|
| 1968 |
-
"learning_rate":
|
| 1969 |
-
"loss": 0.
|
| 1970 |
"step": 2870
|
| 1971 |
},
|
| 1972 |
{
|
| 1973 |
"epoch": 13.52,
|
| 1974 |
-
"learning_rate":
|
| 1975 |
-
"loss": 0.
|
| 1976 |
"step": 2880
|
| 1977 |
},
|
| 1978 |
{
|
| 1979 |
"epoch": 13.57,
|
| 1980 |
-
"learning_rate":
|
| 1981 |
-
"loss": 0.
|
| 1982 |
"step": 2890
|
| 1983 |
},
|
| 1984 |
{
|
| 1985 |
"epoch": 13.62,
|
| 1986 |
-
"learning_rate":
|
| 1987 |
-
"loss": 0.
|
| 1988 |
"step": 2900
|
| 1989 |
},
|
| 1990 |
{
|
| 1991 |
"epoch": 13.66,
|
| 1992 |
-
"learning_rate":
|
| 1993 |
-
"loss": 0.
|
| 1994 |
"step": 2910
|
| 1995 |
},
|
| 1996 |
{
|
| 1997 |
"epoch": 13.71,
|
| 1998 |
-
"learning_rate":
|
| 1999 |
-
"loss": 0.
|
| 2000 |
"step": 2920
|
| 2001 |
},
|
| 2002 |
{
|
| 2003 |
"epoch": 13.76,
|
| 2004 |
-
"learning_rate":
|
| 2005 |
-
"loss": 0.
|
| 2006 |
"step": 2930
|
| 2007 |
},
|
| 2008 |
{
|
| 2009 |
"epoch": 13.8,
|
| 2010 |
-
"learning_rate":
|
| 2011 |
-
"loss": 0.
|
| 2012 |
"step": 2940
|
| 2013 |
},
|
| 2014 |
{
|
| 2015 |
"epoch": 13.85,
|
| 2016 |
-
"learning_rate":
|
| 2017 |
-
"loss": 0.
|
| 2018 |
"step": 2950
|
| 2019 |
},
|
| 2020 |
{
|
| 2021 |
"epoch": 13.9,
|
| 2022 |
-
"learning_rate":
|
| 2023 |
-
"loss": 0.
|
| 2024 |
"step": 2960
|
| 2025 |
},
|
| 2026 |
{
|
| 2027 |
"epoch": 13.94,
|
| 2028 |
-
"learning_rate":
|
| 2029 |
-
"loss": 0.
|
| 2030 |
"step": 2970
|
| 2031 |
},
|
| 2032 |
{
|
| 2033 |
"epoch": 13.99,
|
| 2034 |
-
"learning_rate":
|
| 2035 |
-
"loss": 0.
|
| 2036 |
"step": 2980
|
| 2037 |
},
|
| 2038 |
{
|
| 2039 |
"epoch": 14.0,
|
| 2040 |
-
"eval_corporation_f1": 0.
|
| 2041 |
-
"eval_creative-work_f1": 0.
|
| 2042 |
-
"eval_group_f1": 0.
|
| 2043 |
-
"eval_location_f1": 0.
|
| 2044 |
-
"eval_loss": 0.
|
| 2045 |
-
"eval_overall_accuracy": 0.
|
| 2046 |
-
"eval_overall_f1": 0.
|
| 2047 |
-
"eval_overall_precision": 0.
|
| 2048 |
-
"eval_overall_recall": 0.
|
| 2049 |
-
"eval_person_f1": 0.
|
| 2050 |
-
"eval_product_f1": 0.
|
| 2051 |
-
"eval_runtime":
|
| 2052 |
-
"eval_samples_per_second":
|
| 2053 |
-
"eval_steps_per_second":
|
| 2054 |
"step": 2982
|
| 2055 |
},
|
| 2056 |
{
|
| 2057 |
-
"epoch": 14.
|
| 2058 |
-
"
|
| 2059 |
-
"
|
| 2060 |
-
"
|
| 2061 |
-
|
| 2062 |
-
|
| 2063 |
-
"
|
| 2064 |
-
"learning_rate": 1.7183098591549297e-05,
|
| 2065 |
-
"loss": 0.0072,
|
| 2066 |
-
"step": 3000
|
| 2067 |
-
},
|
| 2068 |
-
{
|
| 2069 |
-
"epoch": 14.13,
|
| 2070 |
-
"learning_rate": 1.7173708920187795e-05,
|
| 2071 |
-
"loss": 0.0081,
|
| 2072 |
-
"step": 3010
|
| 2073 |
-
},
|
| 2074 |
-
{
|
| 2075 |
-
"epoch": 14.18,
|
| 2076 |
-
"learning_rate": 1.7164319248826292e-05,
|
| 2077 |
-
"loss": 0.0046,
|
| 2078 |
-
"step": 3020
|
| 2079 |
-
},
|
| 2080 |
-
{
|
| 2081 |
-
"epoch": 14.23,
|
| 2082 |
-
"learning_rate": 1.715492957746479e-05,
|
| 2083 |
-
"loss": 0.0088,
|
| 2084 |
-
"step": 3030
|
| 2085 |
-
},
|
| 2086 |
-
{
|
| 2087 |
-
"epoch": 14.27,
|
| 2088 |
-
"learning_rate": 1.7145539906103287e-05,
|
| 2089 |
-
"loss": 0.0047,
|
| 2090 |
-
"step": 3040
|
| 2091 |
-
},
|
| 2092 |
-
{
|
| 2093 |
-
"epoch": 14.32,
|
| 2094 |
-
"learning_rate": 1.7136150234741785e-05,
|
| 2095 |
-
"loss": 0.019,
|
| 2096 |
-
"step": 3050
|
| 2097 |
-
},
|
| 2098 |
-
{
|
| 2099 |
-
"epoch": 14.37,
|
| 2100 |
-
"learning_rate": 1.7126760563380282e-05,
|
| 2101 |
-
"loss": 0.0099,
|
| 2102 |
-
"step": 3060
|
| 2103 |
-
},
|
| 2104 |
-
{
|
| 2105 |
-
"epoch": 14.41,
|
| 2106 |
-
"learning_rate": 1.711737089201878e-05,
|
| 2107 |
-
"loss": 0.0038,
|
| 2108 |
-
"step": 3070
|
| 2109 |
-
},
|
| 2110 |
-
{
|
| 2111 |
-
"epoch": 14.46,
|
| 2112 |
-
"learning_rate": 1.7107981220657278e-05,
|
| 2113 |
-
"loss": 0.0114,
|
| 2114 |
-
"step": 3080
|
| 2115 |
-
},
|
| 2116 |
-
{
|
| 2117 |
-
"epoch": 14.51,
|
| 2118 |
-
"learning_rate": 1.7098591549295775e-05,
|
| 2119 |
-
"loss": 0.0146,
|
| 2120 |
-
"step": 3090
|
| 2121 |
-
},
|
| 2122 |
-
{
|
| 2123 |
-
"epoch": 14.55,
|
| 2124 |
-
"learning_rate": 1.7089201877934273e-05,
|
| 2125 |
-
"loss": 0.0117,
|
| 2126 |
-
"step": 3100
|
| 2127 |
-
},
|
| 2128 |
-
{
|
| 2129 |
-
"epoch": 14.6,
|
| 2130 |
-
"learning_rate": 1.7079812206572774e-05,
|
| 2131 |
-
"loss": 0.0056,
|
| 2132 |
-
"step": 3110
|
| 2133 |
-
},
|
| 2134 |
-
{
|
| 2135 |
-
"epoch": 14.65,
|
| 2136 |
-
"learning_rate": 1.7070422535211268e-05,
|
| 2137 |
-
"loss": 0.0059,
|
| 2138 |
-
"step": 3120
|
| 2139 |
-
},
|
| 2140 |
-
{
|
| 2141 |
-
"epoch": 14.69,
|
| 2142 |
-
"learning_rate": 1.7061032863849766e-05,
|
| 2143 |
-
"loss": 0.0122,
|
| 2144 |
-
"step": 3130
|
| 2145 |
-
},
|
| 2146 |
-
{
|
| 2147 |
-
"epoch": 14.74,
|
| 2148 |
-
"learning_rate": 1.7051643192488267e-05,
|
| 2149 |
-
"loss": 0.006,
|
| 2150 |
-
"step": 3140
|
| 2151 |
-
},
|
| 2152 |
-
{
|
| 2153 |
-
"epoch": 14.79,
|
| 2154 |
-
"learning_rate": 1.704225352112676e-05,
|
| 2155 |
-
"loss": 0.0115,
|
| 2156 |
-
"step": 3150
|
| 2157 |
-
},
|
| 2158 |
-
{
|
| 2159 |
-
"epoch": 14.84,
|
| 2160 |
-
"learning_rate": 1.703286384976526e-05,
|
| 2161 |
-
"loss": 0.0037,
|
| 2162 |
-
"step": 3160
|
| 2163 |
-
},
|
| 2164 |
-
{
|
| 2165 |
-
"epoch": 14.88,
|
| 2166 |
-
"learning_rate": 1.702347417840376e-05,
|
| 2167 |
-
"loss": 0.0207,
|
| 2168 |
-
"step": 3170
|
| 2169 |
-
},
|
| 2170 |
-
{
|
| 2171 |
-
"epoch": 14.93,
|
| 2172 |
-
"learning_rate": 1.7014084507042253e-05,
|
| 2173 |
-
"loss": 0.0102,
|
| 2174 |
-
"step": 3180
|
| 2175 |
-
},
|
| 2176 |
-
{
|
| 2177 |
-
"epoch": 14.98,
|
| 2178 |
-
"learning_rate": 1.7004694835680754e-05,
|
| 2179 |
-
"loss": 0.0075,
|
| 2180 |
-
"step": 3190
|
| 2181 |
-
},
|
| 2182 |
-
{
|
| 2183 |
-
"epoch": 15.0,
|
| 2184 |
-
"eval_corporation_f1": 0.27027027027027023,
|
| 2185 |
-
"eval_creative-work_f1": 0.36363636363636365,
|
| 2186 |
-
"eval_group_f1": 0.4029850746268656,
|
| 2187 |
-
"eval_location_f1": 0.7499999999999999,
|
| 2188 |
-
"eval_loss": 0.3238934278488159,
|
| 2189 |
-
"eval_overall_accuracy": 0.9498870526691238,
|
| 2190 |
-
"eval_overall_f1": 0.6374216651745748,
|
| 2191 |
-
"eval_overall_precision": 0.6912621359223301,
|
| 2192 |
-
"eval_overall_recall": 0.5913621262458472,
|
| 2193 |
-
"eval_person_f1": 0.7732919254658385,
|
| 2194 |
-
"eval_product_f1": 0.4152249134948097,
|
| 2195 |
-
"eval_runtime": 2.1734,
|
| 2196 |
-
"eval_samples_per_second": 464.252,
|
| 2197 |
-
"eval_steps_per_second": 29.447,
|
| 2198 |
-
"step": 3195
|
| 2199 |
-
},
|
| 2200 |
-
{
|
| 2201 |
-
"epoch": 15.0,
|
| 2202 |
-
"step": 3195,
|
| 2203 |
-
"total_flos": 1172435714212020.0,
|
| 2204 |
-
"train_loss": 0.06678202886758183,
|
| 2205 |
-
"train_runtime": 827.6844,
|
| 2206 |
-
"train_samples_per_second": 410.06,
|
| 2207 |
-
"train_steps_per_second": 25.734
|
| 2208 |
}
|
| 2209 |
],
|
| 2210 |
"max_steps": 21300,
|
| 2211 |
"num_train_epochs": 100,
|
| 2212 |
-
"total_flos":
|
| 2213 |
"trial_name": null,
|
| 2214 |
"trial_params": null
|
| 2215 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.27115458250045776,
|
| 3 |
+
"best_model_checkpoint": "./fine_tune_bert_output_LP_FP/checkpoint-852",
|
| 4 |
+
"epoch": 14.0,
|
| 5 |
+
"global_step": 2982,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
| 9 |
"log_history": [
|
| 10 |
{
|
| 11 |
"epoch": 0.0,
|
| 12 |
+
"learning_rate": 9.999530516431926e-06,
|
| 13 |
+
"loss": 0.0227,
|
| 14 |
"step": 1
|
| 15 |
},
|
| 16 |
{
|
| 17 |
"epoch": 0.05,
|
| 18 |
+
"learning_rate": 9.99530516431925e-06,
|
| 19 |
+
"loss": 0.0194,
|
| 20 |
"step": 10
|
| 21 |
},
|
| 22 |
{
|
| 23 |
"epoch": 0.09,
|
| 24 |
+
"learning_rate": 9.990610328638498e-06,
|
| 25 |
+
"loss": 0.0265,
|
| 26 |
"step": 20
|
| 27 |
},
|
| 28 |
{
|
| 29 |
"epoch": 0.14,
|
| 30 |
+
"learning_rate": 9.985915492957747e-06,
|
| 31 |
+
"loss": 0.0265,
|
| 32 |
"step": 30
|
| 33 |
},
|
| 34 |
{
|
| 35 |
"epoch": 0.19,
|
| 36 |
+
"learning_rate": 9.981220657276996e-06,
|
| 37 |
+
"loss": 0.0188,
|
| 38 |
"step": 40
|
| 39 |
},
|
| 40 |
{
|
| 41 |
"epoch": 0.23,
|
| 42 |
+
"learning_rate": 9.976525821596245e-06,
|
| 43 |
+
"loss": 0.0188,
|
| 44 |
"step": 50
|
| 45 |
},
|
| 46 |
{
|
| 47 |
"epoch": 0.28,
|
| 48 |
+
"learning_rate": 9.971830985915494e-06,
|
| 49 |
+
"loss": 0.0313,
|
| 50 |
"step": 60
|
| 51 |
},
|
| 52 |
{
|
| 53 |
"epoch": 0.33,
|
| 54 |
+
"learning_rate": 9.967136150234742e-06,
|
| 55 |
+
"loss": 0.0259,
|
| 56 |
"step": 70
|
| 57 |
},
|
| 58 |
{
|
| 59 |
"epoch": 0.38,
