File size: 3,758 Bytes
acdf4a4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
---
license: apache-2.0
base_model: google/mt5-large
tags:
- generated_from_trainer
model-index:
- name: ner_cs
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# ner_cs

This model is a fine-tuned version of [google/mt5-large](https://huggingface.co/google/mt5-large) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5017
- Loc: {'precision': 0.8522895125553914, 'recall': 0.9058084772370487, 'f1': 0.878234398782344, 'number': 637}
- Org: {'precision': 0.8361702127659575, 'recall': 0.8488120950323974, 'f1': 0.8424437299035369, 'number': 463}
- Per: {'precision': 0.9230769230769231, 'recall': 0.9737470167064439, 'f1': 0.9477351916376306, 'number': 419}
- Overall Precision: 0.8672
- Overall Recall: 0.9072
- Overall F1: 0.8867
- Overall Accuracy: 0.9365

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 20

### Training results

| Training Loss | Epoch | Step  | Validation Loss | Loc                                                                                                      | Org                                                                                                      | Per                                                                                                      | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy |
|:-------------:|:-----:|:-----:|:---------------:|:--------------------------------------------------------------------------------------------------------:|:--------------------------------------------------------------------------------------------------------:|:--------------------------------------------------------------------------------------------------------:|:-----------------:|:--------------:|:----------:|:----------------:|
| 0.2192        | 5.71  | 5000  | 0.2824          | {'precision': 0.8384728340675477, 'recall': 0.8963893249607535, 'f1': 0.8664643399089529, 'number': 637} | {'precision': 0.808641975308642, 'recall': 0.8488120950323974, 'f1': 0.8282402528977871, 'number': 463}  | {'precision': 0.9325581395348838, 'recall': 0.9570405727923628, 'f1': 0.944640753828033, 'number': 419}  | 0.8547            | 0.8986         | 0.8761     | 0.9363           |
| 0.0244        | 11.43 | 10000 | 0.4134          | {'precision': 0.8622754491017964, 'recall': 0.9042386185243328, 'f1': 0.8827586206896552, 'number': 637} | {'precision': 0.841991341991342, 'recall': 0.8401727861771058, 'f1': 0.8410810810810811, 'number': 463}  | {'precision': 0.920814479638009, 'recall': 0.9713603818615751, 'f1': 0.9454123112659697, 'number': 419}  | 0.8728            | 0.9032         | 0.8877     | 0.9370           |
| 0.0066        | 17.14 | 15000 | 0.5017          | {'precision': 0.8522895125553914, 'recall': 0.9058084772370487, 'f1': 0.878234398782344, 'number': 637}  | {'precision': 0.8361702127659575, 'recall': 0.8488120950323974, 'f1': 0.8424437299035369, 'number': 463} | {'precision': 0.9230769230769231, 'recall': 0.9737470167064439, 'f1': 0.9477351916376306, 'number': 419} | 0.8672            | 0.9072         | 0.8867     | 0.9365           |


### Framework versions

- Transformers 4.39.3
- Pytorch 1.11.0a0+17540c5
- Datasets 2.20.0
- Tokenizers 0.15.2