|
| 60 |
+
"learning_rate": 9.962441314553991e-06,
|
| 61 |
+
"loss": 0.0249,
|
| 62 |
"step": 80
|
| 63 |
},
|
| 64 |
{
|
| 65 |
"epoch": 0.42,
|
| 66 |
+
"learning_rate": 9.95774647887324e-06,
|
| 67 |
+
"loss": 0.0333,
|
| 68 |
"step": 90
|
| 69 |
},
|
| 70 |
{
|
| 71 |
"epoch": 0.47,
|
| 72 |
+
"learning_rate": 9.953051643192489e-06,
|
| 73 |
+
"loss": 0.0337,
|
| 74 |
"step": 100
|
| 75 |
},
|
| 76 |
{
|
| 77 |
"epoch": 0.52,
|
| 78 |
+
"learning_rate": 9.948356807511738e-06,
|
| 79 |
+
"loss": 0.0229,
|
| 80 |
"step": 110
|
| 81 |
},
|
| 82 |
{
|
| 83 |
"epoch": 0.56,
|
| 84 |
+
"learning_rate": 9.943661971830986e-06,
|
| 85 |
+
"loss": 0.013,
|
| 86 |
"step": 120
|
| 87 |
},
|
| 88 |
{
|
| 89 |
"epoch": 0.61,
|
| 90 |
+
"learning_rate": 9.938967136150237e-06,
|
| 91 |
+
"loss": 0.0332,
|
| 92 |
"step": 130
|
| 93 |
},
|
| 94 |
{
|
| 95 |
"epoch": 0.66,
|
| 96 |
+
"learning_rate": 9.934272300469484e-06,
|
| 97 |
+
"loss": 0.0221,
|
| 98 |
"step": 140
|
| 99 |
},
|
| 100 |
{
|
| 101 |
"epoch": 0.7,
|
| 102 |
+
"learning_rate": 9.929577464788733e-06,
|
| 103 |
+
"loss": 0.0198,
|
| 104 |
"step": 150
|
| 105 |
},
|
| 106 |
{
|
| 107 |
"epoch": 0.75,
|
| 108 |
+
"learning_rate": 9.924882629107983e-06,
|
| 109 |
+
"loss": 0.0286,
|
| 110 |
"step": 160
|
| 111 |
},
|
| 112 |
{
|
| 113 |
"epoch": 0.8,
|
| 114 |
+
"learning_rate": 9.92018779342723e-06,
|
| 115 |
+
"loss": 0.0451,
|
| 116 |
"step": 170
|
| 117 |
},
|
| 118 |
{
|
| 119 |
"epoch": 0.85,
|
| 120 |
+
"learning_rate": 9.915492957746479e-06,
|
| 121 |
+
"loss": 0.0257,
|
| 122 |
"step": 180
|
| 123 |
},
|
| 124 |
{
|
| 125 |
"epoch": 0.89,
|
| 126 |
+
"learning_rate": 9.91079812206573e-06,
|
| 127 |
+
"loss": 0.0185,
|
| 128 |
"step": 190
|
| 129 |
},
|
| 130 |
{
|
| 131 |
"epoch": 0.94,
|
| 132 |
+
"learning_rate": 9.906103286384977e-06,
|
| 133 |
+
"loss": 0.032,
|
| 134 |
"step": 200
|
| 135 |
},
|
| 136 |
{
|
| 137 |
"epoch": 0.99,
|
| 138 |
+
"learning_rate": 9.901408450704227e-06,
|
| 139 |
+
"loss": 0.0215,
|
| 140 |
"step": 210
|
| 141 |
},
|
| 142 |
{
|
| 143 |
"epoch": 1.0,
|
| 144 |
+
"eval_corporation_f1": 0.2831858407079646,
|
| 145 |
+
"eval_creative-work_f1": 0.4444444444444445,
|
| 146 |
+
"eval_group_f1": 0.2975206611570248,
|
| 147 |
+
"eval_location_f1": 0.6853932584269663,
|
| 148 |
+
"eval_loss": 0.2913361191749573,
|
| 149 |
+
"eval_overall_accuracy": 0.9507192961597908,
|
| 150 |
+
"eval_overall_f1": 0.641696750902527,
|
| 151 |
+
"eval_overall_precision": 0.7025691699604744,
|
| 152 |
+
"eval_overall_recall": 0.590531561461794,
|
| 153 |
+
"eval_person_f1": 0.7787878787878788,
|
| 154 |
+
"eval_product_f1": 0.4015444015444016,
|
| 155 |
+
"eval_runtime": 6.2036,
|
| 156 |
+
"eval_samples_per_second": 162.648,
|
| 157 |
+
"eval_steps_per_second": 10.317,
|
| 158 |
"step": 213
|
| 159 |
},
|
| 160 |
{
|
| 161 |
"epoch": 1.03,
|
| 162 |
+
"learning_rate": 9.896713615023476e-06,
|
| 163 |
+
"loss": 0.0237,
|
| 164 |
"step": 220
|
| 165 |
},
|
| 166 |
{
|
| 167 |
"epoch": 1.08,
|
| 168 |
+
"learning_rate": 9.892018779342723e-06,
|
| 169 |
+
"loss": 0.0166,
|
| 170 |
"step": 230
|
| 171 |
},
|
| 172 |
{
|
| 173 |
"epoch": 1.13,
|
| 174 |
+
"learning_rate": 9.887323943661974e-06,
|
| 175 |
+
"loss": 0.0144,
|
| 176 |
"step": 240
|
| 177 |
},
|
| 178 |
{
|
| 179 |
"epoch": 1.17,
|
| 180 |
+
"learning_rate": 9.882629107981222e-06,
|
| 181 |
+
"loss": 0.0198,
|
| 182 |
"step": 250
|
| 183 |
},
|
| 184 |
{
|
| 185 |
"epoch": 1.22,
|
| 186 |
+
"learning_rate": 9.87793427230047e-06,
|
| 187 |
+
"loss": 0.0299,
|
| 188 |
"step": 260
|
| 189 |
},
|
| 190 |
{
|
| 191 |
"epoch": 1.27,
|
| 192 |
+
"learning_rate": 9.87323943661972e-06,
|
| 193 |
+
"loss": 0.0123,
|
| 194 |
"step": 270
|
| 195 |
},
|
| 196 |
{
|
| 197 |
"epoch": 1.31,
|
| 198 |
+
"learning_rate": 9.868544600938969e-06,
|
| 199 |
+
"loss": 0.0143,
|
| 200 |
"step": 280
|
| 201 |
},
|
| 202 |
{
|
| 203 |
"epoch": 1.36,
|
| 204 |
+
"learning_rate": 9.863849765258216e-06,
|
| 205 |
+
"loss": 0.0227,
|
| 206 |
"step": 290
|
| 207 |
},
|
| 208 |
{
|
| 209 |
"epoch": 1.41,
|
| 210 |
+
"learning_rate": 9.859154929577466e-06,
|
| 211 |
+
"loss": 0.0125,
|
| 212 |
"step": 300
|
| 213 |
},
|
| 214 |
{
|
| 215 |
"epoch": 1.46,
|
| 216 |
+
"learning_rate": 9.854460093896713e-06,
|
| 217 |
+
"loss": 0.0374,
|
| 218 |
"step": 310
|
| 219 |
},
|
| 220 |
{
|
| 221 |
"epoch": 1.5,
|
| 222 |
+
"learning_rate": 9.849765258215964e-06,
|
| 223 |
+
"loss": 0.0126,
|
| 224 |
"step": 320
|
| 225 |
},
|
| 226 |
{
|
| 227 |
"epoch": 1.55,
|
| 228 |
+
"learning_rate": 9.845070422535213e-06,
|
| 229 |
+
"loss": 0.0234,
|
| 230 |
"step": 330
|
| 231 |
},
|
| 232 |
{
|
| 233 |
"epoch": 1.6,
|
| 234 |
+
"learning_rate": 9.84037558685446e-06,
|
| 235 |
+
"loss": 0.009,
|
| 236 |
"step": 340
|
| 237 |
},
|
| 238 |
{
|
| 239 |
"epoch": 1.64,
|
| 240 |
+
"learning_rate": 9.83568075117371e-06,
|
| 241 |
+
"loss": 0.0149,
|
| 242 |
"step": 350
|
| 243 |
},
|
| 244 |
{
|
| 245 |
"epoch": 1.69,
|
| 246 |
+
"learning_rate": 9.830985915492959e-06,
|
| 247 |
+
"loss": 0.0289,
|
| 248 |
"step": 360
|
| 249 |
},
|
| 250 |
{
|
| 251 |
"epoch": 1.74,
|
| 252 |
+
"learning_rate": 9.826291079812206e-06,
|
| 253 |
+
"loss": 0.03,
|
| 254 |
"step": 370
|
| 255 |
},
|
| 256 |
{
|
| 257 |
"epoch": 1.78,
|
| 258 |
+
"learning_rate": 9.821596244131457e-06,
|
| 259 |
+
"loss": 0.0278,
|
| 260 |
"step": 380
|
| 261 |
},
|
| 262 |
{
|
| 263 |
"epoch": 1.83,
|
| 264 |
+
"learning_rate": 9.816901408450705e-06,
|
| 265 |
+
"loss": 0.0192,
|
| 266 |
"step": 390
|
| 267 |
},
|
| 268 |
{
|
| 269 |
"epoch": 1.88,
|
| 270 |
+
"learning_rate": 9.812206572769954e-06,
|
| 271 |
+
"loss": 0.0187,
|
| 272 |
"step": 400
|
| 273 |
},
|
| 274 |
{
|
| 275 |
"epoch": 1.92,
|
| 276 |
+
"learning_rate": 9.807511737089203e-06,
|
| 277 |
+
"loss": 0.0184,
|
| 278 |
"step": 410
|
| 279 |
},
|
| 280 |
{
|
| 281 |
"epoch": 1.97,
|
| 282 |
+
"learning_rate": 9.802816901408452e-06,
|
| 283 |
+
"loss": 0.0213,
|
| 284 |
"step": 420
|
| 285 |
},
|
| 286 |
{
|
| 287 |
"epoch": 2.0,
|
| 288 |
+
"eval_corporation_f1": 0.2830188679245283,
|
| 289 |
+
"eval_creative-work_f1": 0.3482587064676617,
|
| 290 |
+
"eval_group_f1": 0.3230769230769231,
|
| 291 |
+
"eval_location_f1": 0.6857142857142858,
|
| 292 |
+
"eval_loss": 0.30522316694259644,
|
| 293 |
+
"eval_overall_accuracy": 0.9494709309237903,
|
| 294 |
+
"eval_overall_f1": 0.6233183856502242,
|
| 295 |
+
"eval_overall_precision": 0.6773879142300195,
|
| 296 |
+
"eval_overall_recall": 0.5772425249169435,
|
| 297 |
+
"eval_person_f1": 0.7727620504973222,
|
| 298 |
+
"eval_product_f1": 0.37942122186495186,
|
| 299 |
+
"eval_runtime": 6.2329,
|
| 300 |
+
"eval_samples_per_second": 161.883,
|
| 301 |
+
"eval_steps_per_second": 10.268,
|
| 302 |
"step": 426
|
| 303 |
},
|
| 304 |
{
|
| 305 |
"epoch": 2.02,
|
| 306 |
+
"learning_rate": 9.7981220657277e-06,
|
| 307 |
+
"loss": 0.0294,
|
| 308 |
"step": 430
|
| 309 |
},
|
| 310 |
{
|
| 311 |
"epoch": 2.07,
|
| 312 |
+
"learning_rate": 9.79342723004695e-06,
|
| 313 |
+
"loss": 0.0179,
|
| 314 |
"step": 440
|
| 315 |
},
|
| 316 |
{
|
| 317 |
"epoch": 2.11,
|
| 318 |
+
"learning_rate": 9.788732394366198e-06,
|
| 319 |
+
"loss": 0.021,
|
| 320 |
"step": 450
|
| 321 |
},
|
| 322 |
{
|
| 323 |
"epoch": 2.16,
|
| 324 |
+
"learning_rate": 9.784037558685447e-06,
|
| 325 |
+
"loss": 0.0128,
|
| 326 |
"step": 460
|
| 327 |
},
|
| 328 |
{
|
| 329 |
"epoch": 2.21,
|
| 330 |
+
"learning_rate": 9.779342723004696e-06,
|
| 331 |
+
"loss": 0.0138,
|
| 332 |
"step": 470
|
| 333 |
},
|
| 334 |
{
|
| 335 |
"epoch": 2.25,
|
| 336 |
+
"learning_rate": 9.774647887323945e-06,
|
| 337 |
+
"loss": 0.0293,
|
| 338 |
"step": 480
|
| 339 |
},
|
| 340 |
{
|
| 341 |
"epoch": 2.3,
|
| 342 |
+
"learning_rate": 9.769953051643193e-06,
|
| 343 |
+
"loss": 0.0099,
|
| 344 |
"step": 490
|
| 345 |
},
|
| 346 |
{
|
| 347 |
"epoch": 2.35,
|
| 348 |
+
"learning_rate": 9.765258215962442e-06,
|
| 349 |
+
"loss": 0.0088,
|
| 350 |
"step": 500
|
| 351 |
},
|
| 352 |
{
|
| 353 |
"epoch": 2.39,
|
| 354 |
+
"learning_rate": 9.760563380281691e-06,
|
| 355 |
+
"loss": 0.0137,
|
| 356 |
"step": 510
|
| 357 |
},
|
| 358 |
{
|
| 359 |
"epoch": 2.44,
|
| 360 |
+
"learning_rate": 9.75586854460094e-06,
|
| 361 |
+
"loss": 0.021,
|
| 362 |
"step": 520
|
| 363 |
},
|
| 364 |
{
|
| 365 |
"epoch": 2.49,
|
| 366 |
+
"learning_rate": 9.751173708920188e-06,
|
| 367 |
+
"loss": 0.0197,
|
| 368 |
"step": 530
|
| 369 |
},
|
| 370 |
{
|
| 371 |
"epoch": 2.54,
|
| 372 |
+
"learning_rate": 9.746478873239437e-06,
|
| 373 |
+
"loss": 0.0208,
|
| 374 |
"step": 540
|
| 375 |
},
|
| 376 |
{
|
| 377 |
"epoch": 2.58,
|
| 378 |
+
"learning_rate": 9.741784037558686e-06,
|
| 379 |
+
"loss": 0.0074,
|
| 380 |
"step": 550
|
| 381 |
},
|
| 382 |
{
|
| 383 |
"epoch": 2.63,
|
| 384 |
+
"learning_rate": 9.737089201877935e-06,
|
| 385 |
+
"loss": 0.017,
|
| 386 |
"step": 560
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 2.68,
|
| 390 |
+
"learning_rate": 9.732394366197184e-06,
|
| 391 |
+
"loss": 0.0205,
|
| 392 |
"step": 570
|
| 393 |
},
|
| 394 |
{
|
| 395 |
"epoch": 2.72,
|
| 396 |
+
"learning_rate": 9.727699530516432e-06,
|
| 397 |
+
"loss": 0.0159,
|
| 398 |
"step": 580
|
| 399 |
},
|
| 400 |
{
|
| 401 |
"epoch": 2.77,
|
| 402 |
+
"learning_rate": 9.723004694835681e-06,
|
| 403 |
+
"loss": 0.0139,
|
| 404 |
"step": 590
|
| 405 |
},
|
| 406 |
{
|
| 407 |
"epoch": 2.82,
|
| 408 |
+
"learning_rate": 9.71830985915493e-06,
|
| 409 |
+
"loss": 0.017,
|
| 410 |
"step": 600
|
| 411 |
},
|
| 412 |
{
|
| 413 |
"epoch": 2.86,
|
| 414 |
+
"learning_rate": 9.713615023474179e-06,
|
| 415 |
+
"loss": 0.0171,
|
| 416 |
"step": 610
|
| 417 |
},
|
| 418 |
{
|
| 419 |
"epoch": 2.91,
|
| 420 |
+
"learning_rate": 9.708920187793428e-06,
|
| 421 |
+
"loss": 0.0164,
|
| 422 |
"step": 620
|
| 423 |
},
|
| 424 |
{
|
| 425 |
"epoch": 2.96,
|
| 426 |
+
"learning_rate": 9.704225352112678e-06,
|
| 427 |
+
"loss": 0.0288,
|
| 428 |
"step": 630
|
| 429 |
},
|
| 430 |
{
|
| 431 |
"epoch": 3.0,
|
| 432 |
+
"eval_corporation_f1": 0.3076923076923077,
|
| 433 |
+
"eval_creative-work_f1": 0.41841004184100417,
|
| 434 |
+
"eval_group_f1": 0.3529411764705882,
|
| 435 |
+
"eval_location_f1": 0.6222222222222222,
|
| 436 |
+
"eval_loss": 0.3378466069698334,
|
| 437 |
+
"eval_overall_accuracy": 0.946736416597313,
|
| 438 |
+
"eval_overall_f1": 0.6187587494167056,
|
| 439 |
+
"eval_overall_precision": 0.7060702875399361,
|
| 440 |
+
"eval_overall_recall": 0.5506644518272426,
|
| 441 |
+
"eval_person_f1": 0.7532051282051282,
|
| 442 |
+
"eval_product_f1": 0.39097744360902253,
|
| 443 |
+
"eval_runtime": 2.031,
|
| 444 |
+
"eval_samples_per_second": 496.798,
|
| 445 |
+
"eval_steps_per_second": 31.511,
|
| 446 |
"step": 639
|
| 447 |
},
|
| 448 |
{
|
| 449 |
"epoch": 3.0,
|
| 450 |
+
"learning_rate": 9.699530516431925e-06,
|
| 451 |
+
"loss": 0.023,
|
| 452 |
"step": 640
|
| 453 |
},
|
| 454 |
{
|
| 455 |
"epoch": 3.05,
|
| 456 |
+
"learning_rate": 9.694835680751174e-06,
|
| 457 |
+
"loss": 0.0173,
|
| 458 |
"step": 650
|
| 459 |
},
|
| 460 |
{
|
| 461 |
"epoch": 3.1,
|
| 462 |
+
"learning_rate": 9.690140845070424e-06,
|
| 463 |
+
"loss": 0.0149,
|
| 464 |
"step": 660
|
| 465 |
},
|
| 466 |
{
|
| 467 |
"epoch": 3.15,
|
| 468 |
+
"learning_rate": 9.685446009389672e-06,
|
| 469 |
+
"loss": 0.0141,
|
| 470 |
"step": 670
|
| 471 |
},
|
| 472 |
{
|
| 473 |
"epoch": 3.19,
|
| 474 |
+
"learning_rate": 9.68075117370892e-06,
|
| 475 |
+
"loss": 0.0172,
|
| 476 |
"step": 680
|
| 477 |
},
|
| 478 |
{
|
| 479 |
"epoch": 3.24,
|
| 480 |
+
"learning_rate": 9.67605633802817e-06,
|
| 481 |
+
"loss": 0.0072,
|
| 482 |
"step": 690
|
| 483 |
},
|
| 484 |
{
|
| 485 |
"epoch": 3.29,
|
| 486 |
+
"learning_rate": 9.671361502347418e-06,
|
| 487 |
+
"loss": 0.015,
|
| 488 |
"step": 700
|
| 489 |
},
|
| 490 |
{
|
| 491 |
"epoch": 3.33,
|
| 492 |
+
"learning_rate": 9.666666666666667e-06,
|
| 493 |
+
"loss": 0.009,
|
| 494 |
"step": 710
|
| 495 |
},
|
| 496 |
{
|
| 497 |
"epoch": 3.38,
|
| 498 |
+
"learning_rate": 9.661971830985917e-06,
|
| 499 |
+
"loss": 0.0098,
|
| 500 |
"step": 720
|
| 501 |
},
|
| 502 |
{
|
| 503 |
"epoch": 3.43,
|
| 504 |
+
"learning_rate": 9.657276995305164e-06,
|
| 505 |
+
"loss": 0.018,
|
| 506 |
"step": 730
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 3.47,
|
| 510 |
+
"learning_rate": 9.652582159624415e-06,
|
| 511 |
+
"loss": 0.0092,
|
| 512 |
"step": 740
|
| 513 |
},
|
| 514 |
{
|
| 515 |
"epoch": 3.52,
|
| 516 |
+
"learning_rate": 9.647887323943664e-06,
|
| 517 |
+
"loss": 0.0166,
|
| 518 |
"step": 750
|
| 519 |
},
|
| 520 |
{
|
| 521 |
"epoch": 3.57,
|
| 522 |
+
"learning_rate": 9.64319248826291e-06,
|
| 523 |
+
"loss": 0.0085,
|
| 524 |
"step": 760
|
| 525 |
},
|
| 526 |
{
|
| 527 |
"epoch": 3.62,
|
| 528 |
+
"learning_rate": 9.638497652582161e-06,
|
| 529 |
+
"loss": 0.0168,
|
| 530 |
"step": 770
|
| 531 |
},
|
| 532 |
{
|
| 533 |
"epoch": 3.66,
|
| 534 |
+
"learning_rate": 9.63380281690141e-06,
|
| 535 |
+
"loss": 0.0201,
|
| 536 |
"step": 780
|
| 537 |
},
|
| 538 |
{
|
| 539 |
"epoch": 3.71,
|
| 540 |
+
"learning_rate": 9.629107981220657e-06,
|
| 541 |
+
"loss": 0.0143,
|
| 542 |
"step": 790
|
| 543 |
},
|
| 544 |
{
|
| 545 |
"epoch": 3.76,
|
| 546 |
+
"learning_rate": 9.624413145539908e-06,
|
| 547 |
+
"loss": 0.0189,
|
| 548 |
"step": 800
|
| 549 |
},
|
| 550 |
{
|
| 551 |
"epoch": 3.8,
|
| 552 |
+
"learning_rate": 9.619718309859156e-06,
|
| 553 |
+
"loss": 0.0099,
|
| 554 |
"step": 810
|
| 555 |
},
|
| 556 |
{
|
| 557 |
"epoch": 3.85,
|
| 558 |
+
"learning_rate": 9.615023474178405e-06,
|
| 559 |
+
"loss": 0.0259,
|
| 560 |
"step": 820
|
| 561 |
},
|
| 562 |
{
|
| 563 |
"epoch": 3.9,
|
| 564 |
+
"learning_rate": 9.610328638497654e-06,
|
| 565 |
+
"loss": 0.0187,
|
| 566 |
"step": 830
|
| 567 |
},
|
| 568 |
{
|
| 569 |
"epoch": 3.94,
|
| 570 |
+
"learning_rate": 9.605633802816903e-06,
|
| 571 |
+
"loss": 0.0102,
|
| 572 |
"step": 840
|
| 573 |
},
|
| 574 |
{
|
| 575 |
"epoch": 3.99,
|
| 576 |
+
"learning_rate": 9.600938967136152e-06,
|
| 577 |
+
"loss": 0.0124,
|
| 578 |
"step": 850
|
| 579 |
},
|
| 580 |
{
|
| 581 |
"epoch": 4.0,
|
| 582 |
+
"eval_corporation_f1": 0.30769230769230765,
|
| 583 |
+
"eval_creative-work_f1": 0.48421052631578954,
|
| 584 |
+
"eval_group_f1": 0.31666666666666665,
|
| 585 |
+
"eval_location_f1": 0.6808510638297872,
|
| 586 |
+
"eval_loss": 0.27115458250045776,
|
| 587 |
+
"eval_overall_accuracy": 0.9502437284508382,
|
| 588 |
+
"eval_overall_f1": 0.6339784946236559,
|
| 589 |
+
"eval_overall_precision": 0.6574487065120428,
|
| 590 |
+
"eval_overall_recall": 0.6121262458471761,
|
| 591 |
+
"eval_person_f1": 0.7734553775743707,
|
| 592 |
+
"eval_product_f1": 0.3986254295532646,
|
| 593 |
+
"eval_runtime": 5.0216,
|
| 594 |
+
"eval_samples_per_second": 200.934,
|
| 595 |
+
"eval_steps_per_second": 12.745,
|
| 596 |
"step": 852
|
| 597 |
},
|
| 598 |
{
|
| 599 |
"epoch": 4.04,
|
| 600 |
+
"learning_rate": 9.5962441314554e-06,
|
| 601 |
+
"loss": 0.017,
|
| 602 |
"step": 860
|
| 603 |
},
|
| 604 |
{
|
| 605 |
"epoch": 4.08,
|
| 606 |
+
"learning_rate": 9.591549295774649e-06,
|
| 607 |
+
"loss": 0.0138,
|
| 608 |
"step": 870
|
| 609 |
},
|
| 610 |
{
|
| 611 |
"epoch": 4.13,
|
| 612 |
+
"learning_rate": 9.586854460093898e-06,
|
| 613 |
+
"loss": 0.0122,
|
| 614 |
"step": 880
|
| 615 |
},
|
| 616 |
{
|
| 617 |
"epoch": 4.18,
|
| 618 |
+
"learning_rate": 9.582159624413147e-06,
|
| 619 |
+
"loss": 0.0116,
|
| 620 |
"step": 890
|
| 621 |
},
|
| 622 |
{
|
| 623 |
"epoch": 4.23,
|
| 624 |
+
"learning_rate": 9.577464788732394e-06,
|
| 625 |
+
"loss": 0.0066,
|
| 626 |
"step": 900
|
| 627 |
},
|
| 628 |
{
|
| 629 |
"epoch": 4.27,
|
| 630 |
+
"learning_rate": 9.572769953051644e-06,
|
| 631 |
+
"loss": 0.016,
|
| 632 |
"step": 910
|
| 633 |
},
|
| 634 |
{
|
| 635 |
"epoch": 4.32,
|
| 636 |
+
"learning_rate": 9.568075117370893e-06,
|
| 637 |
+
"loss": 0.0203,
|
| 638 |
"step": 920
|
| 639 |
},
|
| 640 |
{
|
| 641 |
"epoch": 4.37,
|
| 642 |
+
"learning_rate": 9.563380281690142e-06,
|
| 643 |
+
"loss": 0.0092,
|
| 644 |
"step": 930
|
| 645 |
},
|
| 646 |
{
|
| 647 |
"epoch": 4.41,
|
| 648 |
+
"learning_rate": 9.55868544600939e-06,
|
| 649 |
+
"loss": 0.0098,
|
| 650 |
"step": 940
|
| 651 |
},
|
| 652 |
{
|
| 653 |
"epoch": 4.46,
|
| 654 |
+
"learning_rate": 9.55399061032864e-06,
|
| 655 |
+
"loss": 0.0083,
|
| 656 |
"step": 950
|
| 657 |
},
|
| 658 |
{
|
| 659 |
"epoch": 4.51,
|
| 660 |
+
"learning_rate": 9.549295774647888e-06,
|
| 661 |
+
"loss": 0.0244,
|
| 662 |
"step": 960
|
| 663 |
},
|
| 664 |
{
|
| 665 |
"epoch": 4.55,
|
| 666 |
+
"learning_rate": 9.544600938967137e-06,
|
| 667 |
+
"loss": 0.0096,
|
| 668 |
"step": 970
|
| 669 |
},
|
| 670 |
{
|
| 671 |
"epoch": 4.6,
|
| 672 |
+
"learning_rate": 9.539906103286386e-06,
|
| 673 |
+
"loss": 0.0109,
|
| 674 |
"step": 980
|
| 675 |
},
|
| 676 |
{
|
| 677 |
"epoch": 4.65,
|
| 678 |
+
"learning_rate": 9.535211267605635e-06,
|
| 679 |
+
"loss": 0.0114,
|
| 680 |
"step": 990
|
| 681 |
},
|
| 682 |
{
|
| 683 |
"epoch": 4.69,
|
| 684 |
+
"learning_rate": 9.530516431924883e-06,
|
| 685 |
+
"loss": 0.0214,
|
| 686 |
"step": 1000
|
| 687 |
},
|
| 688 |
{
|
| 689 |
"epoch": 4.74,
|
| 690 |
+
"learning_rate": 9.525821596244132e-06,
|
| 691 |
+
"loss": 0.0061,
|
| 692 |
"step": 1010
|
| 693 |
},
|
| 694 |
{
|
| 695 |
"epoch": 4.79,
|
| 696 |
+
"learning_rate": 9.521126760563381e-06,
|
| 697 |
+
"loss": 0.011,
|
| 698 |
"step": 1020
|
| 699 |
},
|
| 700 |
{
|
| 701 |
"epoch": 4.84,
|
| 702 |
+
"learning_rate": 9.51643192488263e-06,
|
| 703 |
+
"loss": 0.0061,
|
| 704 |
"step": 1030
|
| 705 |
},
|
| 706 |
{
|
| 707 |
"epoch": 4.88,
|
| 708 |
+
"learning_rate": 9.511737089201879e-06,
|
| 709 |
+
"loss": 0.0202,
|
| 710 |
"step": 1040
|
| 711 |
},
|
| 712 |
{
|
| 713 |
"epoch": 4.93,
|
| 714 |
+
"learning_rate": 9.507042253521127e-06,
|
| 715 |
+
"loss": 0.0249,
|
| 716 |
"step": 1050
|
| 717 |
},
|
| 718 |
{
|
| 719 |
"epoch": 4.98,
|
| 720 |
+
"learning_rate": 9.502347417840376e-06,
|
| 721 |
+
"loss": 0.0208,
|
| 722 |
"step": 1060
|
| 723 |
},
|
| 724 |
{
|
| 725 |
"epoch": 5.0,
|
| 726 |
+
"eval_corporation_f1": 0.3063063063063063,
|
| 727 |
+
"eval_creative-work_f1": 0.42857142857142855,
|
| 728 |
+
"eval_group_f1": 0.3418803418803419,
|
| 729 |
+
"eval_location_f1": 0.7052023121387283,
|
| 730 |
+
"eval_loss": 0.29045674204826355,
|
| 731 |
+
"eval_overall_accuracy": 0.9518487694685531,
|
| 732 |
+
"eval_overall_f1": 0.6544150605109816,
|
| 733 |
+
"eval_overall_precision": 0.7108081791626095,
|
| 734 |
+
"eval_overall_recall": 0.606312292358804,
|
| 735 |
+
"eval_person_f1": 0.7912584777694045,
|
| 736 |
+
"eval_product_f1": 0.42231075697211157,
|
| 737 |
+
"eval_runtime": 4.8738,
|
| 738 |
+
"eval_samples_per_second": 207.026,
|
| 739 |
+
"eval_steps_per_second": 13.131,
|
| 740 |
"step": 1065
|
| 741 |
},
|
| 742 |
{
|
| 743 |
"epoch": 5.02,
|
| 744 |
+
"learning_rate": 9.497652582159625e-06,
|
| 745 |
+
"loss": 0.0067,
|
| 746 |
"step": 1070
|
| 747 |
},
|
| 748 |
{
|
| 749 |
"epoch": 5.07,
|
| 750 |
+
"learning_rate": 9.492957746478874e-06,
|
| 751 |
+
"loss": 0.017,
|
| 752 |
"step": 1080
|
| 753 |
},
|
| 754 |
{
|
| 755 |
"epoch": 5.12,
|
| 756 |
+
"learning_rate": 9.488262910798123e-06,
|
| 757 |
+
"loss": 0.0069,
|
| 758 |
"step": 1090
|
| 759 |
},
|
| 760 |
{
|
| 761 |
"epoch": 5.16,
|
| 762 |
+
"learning_rate": 9.483568075117371e-06,
|
| 763 |
+
"loss": 0.0109,
|
| 764 |
"step": 1100
|
| 765 |
},
|
| 766 |
{
|
| 767 |
"epoch": 5.21,
|
| 768 |
+
"learning_rate": 9.47887323943662e-06,
|
| 769 |
+
"loss": 0.0108,
|
| 770 |
"step": 1110
|
| 771 |
},
|
| 772 |
{
|
| 773 |
"epoch": 5.26,
|
| 774 |
+
"learning_rate": 9.474178403755869e-06,
|
| 775 |
+
"loss": 0.0112,
|
| 776 |
"step": 1120
|
| 777 |
},
|
| 778 |
{
|
| 779 |
"epoch": 5.31,
|
| 780 |
+
"learning_rate": 9.469483568075118e-06,
|
| 781 |
+
"loss": 0.0042,
|
| 782 |
"step": 1130
|
| 783 |
},
|
| 784 |
{
|
| 785 |
"epoch": 5.35,
|
| 786 |
+
"learning_rate": 9.464788732394366e-06,
|
| 787 |
+
"loss": 0.0086,
|
| 788 |
"step": 1140
|
| 789 |
},
|
| 790 |
{
|
| 791 |
"epoch": 5.4,
|
| 792 |
+
"learning_rate": 9.460093896713615e-06,
|
| 793 |
+
"loss": 0.0067,
|
| 794 |
"step": 1150
|
| 795 |
},
|
| 796 |
{
|
| 797 |
"epoch": 5.45,
|
| 798 |
+
"learning_rate": 9.455399061032866e-06,
|
| 799 |
+
"loss": 0.0099,
|
| 800 |
"step": 1160
|
| 801 |
},
|
| 802 |
{
|
| 803 |
"epoch": 5.49,
|
| 804 |
+
"learning_rate": 9.450704225352113e-06,
|
| 805 |
+
"loss": 0.0071,
|
| 806 |
"step": 1170
|
| 807 |
},
|
| 808 |
{
|
| 809 |
"epoch": 5.54,
|
| 810 |
+
"learning_rate": 9.446009389671362e-06,
|
| 811 |
+
"loss": 0.0061,
|
| 812 |
"step": 1180
|
| 813 |
},
|
| 814 |
{
|
| 815 |
"epoch": 5.59,
|
| 816 |
+
"learning_rate": 9.441314553990612e-06,
|
| 817 |
+
"loss": 0.0088,
|
| 818 |
"step": 1190
|
| 819 |
},
|
| 820 |
{
|
| 821 |
"epoch": 5.63,
|
| 822 |
+
"learning_rate": 9.43661971830986e-06,
|
| 823 |
+
"loss": 0.0078,
|
| 824 |
"step": 1200
|
| 825 |
},
|
| 826 |
{
|
| 827 |
"epoch": 5.68,
|
| 828 |
+
"learning_rate": 9.431924882629108e-06,
|
| 829 |
+
"loss": 0.0041,
|
| 830 |
"step": 1210
|
| 831 |
},
|
| 832 |
{
|
| 833 |
"epoch": 5.73,
|
| 834 |
+
"learning_rate": 9.427230046948358e-06,
|
| 835 |
+
"loss": 0.019,
|
| 836 |
"step": 1220
|
| 837 |
},
|
| 838 |
{
|
| 839 |
"epoch": 5.77,
|
| 840 |
+
"learning_rate": 9.422535211267606e-06,
|
| 841 |
+
"loss": 0.0113,
|
| 842 |
"step": 1230
|
| 843 |
},
|
| 844 |
{
|
| 845 |
"epoch": 5.82,
|
| 846 |
+
"learning_rate": 9.417840375586856e-06,
|
| 847 |
+
"loss": 0.0245,
|
| 848 |
"step": 1240
|
| 849 |
},
|
| 850 |
{
|
| 851 |
"epoch": 5.87,
|
| 852 |
+
"learning_rate": 9.413145539906105e-06,
|
| 853 |
+
"loss": 0.0195,
|
| 854 |
"step": 1250
|
| 855 |
},
|
| 856 |
{
|
| 857 |
"epoch": 5.92,
|
| 858 |
+
"learning_rate": 9.408450704225352e-06,
|
| 859 |
+
"loss": 0.007,
|
| 860 |
"step": 1260
|
| 861 |
},
|
| 862 |
{
|
| 863 |
"epoch": 5.96,
|
| 864 |
+
"learning_rate": 9.403755868544602e-06,
|
| 865 |
+
"loss": 0.0071,
|
| 866 |
"step": 1270
|
| 867 |
},
|
| 868 |
{
|
| 869 |
"epoch": 6.0,
|
| 870 |
+
"eval_corporation_f1": 0.27586206896551724,
|
| 871 |
+
"eval_creative-work_f1": 0.4380165289256198,
|
| 872 |
+
"eval_group_f1": 0.32558139534883723,
|
| 873 |
+
"eval_location_f1": 0.6744186046511628,
|
| 874 |
+
"eval_loss": 0.3188508152961731,
|
| 875 |
+
"eval_overall_accuracy": 0.9494114849601712,
|
| 876 |
+
"eval_overall_f1": 0.6268922528940339,
|
| 877 |
+
"eval_overall_precision": 0.6756238003838771,
|
| 878 |
+
"eval_overall_recall": 0.584717607973422,
|
| 879 |
+
"eval_person_f1": 0.7781250000000001,
|
| 880 |
+
"eval_product_f1": 0.37785016286644946,
|
| 881 |
+
"eval_runtime": 5.2125,
|
| 882 |
+
"eval_samples_per_second": 193.573,
|
| 883 |
+
"eval_steps_per_second": 12.278,
|
| 884 |
"step": 1278
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 6.01,
|
| 888 |
+
"learning_rate": 9.399061032863851e-06,
|
| 889 |
+
"loss": 0.0069,
|
| 890 |
"step": 1280
|
| 891 |
},
|
| 892 |
{
|
| 893 |
"epoch": 6.06,
|
| 894 |
+
"learning_rate": 9.394366197183098e-06,
|
| 895 |
+
"loss": 0.0062,
|
| 896 |
"step": 1290
|
| 897 |
},
|
| 898 |
{
|
| 899 |
"epoch": 6.1,
|
| 900 |
+
"learning_rate": 9.389671361502349e-06,
|
| 901 |
+
"loss": 0.0096,
|
| 902 |
"step": 1300
|
| 903 |
},
|
| 904 |
{
|
| 905 |
"epoch": 6.15,
|
| 906 |
+
"learning_rate": 9.384976525821598e-06,
|
| 907 |
+
"loss": 0.0184,
|
| 908 |
"step": 1310
|
| 909 |
},
|
| 910 |
{
|
| 911 |
"epoch": 6.2,
|
| 912 |
+
"learning_rate": 9.380281690140845e-06,
|
| 913 |
+
"loss": 0.0116,
|
| 914 |
"step": 1320
|
| 915 |
},
|
| 916 |
{
|
| 917 |
"epoch": 6.24,
|
| 918 |
+
"learning_rate": 9.375586854460095e-06,
|
| 919 |
+
"loss": 0.0092,
|
| 920 |
"step": 1330
|
| 921 |
},
|
| 922 |
{
|
| 923 |
"epoch": 6.29,
|
| 924 |
+
"learning_rate": 9.370892018779344e-06,
|
| 925 |
+
"loss": 0.0075,
|
| 926 |
"step": 1340
|
| 927 |
},
|
| 928 |
{
|
| 929 |
"epoch": 6.34,
|
| 930 |
+
"learning_rate": 9.366197183098593e-06,
|
| 931 |
+
"loss": 0.0101,
|
| 932 |
"step": 1350
|
| 933 |
},
|
| 934 |
{
|
| 935 |
"epoch": 6.38,
|
| 936 |
+
"learning_rate": 9.361502347417842e-06,
|
| 937 |
+
"loss": 0.0106,
|
| 938 |
"step": 1360
|
| 939 |
},
|
| 940 |
{
|
| 941 |
"epoch": 6.43,
|
| 942 |
+
"learning_rate": 9.35680751173709e-06,
|
| 943 |
+
"loss": 0.0069,
|
| 944 |
"step": 1370
|
| 945 |
},
|
| 946 |
{
|
| 947 |
"epoch": 6.48,
|
| 948 |
+
"learning_rate": 9.35211267605634e-06,
|
| 949 |
+
"loss": 0.0064,
|
| 950 |
"step": 1380
|
| 951 |
},
|
| 952 |
{
|
| 953 |
"epoch": 6.53,
|
| 954 |
+
"learning_rate": 9.347417840375588e-06,
|
| 955 |
+
"loss": 0.0112,
|
| 956 |
"step": 1390
|
| 957 |
},
|
| 958 |
{
|
| 959 |
"epoch": 6.57,
|
| 960 |
+
"learning_rate": 9.342723004694837e-06,
|
| 961 |
+
"loss": 0.0059,
|
| 962 |
"step": 1400
|
| 963 |
},
|
| 964 |
{
|
| 965 |
"epoch": 6.62,
|
| 966 |
+
"learning_rate": 9.338028169014086e-06,
|
| 967 |
+
"loss": 0.006,
|
| 968 |
"step": 1410
|
| 969 |
},
|
| 970 |
{
|
| 971 |
"epoch": 6.67,
|
| 972 |
+
"learning_rate": 9.333333333333334e-06,
|
| 973 |
+
"loss": 0.0089,
|
| 974 |
"step": 1420
|
| 975 |
},
|
| 976 |
{
|
| 977 |
"epoch": 6.71,
|
| 978 |
+
"learning_rate": 9.328638497652583e-06,
|
| 979 |
+
"loss": 0.0068,
|
| 980 |
"step": 1430
|
| 981 |
},
|
| 982 |
{
|
| 983 |
"epoch": 6.76,
|
| 984 |
+
"learning_rate": 9.323943661971832e-06,
|
| 985 |
+
"loss": 0.0041,
|
| 986 |
"step": 1440
|
| 987 |
},
|
| 988 |
{
|
| 989 |
"epoch": 6.81,
|
| 990 |
+
"learning_rate": 9.31924882629108e-06,
|
| 991 |
+
"loss": 0.0191,
|
| 992 |
"step": 1450
|
| 993 |
},
|
| 994 |
{
|
| 995 |
"epoch": 6.85,
|
| 996 |
+
"learning_rate": 9.31455399061033e-06,
|
| 997 |
+
"loss": 0.0123,
|
| 998 |
"step": 1460
|
| 999 |
},
|
| 1000 |
{
|
| 1001 |
"epoch": 6.9,
|
| 1002 |
+
"learning_rate": 9.309859154929578e-06,
|
| 1003 |
+
"loss": 0.0033,
|
| 1004 |
"step": 1470
|
| 1005 |
},
|
| 1006 |
{
|
| 1007 |
"epoch": 6.95,
|
| 1008 |
+
"learning_rate": 9.305164319248827e-06,
|
| 1009 |
+
"loss": 0.0085,
|
| 1010 |
"step": 1480
|
| 1011 |
},
|
| 1012 |
{
|
| 1013 |
"epoch": 7.0,
|
| 1014 |
+
"learning_rate": 9.300469483568076e-06,
|
| 1015 |
+
"loss": 0.0073,
|
| 1016 |
"step": 1490
|
| 1017 |
},
|
| 1018 |
{
|
| 1019 |
"epoch": 7.0,
|
| 1020 |
+
"eval_corporation_f1": 0.3061224489795918,
|
| 1021 |
+
"eval_creative-work_f1": 0.4388185654008439,
|
| 1022 |
+
"eval_group_f1": 0.37837837837837834,
|
| 1023 |
+
"eval_location_f1": 0.6946107784431138,
|
| 1024 |
+
"eval_loss": 0.35933107137680054,
|
| 1025 |
+
"eval_overall_accuracy": 0.94756866008798,
|
| 1026 |
+
"eval_overall_f1": 0.631031220435194,
|
| 1027 |
+
"eval_overall_precision": 0.7329670329670329,
|
| 1028 |
+
"eval_overall_recall": 0.5539867109634552,
|
| 1029 |
+
"eval_person_f1": 0.7631160572337042,
|
| 1030 |
+
"eval_product_f1": 0.3374485596707819,
|
| 1031 |
+
"eval_runtime": 4.8206,
|
| 1032 |
+
"eval_samples_per_second": 209.311,
|
| 1033 |
+
"eval_steps_per_second": 13.276,
|
| 1034 |
"step": 1491
|
| 1035 |
},
|
| 1036 |
{
|
| 1037 |
"epoch": 7.04,
|
| 1038 |
+
"learning_rate": 9.295774647887325e-06,
|
| 1039 |
+
"loss": 0.0109,
|
| 1040 |
"step": 1500
|
| 1041 |
},
|
| 1042 |
{
|
| 1043 |
"epoch": 7.09,
|
| 1044 |
+
"learning_rate": 9.291079812206573e-06,
|
| 1045 |
+
"loss": 0.0049,
|
| 1046 |
"step": 1510
|
| 1047 |
},
|
| 1048 |
{
|
| 1049 |
"epoch": 7.14,
|
| 1050 |
+
"learning_rate": 9.286384976525822e-06,
|
| 1051 |
+
"loss": 0.0098,
|
| 1052 |
"step": 1520
|
| 1053 |
},
|
| 1054 |
{
|
| 1055 |
"epoch": 7.18,
|
| 1056 |
+
"learning_rate": 9.281690140845071e-06,
|
| 1057 |
+
"loss": 0.0053,
|
| 1058 |
"step": 1530
|
| 1059 |
},
|
| 1060 |
{
|
| 1061 |
"epoch": 7.23,
|
| 1062 |
+
"learning_rate": 9.27699530516432e-06,
|
| 1063 |
+
"loss": 0.0054,
|
| 1064 |
"step": 1540
|
| 1065 |
},
|
| 1066 |
{
|
| 1067 |
"epoch": 7.28,
|
| 1068 |
+
"learning_rate": 9.272300469483569e-06,
|
| 1069 |
+
"loss": 0.0123,
|
| 1070 |
"step": 1550
|
| 1071 |
},
|
| 1072 |
{
|
| 1073 |
"epoch": 7.32,
|
| 1074 |
+
"learning_rate": 9.267605633802817e-06,
|
| 1075 |
+
"loss": 0.0037,
|
| 1076 |
"step": 1560
|
| 1077 |
},
|
| 1078 |
{
|
| 1079 |
"epoch": 7.37,
|
| 1080 |
+
"learning_rate": 9.262910798122066e-06,
|
| 1081 |
+
"loss": 0.0029,
|
| 1082 |
"step": 1570
|
| 1083 |
},
|
| 1084 |
{
|
| 1085 |
"epoch": 7.42,
|
| 1086 |
+
"learning_rate": 9.258215962441315e-06,
|
| 1087 |
+
"loss": 0.0186,
|
| 1088 |
"step": 1580
|
| 1089 |
},
|
| 1090 |
{
|
| 1091 |
"epoch": 7.46,
|
| 1092 |
+
"learning_rate": 9.253521126760564e-06,
|
| 1093 |
+
"loss": 0.0089,
|
| 1094 |
"step": 1590
|
| 1095 |
},
|
| 1096 |
{
|
| 1097 |
"epoch": 7.51,
|
| 1098 |
+
"learning_rate": 9.248826291079813e-06,
|
| 1099 |
+
"loss": 0.006,
|
| 1100 |
"step": 1600
|
| 1101 |
},
|
| 1102 |
{
|
| 1103 |
"epoch": 7.56,
|
| 1104 |
+
"learning_rate": 9.244131455399061e-06,
|
| 1105 |
+
"loss": 0.0087,
|
| 1106 |
"step": 1610
|
| 1107 |
},
|
| 1108 |
{
|
| 1109 |
"epoch": 7.61,
|
| 1110 |
+
"learning_rate": 9.23943661971831e-06,
|
| 1111 |
+
"loss": 0.0099,
|
| 1112 |
"step": 1620
|
| 1113 |
},
|
| 1114 |
{
|
| 1115 |
"epoch": 7.65,
|
| 1116 |
+
"learning_rate": 9.234741784037559e-06,
|
| 1117 |
+
"loss": 0.0293,
|
| 1118 |
"step": 1630
|
| 1119 |
},
|
| 1120 |
{
|
| 1121 |
"epoch": 7.7,
|
| 1122 |
+
"learning_rate": 9.230046948356808e-06,
|
| 1123 |
+
"loss": 0.0129,
|
| 1124 |
"step": 1640
|
| 1125 |
},
|
| 1126 |
{
|
| 1127 |
"epoch": 7.75,
|
| 1128 |
+
"learning_rate": 9.225352112676057e-06,
|
| 1129 |
+
"loss": 0.0207,
|
| 1130 |
"step": 1650
|
| 1131 |
},
|
| 1132 |
{
|
| 1133 |
"epoch": 7.79,
|
| 1134 |
+
"learning_rate": 9.220657276995307e-06,
|
| 1135 |
+
"loss": 0.0055,
|
| 1136 |
"step": 1660
|
| 1137 |
},
|
| 1138 |
{
|
| 1139 |
"epoch": 7.84,
|
| 1140 |
+
"learning_rate": 9.215962441314554e-06,
|
| 1141 |
+
"loss": 0.0059,
|
| 1142 |
"step": 1670
|
| 1143 |
},
|
| 1144 |
{
|
| 1145 |
"epoch": 7.89,
|
| 1146 |
+
"learning_rate": 9.211267605633803e-06,
|
| 1147 |
+
"loss": 0.0103,
|
| 1148 |
"step": 1680
|
| 1149 |
},
|
| 1150 |
{
|
| 1151 |
"epoch": 7.93,
|
| 1152 |
+
"learning_rate": 9.206572769953053e-06,
|
| 1153 |
+
"loss": 0.018,
|
| 1154 |
"step": 1690
|
| 1155 |
},
|
| 1156 |
{
|
| 1157 |
"epoch": 7.98,
|
| 1158 |
+
"learning_rate": 9.2018779342723e-06,
|
| 1159 |
+
"loss": 0.0135,
|
| 1160 |
"step": 1700
|
| 1161 |
},
|
| 1162 |
{
|
| 1163 |
"epoch": 8.0,
|
| 1164 |
+
"eval_corporation_f1": 0.34,
|
| 1165 |
+
"eval_creative-work_f1": 0.417910447761194,
|
| 1166 |
+
"eval_group_f1": 0.3088235294117647,
|
| 1167 |
+
"eval_location_f1": 0.6631578947368422,
|
| 1168 |
+
"eval_loss": 0.35640954971313477,
|
| 1169 |
+
"eval_overall_accuracy": 0.9470930923790275,
|
| 1170 |
+
"eval_overall_f1": 0.609981515711645,
|
| 1171 |
+
"eval_overall_precision": 0.6875,
|
| 1172 |
+
"eval_overall_recall": 0.5481727574750831,
|
| 1173 |
+
"eval_person_f1": 0.7485667485667485,
|
| 1174 |
+
"eval_product_f1": 0.3694779116465864,
|
| 1175 |
+
"eval_runtime": 6.3558,
|
| 1176 |
+
"eval_samples_per_second": 158.753,
|
| 1177 |
+
"eval_steps_per_second": 10.07,
|
| 1178 |
"step": 1704
|
| 1179 |
},
|
| 1180 |
{
|
| 1181 |
"epoch": 8.03,
|
| 1182 |
+
"learning_rate": 9.19718309859155e-06,
|
| 1183 |
+
"loss": 0.006,
|
| 1184 |
"step": 1710
|
| 1185 |
},
|
| 1186 |
{
|
| 1187 |
"epoch": 8.08,
|
| 1188 |
+
"learning_rate": 9.1924882629108e-06,
|
| 1189 |
+
"loss": 0.0065,
|
| 1190 |
"step": 1720
|
| 1191 |
},
|
| 1192 |
{
|
| 1193 |
"epoch": 8.12,
|
| 1194 |
+
"learning_rate": 9.187793427230047e-06,
|
| 1195 |
+
"loss": 0.014,
|
| 1196 |
"step": 1730
|
| 1197 |
},
|
| 1198 |
{
|
| 1199 |
"epoch": 8.17,
|
| 1200 |
+
"learning_rate": 9.183098591549296e-06,
|
| 1201 |
+
"loss": 0.0077,
|
| 1202 |
"step": 1740
|
| 1203 |
},
|
| 1204 |
{
|
| 1205 |
"epoch": 8.22,
|
| 1206 |
+
"learning_rate": 9.178403755868546e-06,
|
| 1207 |
+
"loss": 0.006,
|
| 1208 |
"step": 1750
|
| 1209 |
},
|
| 1210 |
{
|
| 1211 |
"epoch": 8.26,
|
| 1212 |
+
"learning_rate": 9.173708920187793e-06,
|
| 1213 |
+
"loss": 0.0062,
|
| 1214 |
"step": 1760
|
| 1215 |
},
|
| 1216 |
{
|
| 1217 |
"epoch": 8.31,
|
| 1218 |
+
"learning_rate": 9.169014084507044e-06,
|
| 1219 |
+
"loss": 0.0082,
|
| 1220 |
"step": 1770
|
| 1221 |
},
|
| 1222 |
{
|
| 1223 |
"epoch": 8.36,
|
| 1224 |
+
"learning_rate": 9.164319248826293e-06,
|
| 1225 |
+
"loss": 0.0064,
|
| 1226 |
"step": 1780
|
| 1227 |
},
|
| 1228 |
{
|
| 1229 |
"epoch": 8.4,
|
| 1230 |
+
"learning_rate": 9.15962441314554e-06,
|
| 1231 |
+
"loss": 0.0089,
|
| 1232 |
"step": 1790
|
| 1233 |
},
|
| 1234 |
{
|
| 1235 |
"epoch": 8.45,
|
| 1236 |
+
"learning_rate": 9.15492957746479e-06,
|
| 1237 |
+
"loss": 0.004,
|
| 1238 |
"step": 1800
|
| 1239 |
},
|
| 1240 |
{
|
| 1241 |
"epoch": 8.5,
|
| 1242 |
+
"learning_rate": 9.150234741784039e-06,
|
| 1243 |
+
"loss": 0.0082,
|
| 1244 |
"step": 1810
|
| 1245 |
},
|
| 1246 |
{
|
| 1247 |
"epoch": 8.54,
|
| 1248 |
+
"learning_rate": 9.145539906103286e-06,
|
| 1249 |
+
"loss": 0.0129,
|
| 1250 |
"step": 1820
|
| 1251 |
},
|
| 1252 |
{
|
| 1253 |
"epoch": 8.59,
|
| 1254 |
+
"learning_rate": 9.140845070422536e-06,
|
| 1255 |
+
"loss": 0.0141,
|
| 1256 |
"step": 1830
|
| 1257 |
},
|
| 1258 |
{
|
| 1259 |
"epoch": 8.64,
|
| 1260 |
+
"learning_rate": 9.136150234741785e-06,
|
| 1261 |
+
"loss": 0.0121,
|
| 1262 |
"step": 1840
|
| 1263 |
},
|
| 1264 |
{
|
| 1265 |
"epoch": 8.69,
|
| 1266 |
+
"learning_rate": 9.131455399061034e-06,
|
| 1267 |
+
"loss": 0.0054,
|
| 1268 |
"step": 1850
|
| 1269 |
},
|
| 1270 |
{
|
| 1271 |
"epoch": 8.73,
|
| 1272 |
+
"learning_rate": 9.126760563380283e-06,
|
| 1273 |
+
"loss": 0.0136,
|
| 1274 |
"step": 1860
|
| 1275 |
},
|
| 1276 |
{
|
| 1277 |
"epoch": 8.78,
|
| 1278 |
+
"learning_rate": 9.122065727699532e-06,
|
| 1279 |
+
"loss": 0.0036,
|
| 1280 |
"step": 1870
|
| 1281 |
},
|
| 1282 |
{
|
| 1283 |
"epoch": 8.83,
|
| 1284 |
+
"learning_rate": 9.11737089201878e-06,
|
| 1285 |
+
"loss": 0.0065,
|
| 1286 |
"step": 1880
|
| 1287 |
},
|
| 1288 |
{
|
| 1289 |
"epoch": 8.87,
|
| 1290 |
+
"learning_rate": 9.11267605633803e-06,
|
| 1291 |
+
"loss": 0.0079,
|
| 1292 |
"step": 1890
|
| 1293 |
},
|
| 1294 |
{
|
| 1295 |
"epoch": 8.92,
|
| 1296 |
+
"learning_rate": 9.107981220657278e-06,
|
| 1297 |
+
"loss": 0.0154,
|
| 1298 |
"step": 1900
|
| 1299 |
},
|
| 1300 |
{
|
| 1301 |
"epoch": 8.97,
|
| 1302 |
+
"learning_rate": 9.103286384976527e-06,
|
| 1303 |
+
"loss": 0.0097,
|
| 1304 |
"step": 1910
|
| 1305 |
},
|
| 1306 |
{
|
| 1307 |
"epoch": 9.0,
|
| 1308 |
+
"eval_corporation_f1": 0.3111111111111111,
|
| 1309 |
+
"eval_creative-work_f1": 0.46093750000000006,
|
| 1310 |
+
"eval_group_f1": 0.3835616438356164,
|
| 1311 |
+
"eval_location_f1": 0.7089947089947088,
|
| 1312 |
+
"eval_loss": 0.30845504999160767,
|
| 1313 |
+
"eval_overall_accuracy": 0.9515515396504577,
|
| 1314 |
+
"eval_overall_f1": 0.6495149725854069,
|
| 1315 |
+
"eval_overall_precision": 0.6598114824335904,
|
| 1316 |
+
"eval_overall_recall": 0.6395348837209303,
|
| 1317 |
+
"eval_person_f1": 0.7905604719764011,
|
| 1318 |
+
"eval_product_f1": 0.40830449826989623,
|
| 1319 |
+
"eval_runtime": 5.2314,
|
| 1320 |
+
"eval_samples_per_second": 192.875,
|
| 1321 |
+
"eval_steps_per_second": 12.234,
|
| 1322 |
"step": 1917
|
| 1323 |
},
|
| 1324 |
{
|
| 1325 |
"epoch": 9.01,
|
| 1326 |
+
"learning_rate": 9.098591549295776e-06,
|
| 1327 |
+
"loss": 0.005,
|
| 1328 |
"step": 1920
|
| 1329 |
},
|
| 1330 |
{
|
| 1331 |
"epoch": 9.06,
|
| 1332 |
+
"learning_rate": 9.093896713615024e-06,
|
| 1333 |
+
"loss": 0.0036,
|
| 1334 |
"step": 1930
|
| 1335 |
},
|
| 1336 |
{
|
| 1337 |
"epoch": 9.11,
|
| 1338 |
+
"learning_rate": 9.089201877934273e-06,
|
| 1339 |
+
"loss": 0.0157,
|
| 1340 |
"step": 1940
|
| 1341 |
},
|
| 1342 |
{
|
| 1343 |
"epoch": 9.15,
|
| 1344 |
+
"learning_rate": 9.084507042253522e-06,
|
| 1345 |
+
"loss": 0.0044,
|
| 1346 |
"step": 1950
|
| 1347 |
},
|
| 1348 |
{
|
| 1349 |
"epoch": 9.2,
|
| 1350 |
+
"learning_rate": 9.07981220657277e-06,
|
| 1351 |
+
"loss": 0.0139,
|
| 1352 |
"step": 1960
|
| 1353 |
},
|
| 1354 |
{
|
| 1355 |
"epoch": 9.25,
|
| 1356 |
+
"learning_rate": 9.07511737089202e-06,
|
| 1357 |
+
"loss": 0.0135,
|
| 1358 |
"step": 1970
|
| 1359 |
},
|
| 1360 |
{
|
| 1361 |
"epoch": 9.3,
|
| 1362 |
+
"learning_rate": 9.070422535211268e-06,
|
| 1363 |
+
"loss": 0.003,
|
| 1364 |
"step": 1980
|
| 1365 |
},
|
| 1366 |
{
|
| 1367 |
"epoch": 9.34,
|
| 1368 |
+
"learning_rate": 9.065727699530517e-06,
|
| 1369 |
+
"loss": 0.0078,
|
| 1370 |
"step": 1990
|
| 1371 |
},
|
| 1372 |
{
|
| 1373 |
"epoch": 9.39,
|
| 1374 |
+
"learning_rate": 9.061032863849766e-06,
|
| 1375 |
+
"loss": 0.0052,
|
| 1376 |
"step": 2000
|
| 1377 |
},
|
| 1378 |
{
|
| 1379 |
"epoch": 9.44,
|
| 1380 |
+
"learning_rate": 9.056338028169015e-06,
|
| 1381 |
+
"loss": 0.0043,
|
| 1382 |
"step": 2010
|
| 1383 |
},
|
| 1384 |
{
|
| 1385 |
"epoch": 9.48,
|
| 1386 |
+
"learning_rate": 9.051643192488264e-06,
|
| 1387 |
+
"loss": 0.0075,
|
| 1388 |
"step": 2020
|
| 1389 |
},
|
| 1390 |
{
|
| 1391 |
"epoch": 9.53,
|
| 1392 |
+
"learning_rate": 9.046948356807512e-06,
|
| 1393 |
+
"loss": 0.0041,
|
| 1394 |
"step": 2030
|
| 1395 |
},
|
| 1396 |
{
|
| 1397 |
"epoch": 9.58,
|
| 1398 |
+
"learning_rate": 9.042253521126761e-06,
|
| 1399 |
+
"loss": 0.0042,
|
| 1400 |
"step": 2040
|
| 1401 |
},
|
| 1402 |
{
|
| 1403 |
"epoch": 9.62,
|
| 1404 |
+
"learning_rate": 9.03755868544601e-06,
|
| 1405 |
+
"loss": 0.0064,
|
| 1406 |
"step": 2050
|
| 1407 |
},
|
| 1408 |
{
|
| 1409 |
"epoch": 9.67,
|
| 1410 |
+
"learning_rate": 9.032863849765259e-06,
|
| 1411 |
+
"loss": 0.0045,
|
| 1412 |
"step": 2060
|
| 1413 |
},
|
| 1414 |
{
|
| 1415 |
"epoch": 9.72,
|
| 1416 |
+
"learning_rate": 9.028169014084507e-06,
|
| 1417 |
+
"loss": 0.0074,
|
| 1418 |
"step": 2070
|
| 1419 |
},
|
| 1420 |
{
|
| 1421 |
"epoch": 9.77,
|
| 1422 |
+
"learning_rate": 9.023474178403756e-06,
|
| 1423 |
+
"loss": 0.0048,
|
| 1424 |
"step": 2080
|
| 1425 |
},
|
| 1426 |
{
|
| 1427 |
"epoch": 9.81,
|
| 1428 |
+
"learning_rate": 9.018779342723005e-06,
|
| 1429 |
+
"loss": 0.0027,
|
| 1430 |
"step": 2090
|
| 1431 |
},
|
| 1432 |
{
|
| 1433 |
"epoch": 9.86,
|
| 1434 |
+
"learning_rate": 9.014084507042254e-06,
|
| 1435 |
+
"loss": 0.0028,
|
| 1436 |
"step": 2100
|
| 1437 |
},
|
| 1438 |
{
|
| 1439 |
"epoch": 9.91,
|
| 1440 |
+
"learning_rate": 9.009389671361503e-06,
|
| 1441 |
+
"loss": 0.0096,
|
| 1442 |
"step": 2110
|
| 1443 |
},
|
| 1444 |
{
|
| 1445 |
"epoch": 9.95,
|
| 1446 |
+
"learning_rate": 9.004694835680751e-06,
|
| 1447 |
+
"loss": 0.0058,
|
| 1448 |
"step": 2120
|
| 1449 |
},
|
| 1450 |
{
|
| 1451 |
"epoch": 10.0,
|
| 1452 |
+
"learning_rate": 9e-06,
|
| 1453 |
+
"loss": 0.0108,
|
| 1454 |
"step": 2130
|
| 1455 |
},
|
| 1456 |
{
|
| 1457 |
"epoch": 10.0,
|
| 1458 |
+
"eval_corporation_f1": 0.35294117647058826,
|
| 1459 |
+
"eval_creative-work_f1": 0.45801526717557256,
|
| 1460 |
+
"eval_group_f1": 0.36486486486486486,
|
| 1461 |
+
"eval_location_f1": 0.689655172413793,
|
| 1462 |
+
"eval_loss": 0.30454060435295105,
|
| 1463 |
+
"eval_overall_accuracy": 0.950897634050648,
|
| 1464 |
+
"eval_overall_f1": 0.6540880503144654,
|
| 1465 |
+
"eval_overall_precision": 0.6604572396274344,
|
| 1466 |
+
"eval_overall_recall": 0.6478405315614618,
|
| 1467 |
+
"eval_person_f1": 0.784313725490196,
|
| 1468 |
+
"eval_product_f1": 0.43870967741935485,
|
| 1469 |
+
"eval_runtime": 6.5888,
|
| 1470 |
+
"eval_samples_per_second": 153.138,
|
| 1471 |
+
"eval_steps_per_second": 9.713,
|
| 1472 |
"step": 2130
|
| 1473 |
},
|
| 1474 |
{
|
| 1475 |
"epoch": 10.05,
|
| 1476 |
+
"learning_rate": 8.995305164319249e-06,
|
| 1477 |
+
"loss": 0.0035,
|
| 1478 |
"step": 2140
|
| 1479 |
},
|
| 1480 |
{
|
| 1481 |
"epoch": 10.09,
|
| 1482 |
+
"learning_rate": 8.990610328638498e-06,
|
| 1483 |
+
"loss": 0.003,
|
| 1484 |
"step": 2150
|
| 1485 |
},
|
| 1486 |
{
|
| 1487 |
"epoch": 10.14,
|
| 1488 |
+
"learning_rate": 8.985915492957748e-06,
|
| 1489 |
+
"loss": 0.0117,
|
| 1490 |
"step": 2160
|
| 1491 |
},
|
| 1492 |
{
|
| 1493 |
"epoch": 10.19,
|
| 1494 |
+
"learning_rate": 8.981220657276995e-06,
|
| 1495 |
+
"loss": 0.004,
|
| 1496 |
"step": 2170
|
| 1497 |
},
|
| 1498 |
{
|
| 1499 |
"epoch": 10.23,
|
| 1500 |
+
"learning_rate": 8.976525821596244e-06,
|
| 1501 |
+
"loss": 0.0038,
|
| 1502 |
"step": 2180
|
| 1503 |
},
|
| 1504 |
{
|
| 1505 |
"epoch": 10.28,
|
| 1506 |
+
"learning_rate": 8.971830985915495e-06,
|
| 1507 |
+
"loss": 0.0057,
|
| 1508 |
"step": 2190
|
| 1509 |
},
|
| 1510 |
{
|
| 1511 |
"epoch": 10.33,
|
| 1512 |
+
"learning_rate": 8.967136150234742e-06,
|
| 1513 |
+
"loss": 0.0025,
|
| 1514 |
"step": 2200
|
| 1515 |
},
|
| 1516 |
{
|
| 1517 |
"epoch": 10.38,
|
| 1518 |
+
"learning_rate": 8.96244131455399e-06,
|
| 1519 |
+
"loss": 0.0035,
|
| 1520 |
"step": 2210
|
| 1521 |
},
|
| 1522 |
{
|
| 1523 |
"epoch": 10.42,
|
| 1524 |
+
"learning_rate": 8.957746478873241e-06,
|
| 1525 |
+
"loss": 0.008,
|
| 1526 |
"step": 2220
|
| 1527 |
},
|
| 1528 |
{
|
| 1529 |
"epoch": 10.47,
|
| 1530 |
+
"learning_rate": 8.953051643192488e-06,
|
| 1531 |
+
"loss": 0.0025,
|
| 1532 |
"step": 2230
|
| 1533 |
},
|
| 1534 |
{
|
| 1535 |
"epoch": 10.52,
|
| 1536 |
+
"learning_rate": 8.948356807511737e-06,
|
| 1537 |
+
"loss": 0.0023,
|
| 1538 |
"step": 2240
|
| 1539 |
},
|
| 1540 |
{
|
| 1541 |
"epoch": 10.56,
|
| 1542 |
+
"learning_rate": 8.943661971830987e-06,
|
| 1543 |
+
"loss": 0.0086,
|
| 1544 |
"step": 2250
|
| 1545 |
},
|
| 1546 |
{
|
| 1547 |
"epoch": 10.61,
|
| 1548 |
+
"learning_rate": 8.938967136150235e-06,
|
| 1549 |
+
"loss": 0.0101,
|
| 1550 |
"step": 2260
|
| 1551 |
},
|
| 1552 |
{
|
| 1553 |
"epoch": 10.66,
|
| 1554 |
+
"learning_rate": 8.934272300469485e-06,
|
| 1555 |
+
"loss": 0.0072,
|
| 1556 |
"step": 2270
|
| 1557 |
},
|
| 1558 |
{
|
| 1559 |
"epoch": 10.7,
|
| 1560 |
+
"learning_rate": 8.929577464788734e-06,
|
| 1561 |
+
"loss": 0.004,
|
| 1562 |
"step": 2280
|
| 1563 |
},
|
| 1564 |
{
|
| 1565 |
"epoch": 10.75,
|
| 1566 |
+
"learning_rate": 8.924882629107981e-06,
|
| 1567 |
+
"loss": 0.0026,
|
| 1568 |
"step": 2290
|
| 1569 |
},
|
| 1570 |
{
|
| 1571 |
"epoch": 10.8,
|
| 1572 |
+
"learning_rate": 8.920187793427231e-06,
|
| 1573 |
+
"loss": 0.0047,
|
| 1574 |
"step": 2300
|
| 1575 |
},
|
| 1576 |
{
|
| 1577 |
"epoch": 10.85,
|
| 1578 |
+
"learning_rate": 8.91549295774648e-06,
|
| 1579 |
+
"loss": 0.002,
|
| 1580 |
"step": 2310
|
| 1581 |
},
|
| 1582 |
{
|
| 1583 |
"epoch": 10.89,
|
| 1584 |
+
"learning_rate": 8.910798122065727e-06,
|
| 1585 |
+
"loss": 0.0116,
|
| 1586 |
"step": 2320
|
| 1587 |
},
|
| 1588 |
{
|
| 1589 |
"epoch": 10.94,
|
| 1590 |
+
"learning_rate": 8.906103286384978e-06,
|
| 1591 |
+
"loss": 0.0057,
|
| 1592 |
"step": 2330
|
| 1593 |
},
|
| 1594 |
{
|
| 1595 |
"epoch": 10.99,
|
| 1596 |
+
"learning_rate": 8.901408450704227e-06,
|
| 1597 |
+
"loss": 0.013,
|
| 1598 |
"step": 2340
|
| 1599 |
},
|
| 1600 |
{
|
| 1601 |
"epoch": 11.0,
|
| 1602 |
+
"eval_corporation_f1": 0.2782608695652174,
|
| 1603 |
+
"eval_creative-work_f1": 0.4247787610619469,
|
| 1604 |
+
"eval_group_f1": 0.3357664233576642,
|
| 1605 |
+
"eval_location_f1": 0.736842105263158,
|
| 1606 |
+
"eval_loss": 0.3382622003555298,
|
| 1607 |
+
"eval_overall_accuracy": 0.9507192961597908,
|
| 1608 |
+
"eval_overall_f1": 0.6469565217391304,
|
| 1609 |
+
"eval_overall_precision": 0.6788321167883211,
|
| 1610 |
+
"eval_overall_recall": 0.6179401993355482,
|
| 1611 |
+
"eval_person_f1": 0.7958271236959762,
|
| 1612 |
+
"eval_product_f1": 0.3655172413793103,
|
| 1613 |
+
"eval_runtime": 5.3383,
|
| 1614 |
+
"eval_samples_per_second": 189.013,
|
| 1615 |
+
"eval_steps_per_second": 11.989,
|
| 1616 |
"step": 2343
|
| 1617 |
},
|
| 1618 |
{
|
| 1619 |
"epoch": 11.03,
|
| 1620 |
+
"learning_rate": 8.896713615023475e-06,
|
| 1621 |
+
"loss": 0.0059,
|
| 1622 |
"step": 2350
|
| 1623 |
},
|
| 1624 |
{
|
| 1625 |
"epoch": 11.08,
|
| 1626 |
+
"learning_rate": 8.892018779342724e-06,
|
| 1627 |
+
"loss": 0.0153,
|
| 1628 |
"step": 2360
|
| 1629 |
},
|
| 1630 |
{
|
| 1631 |
"epoch": 11.13,
|
| 1632 |
+
"learning_rate": 8.887323943661973e-06,
|
| 1633 |
+
"loss": 0.0057,
|
| 1634 |
"step": 2370
|
| 1635 |
},
|
| 1636 |
{
|
| 1637 |
"epoch": 11.17,
|
| 1638 |
+
"learning_rate": 8.882629107981222e-06,
|
| 1639 |
+
"loss": 0.0048,
|
| 1640 |
"step": 2380
|
| 1641 |
},
|
| 1642 |
{
|
| 1643 |
"epoch": 11.22,
|
| 1644 |
+
"learning_rate": 8.87793427230047e-06,
|
| 1645 |
+
"loss": 0.0118,
|
| 1646 |
"step": 2390
|
| 1647 |
},
|
| 1648 |
{
|
| 1649 |
"epoch": 11.27,
|
| 1650 |
+
"learning_rate": 8.87323943661972e-06,
|
| 1651 |
+
"loss": 0.0071,
|
| 1652 |
"step": 2400
|
| 1653 |
},
|
| 1654 |
{
|
| 1655 |
"epoch": 11.31,
|
| 1656 |
+
"learning_rate": 8.868544600938968e-06,
|
| 1657 |
+
"loss": 0.0098,
|
| 1658 |
"step": 2410
|
| 1659 |
},
|
| 1660 |
{
|
| 1661 |
"epoch": 11.36,
|
| 1662 |
+
"learning_rate": 8.863849765258217e-06,
|
| 1663 |
+
"loss": 0.0065,
|
| 1664 |
"step": 2420
|
| 1665 |
},
|
| 1666 |
{
|
| 1667 |
"epoch": 11.41,
|
| 1668 |
+
"learning_rate": 8.859154929577466e-06,
|
| 1669 |
+
"loss": 0.0051,
|
| 1670 |
"step": 2430
|
| 1671 |
},
|
| 1672 |
{
|
| 1673 |
"epoch": 11.46,
|
| 1674 |
+
"learning_rate": 8.854460093896714e-06,
|
| 1675 |
+
"loss": 0.0053,
|
| 1676 |
"step": 2440
|
| 1677 |
},
|
| 1678 |
{
|
| 1679 |
"epoch": 11.5,
|
| 1680 |
+
"learning_rate": 8.849765258215963e-06,
|
| 1681 |
+
"loss": 0.0024,
|
| 1682 |
"step": 2450
|
| 1683 |
},
|
| 1684 |
{
|
| 1685 |
"epoch": 11.55,
|
| 1686 |
+
"learning_rate": 8.845070422535212e-06,
|
| 1687 |
+
"loss": 0.0045,
|
| 1688 |
"step": 2460
|
| 1689 |
},
|
| 1690 |
{
|
| 1691 |
"epoch": 11.6,
|
| 1692 |
+
"learning_rate": 8.84037558685446e-06,
|
| 1693 |
+
"loss": 0.0172,
|
| 1694 |
"step": 2470
|
| 1695 |
},
|
| 1696 |
{
|
| 1697 |
"epoch": 11.64,
|
| 1698 |
+
"learning_rate": 8.83568075117371e-06,
|
| 1699 |
+
"loss": 0.0053,
|
| 1700 |
"step": 2480
|
| 1701 |
},
|
| 1702 |
{
|
| 1703 |
"epoch": 11.69,
|
| 1704 |
+
"learning_rate": 8.830985915492958e-06,
|
| 1705 |
+
"loss": 0.014,
|
| 1706 |
"step": 2490
|
| 1707 |
},
|
| 1708 |
{
|
| 1709 |
"epoch": 11.74,
|
| 1710 |
+
"learning_rate": 8.826291079812207e-06,
|
| 1711 |
+
"loss": 0.0178,
|
| 1712 |
"step": 2500
|
| 1713 |
},
|
| 1714 |
{
|
| 1715 |
"epoch": 11.78,
|
| 1716 |
+
"learning_rate": 8.821596244131456e-06,
|
| 1717 |
+
"loss": 0.0071,
|
| 1718 |
"step": 2510
|
| 1719 |
},
|
| 1720 |
{
|
| 1721 |
"epoch": 11.83,
|
| 1722 |
+
"learning_rate": 8.816901408450705e-06,
|
| 1723 |
+
"loss": 0.0086,
|
| 1724 |
"step": 2520
|
| 1725 |
},
|
| 1726 |
{
|
| 1727 |
"epoch": 11.88,
|
| 1728 |
+
"learning_rate": 8.812206572769954e-06,
|
| 1729 |
+
"loss": 0.0056,
|
| 1730 |
"step": 2530
|
| 1731 |
},
|
| 1732 |
{
|
| 1733 |
"epoch": 11.92,
|
| 1734 |
+
"learning_rate": 8.807511737089202e-06,
|
| 1735 |
+
"loss": 0.0029,
|
| 1736 |
"step": 2540
|
| 1737 |
},
|
| 1738 |
{
|
| 1739 |
"epoch": 11.97,
|
| 1740 |
+
"learning_rate": 8.802816901408451e-06,
|
| 1741 |
+
"loss": 0.0076,
|
| 1742 |
"step": 2550
|
| 1743 |
},
|
| 1744 |
{
|
| 1745 |
"epoch": 12.0,
|
| 1746 |
+
"eval_corporation_f1": 0.2708333333333333,
|
| 1747 |
+
"eval_creative-work_f1": 0.39852398523985244,
|
| 1748 |
+
"eval_group_f1": 0.33333333333333337,
|
| 1749 |
+
"eval_location_f1": 0.6740331491712707,
|
| 1750 |
+
"eval_loss": 0.36174312233924866,
|
| 1751 |
+
"eval_overall_accuracy": 0.9473903221971228,
|
| 1752 |
+
"eval_overall_f1": 0.6143187066974597,
|
| 1753 |
+
"eval_overall_precision": 0.6919875130072841,
|
| 1754 |
+
"eval_overall_recall": 0.5523255813953488,
|
| 1755 |
+
"eval_person_f1": 0.75658419792498,
|
| 1756 |
+
"eval_product_f1": 0.3524590163934426,
|
| 1757 |
+
"eval_runtime": 5.9165,
|
| 1758 |
+
"eval_samples_per_second": 170.539,
|
| 1759 |
+
"eval_steps_per_second": 10.817,
|
| 1760 |
"step": 2556
|
| 1761 |
},
|
| 1762 |
{
|
| 1763 |
"epoch": 12.02,
|
| 1764 |
+
"learning_rate": 8.7981220657277e-06,
|
| 1765 |
+
"loss": 0.0034,
|
| 1766 |
"step": 2560
|
| 1767 |
},
|
| 1768 |
{
|
| 1769 |
"epoch": 12.07,
|
| 1770 |
+
"learning_rate": 8.793427230046949e-06,
|
| 1771 |
+
"loss": 0.0029,
|
| 1772 |
"step": 2570
|
| 1773 |
},
|
| 1774 |
{
|
| 1775 |
"epoch": 12.11,
|
| 1776 |
+
"learning_rate": 8.7887323943662e-06,
|
| 1777 |
+
"loss": 0.0084,
|
| 1778 |
"step": 2580
|
| 1779 |
},
|
| 1780 |
{
|
| 1781 |
"epoch": 12.16,
|
| 1782 |
+
"learning_rate": 8.784037558685446e-06,
|
| 1783 |
+
"loss": 0.0058,
|
| 1784 |
"step": 2590
|
| 1785 |
},
|
| 1786 |
{
|
| 1787 |
"epoch": 12.21,
|
| 1788 |
+
"learning_rate": 8.779342723004695e-06,
|
| 1789 |
+
"loss": 0.0025,
|
| 1790 |
"step": 2600
|
| 1791 |
},
|
| 1792 |
{
|
| 1793 |
"epoch": 12.25,
|
| 1794 |
+
"learning_rate": 8.774647887323946e-06,
|
| 1795 |
+
"loss": 0.0094,
|
| 1796 |
"step": 2610
|
| 1797 |
},
|
| 1798 |
{
|
| 1799 |
"epoch": 12.3,
|
| 1800 |
+
"learning_rate": 8.769953051643193e-06,
|
| 1801 |
+
"loss": 0.0051,
|
| 1802 |
"step": 2620
|
| 1803 |
},
|
| 1804 |
{
|
| 1805 |
"epoch": 12.35,
|
| 1806 |
+
"learning_rate": 8.765258215962442e-06,
|
| 1807 |
+
"loss": 0.0064,
|
| 1808 |
"step": 2630
|
| 1809 |
},
|
| 1810 |
{
|
| 1811 |
"epoch": 12.39,
|
| 1812 |
+
"learning_rate": 8.760563380281692e-06,
|
| 1813 |
+
"loss": 0.0095,
|
| 1814 |
"step": 2640
|
| 1815 |
},
|
| 1816 |
{
|
| 1817 |
"epoch": 12.44,
|
| 1818 |
+
"learning_rate": 8.755868544600939e-06,
|
| 1819 |
+
"loss": 0.0049,
|
| 1820 |
"step": 2650
|
| 1821 |
},
|
| 1822 |
{
|
| 1823 |
"epoch": 12.49,
|
| 1824 |
+
"learning_rate": 8.751173708920188e-06,
|
| 1825 |
+
"loss": 0.0114,
|
| 1826 |
"step": 2660
|
| 1827 |
},
|
| 1828 |
{
|
| 1829 |
"epoch": 12.54,
|
| 1830 |
+
"learning_rate": 8.746478873239437e-06,
|
| 1831 |
+
"loss": 0.0054,
|
| 1832 |
"step": 2670
|
| 1833 |
},
|
| 1834 |
{
|
| 1835 |
"epoch": 12.58,
|
| 1836 |
+
"learning_rate": 8.741784037558685e-06,
|
| 1837 |
+
"loss": 0.0108,
|
| 1838 |
"step": 2680
|
| 1839 |
},
|
| 1840 |
{
|
| 1841 |
"epoch": 12.63,
|
| 1842 |
+
"learning_rate": 8.737089201877936e-06,
|
| 1843 |
+
"loss": 0.0052,
|
| 1844 |
"step": 2690
|
| 1845 |
},
|
| 1846 |
{
|
| 1847 |
"epoch": 12.68,
|
| 1848 |
+
"learning_rate": 8.732394366197183e-06,
|
| 1849 |
+
"loss": 0.0086,
|
| 1850 |
"step": 2700
|
| 1851 |
},
|
| 1852 |
{
|
| 1853 |
"epoch": 12.72,
|
| 1854 |
+
"learning_rate": 8.727699530516432e-06,
|
| 1855 |
+
"loss": 0.0019,
|
| 1856 |
"step": 2710
|
| 1857 |
},
|
| 1858 |
{
|
| 1859 |
"epoch": 12.77,
|
| 1860 |
+
"learning_rate": 8.723004694835682e-06,
|
| 1861 |
+
"loss": 0.0024,
|
| 1862 |
"step": 2720
|
| 1863 |
},
|
| 1864 |
{
|
| 1865 |
"epoch": 12.82,
|
| 1866 |
+
"learning_rate": 8.71830985915493e-06,
|
| 1867 |
+
"loss": 0.0038,
|
| 1868 |
"step": 2730
|
| 1869 |
},
|
| 1870 |
{
|
| 1871 |
"epoch": 12.86,
|
| 1872 |
+
"learning_rate": 8.713615023474178e-06,
|
| 1873 |
+
"loss": 0.0097,
|
| 1874 |
"step": 2740
|
| 1875 |
},
|
| 1876 |
{
|
| 1877 |
"epoch": 12.91,
|
| 1878 |
+
"learning_rate": 8.708920187793429e-06,
|
| 1879 |
+
"loss": 0.0054,
|
| 1880 |
"step": 2750
|
| 1881 |
},
|
| 1882 |
{
|
| 1883 |
"epoch": 12.96,
|
| 1884 |
+
"learning_rate": 8.704225352112676e-06,
|
| 1885 |
+
"loss": 0.0042,
|
| 1886 |
"step": 2760
|
| 1887 |
},
|
| 1888 |
{
|
| 1889 |
"epoch": 13.0,
|
| 1890 |
+
"eval_corporation_f1": 0.24778761061946902,
|
| 1891 |
+
"eval_creative-work_f1": 0.39148936170212767,
|
| 1892 |
+
"eval_group_f1": 0.35211267605633806,
|
| 1893 |
+
"eval_location_f1": 0.656084656084656,
|
| 1894 |
+
"eval_loss": 0.3746570348739624,
|
| 1895 |
+
"eval_overall_accuracy": 0.9472714302698847,
|
| 1896 |
+
"eval_overall_f1": 0.6219790241678067,
|
| 1897 |
+
"eval_overall_precision": 0.6895854398382204,
|
| 1898 |
+
"eval_overall_recall": 0.5664451827242525,
|
| 1899 |
+
"eval_person_f1": 0.7741935483870966,
|
| 1900 |
+
"eval_product_f1": 0.35390946502057613,
|
| 1901 |
+
"eval_runtime": 6.4418,
|
| 1902 |
+
"eval_samples_per_second": 156.633,
|
| 1903 |
+
"eval_steps_per_second": 9.935,
|
| 1904 |
"step": 2769
|
| 1905 |
},
|
| 1906 |
{
|
| 1907 |
"epoch": 13.0,
|
| 1908 |
+
"learning_rate": 8.699530516431926e-06,
|
| 1909 |
+
"loss": 0.0039,
|
| 1910 |
"step": 2770
|
| 1911 |
},
|
| 1912 |
{
|
| 1913 |
"epoch": 13.05,
|
| 1914 |
+
"learning_rate": 8.694835680751175e-06,
|
| 1915 |
+
"loss": 0.0017,
|
| 1916 |
"step": 2780
|
| 1917 |
},
|
| 1918 |
{
|
| 1919 |
"epoch": 13.1,
|
| 1920 |
+
"learning_rate": 8.690140845070422e-06,
|
| 1921 |
+
"loss": 0.0102,
|
| 1922 |
"step": 2790
|
| 1923 |
},
|
| 1924 |
{
|
| 1925 |
"epoch": 13.15,
|
| 1926 |
+
"learning_rate": 8.685446009389673e-06,
|
| 1927 |
+
"loss": 0.0013,
|
| 1928 |
"step": 2800
|
| 1929 |
},
|
| 1930 |
{
|
| 1931 |
"epoch": 13.19,
|
| 1932 |
+
"learning_rate": 8.680751173708921e-06,
|
| 1933 |
+
"loss": 0.0076,
|
| 1934 |
"step": 2810
|
| 1935 |
},
|
| 1936 |
{
|
| 1937 |
"epoch": 13.24,
|
| 1938 |
+
"learning_rate": 8.676056338028169e-06,
|
| 1939 |
+
"loss": 0.0108,
|
| 1940 |
"step": 2820
|
| 1941 |
},
|
| 1942 |
{
|
| 1943 |
"epoch": 13.29,
|
| 1944 |
+
"learning_rate": 8.671361502347419e-06,
|
| 1945 |
+
"loss": 0.0107,
|
| 1946 |
"step": 2830
|
| 1947 |
},
|
| 1948 |
{
|
| 1949 |
"epoch": 13.33,
|
| 1950 |
+
"learning_rate": 8.666666666666668e-06,
|
| 1951 |
+
"loss": 0.0058,
|
| 1952 |
"step": 2840
|
| 1953 |
},
|
| 1954 |
{
|
| 1955 |
"epoch": 13.38,
|
| 1956 |
+
"learning_rate": 8.661971830985915e-06,
|
| 1957 |
+
"loss": 0.0033,
|
| 1958 |
"step": 2850
|
| 1959 |
},
|
| 1960 |
{
|
| 1961 |
"epoch": 13.43,
|
| 1962 |
+
"learning_rate": 8.657276995305165e-06,
|
| 1963 |
+
"loss": 0.0028,
|
| 1964 |
"step": 2860
|
| 1965 |
},
|
| 1966 |
{
|
| 1967 |
"epoch": 13.47,
|
| 1968 |
+
"learning_rate": 8.652582159624414e-06,
|
| 1969 |
+
"loss": 0.0046,
|
| 1970 |
"step": 2870
|
| 1971 |
},
|
| 1972 |
{
|
| 1973 |
"epoch": 13.52,
|
| 1974 |
+
"learning_rate": 8.647887323943663e-06,
|
| 1975 |
+
"loss": 0.002,
|
| 1976 |
"step": 2880
|
| 1977 |
},
|
| 1978 |
{
|
| 1979 |
"epoch": 13.57,
|
| 1980 |
+
"learning_rate": 8.643192488262912e-06,
|
| 1981 |
+
"loss": 0.0042,
|
| 1982 |
"step": 2890
|
| 1983 |
},
|
| 1984 |
{
|
| 1985 |
"epoch": 13.62,
|
| 1986 |
+
"learning_rate": 8.63849765258216e-06,
|
| 1987 |
+
"loss": 0.0066,
|
| 1988 |
"step": 2900
|
| 1989 |
},
|
| 1990 |
{
|
| 1991 |
"epoch": 13.66,
|
| 1992 |
+
"learning_rate": 8.63380281690141e-06,
|
| 1993 |
+
"loss": 0.0022,
|
| 1994 |
"step": 2910
|
| 1995 |
},
|
| 1996 |
{
|
| 1997 |
"epoch": 13.71,
|
| 1998 |
+
"learning_rate": 8.629107981220658e-06,
|
| 1999 |
+
"loss": 0.008,
|
| 2000 |
"step": 2920
|
| 2001 |
},
|
| 2002 |
{
|
| 2003 |
"epoch": 13.76,
|
| 2004 |
+
"learning_rate": 8.624413145539907e-06,
|
| 2005 |
+
"loss": 0.0037,
|
| 2006 |
"step": 2930
|
| 2007 |
},
|
| 2008 |
{
|
| 2009 |
"epoch": 13.8,
|
| 2010 |
+
"learning_rate": 8.619718309859156e-06,
|
| 2011 |
+
"loss": 0.0106,
|
| 2012 |
"step": 2940
|
| 2013 |
},
|
| 2014 |
{
|
| 2015 |
"epoch": 13.85,
|
| 2016 |
+
"learning_rate": 8.615023474178405e-06,
|
| 2017 |
+
"loss": 0.0031,
|
| 2018 |
"step": 2950
|
| 2019 |
},
|
| 2020 |
{
|
| 2021 |
"epoch": 13.9,
|
| 2022 |
+
"learning_rate": 8.610328638497653e-06,
|
| 2023 |
+
"loss": 0.0023,
|
| 2024 |
"step": 2960
|
| 2025 |
},
|
| 2026 |
{
|
| 2027 |
"epoch": 13.94,
|
| 2028 |
+
"learning_rate": 8.605633802816902e-06,
|
| 2029 |
+
"loss": 0.0118,
|
| 2030 |
"step": 2970
|
| 2031 |
},
|
| 2032 |
{
|
| 2033 |
"epoch": 13.99,
|
| 2034 |
+
"learning_rate": 8.600938967136151e-06,
|
| 2035 |
+
"loss": 0.0049,
|
| 2036 |
"step": 2980
|
| 2037 |
},
|
| 2038 |
{
|
| 2039 |
"epoch": 14.0,
|
| 2040 |
+
"eval_corporation_f1": 0.29752066115702475,
|
| 2041 |
+
"eval_creative-work_f1": 0.44360902255639095,
|
| 2042 |
+
"eval_group_f1": 0.36241610738255037,
|
| 2043 |
+
"eval_location_f1": 0.6834170854271356,
|
| 2044 |
+
"eval_loss": 0.3375568985939026,
|
| 2045 |
+
"eval_overall_accuracy": 0.950897634050648,
|
| 2046 |
+
"eval_overall_f1": 0.6430135786246168,
|
| 2047 |
+
"eval_overall_precision": 0.680259499536608,
|
| 2048 |
+
"eval_overall_recall": 0.6096345514950167,
|
| 2049 |
+
"eval_person_f1": 0.7901614142966948,
|
| 2050 |
+
"eval_product_f1": 0.388663967611336,
|
| 2051 |
+
"eval_runtime": 6.509,
|
| 2052 |
+
"eval_samples_per_second": 155.016,
|
| 2053 |
+
"eval_steps_per_second": 9.833,
|
| 2054 |
"step": 2982
|
| 2055 |
},
|
| 2056 |
{
|
| 2057 |
+
"epoch": 14.0,
|
| 2058 |
+
"step": 2982,
|
| 2059 |
+
"total_flos": 1093917417133104.0,
|
| 2060 |
+
"train_loss": 0.01162932096780006,
|
| 2061 |
+
"train_runtime": 800.644,
|
| 2062 |
+
"train_samples_per_second": 423.909,
|
| 2063 |
+
"train_steps_per_second": 26.604
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2064 |
}
|
| 2065 |
],
|
| 2066 |
"max_steps": 21300,
|
| 2067 |
"num_train_epochs": 100,
|
| 2068 |
+
"total_flos": 1093917417133104.0,
|
| 2069 |
"trial_name": null,
|
| 2070 |
"trial_params": null
|
| 2071 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3119
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9b2b8b76170893b10db52fb47f8d28118527ff73615505009abf7d395b897f6
|
| 3 |
size 3119
|
validation_results.json
CHANGED
|
@@ -1,17 +1,17 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"validation_corporation_f1": 0.
|
| 4 |
-
"validation_creative-work_f1": 0.
|
| 5 |
-
"validation_group_f1": 0.
|
| 6 |
-
"validation_location_f1": 0.
|
| 7 |
-
"validation_loss": 0.
|
| 8 |
-
"validation_overall_accuracy": 0.
|
| 9 |
-
"validation_overall_f1": 0.
|
| 10 |
-
"validation_overall_precision": 0.
|
| 11 |
-
"validation_overall_recall": 0.
|
| 12 |
-
"validation_person_f1": 0.
|
| 13 |
-
"validation_product_f1": 0.
|
| 14 |
-
"validation_runtime":
|
| 15 |
-
"validation_samples_per_second":
|
| 16 |
-
"validation_steps_per_second":
|
| 17 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 14.0,
|
| 3 |
+
"validation_corporation_f1": 0.30769230769230765,
|
| 4 |
+
"validation_creative-work_f1": 0.48421052631578954,
|
| 5 |
+
"validation_group_f1": 0.31666666666666665,
|
| 6 |
+
"validation_location_f1": 0.6808510638297872,
|
| 7 |
+
"validation_loss": 0.27115458250045776,
|
| 8 |
+
"validation_overall_accuracy": 0.9502437284508382,
|
| 9 |
+
"validation_overall_f1": 0.6339784946236559,
|
| 10 |
+
"validation_overall_precision": 0.6574487065120428,
|
| 11 |
+
"validation_overall_recall": 0.6121262458471761,
|
| 12 |
+
"validation_person_f1": 0.7734553775743707,
|
| 13 |
+
"validation_product_f1": 0.3986254295532646,
|
| 14 |
+
"validation_runtime": 5.0688,
|
| 15 |
+
"validation_samples_per_second": 199.06,
|
| 16 |
+
"validation_steps_per_second": 12.626
|
| 17 |
}
|