File size: 2,961 Bytes
1a490d1
 
 
 
990a884
 
 
 
 
1a490d1
 
 
 
 
 
 
 
 
 
990a884
 
 
 
 
 
 
1a490d1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
990a884
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1a490d1
 
 
 
 
 
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
68
69
70
71
72
73
74
75
76
77
78
79
80
81
---
license: mit
tags:
- generated_from_trainer
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: tmvar_0.0001_ES12
  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. -->

# tmvar_0.0001_ES12

This model is a fine-tuned version of [microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext](https://huggingface.co/microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0194
- Precision: 0.8877
- Recall: 0.8973
- F1: 0.8925
- Accuracy: 0.9968

## 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: 0.0001
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- training_steps: 1000

### Training results

| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
| 0.2263        | 1.47  | 25   | 0.0788          | 0.0       | 0.0    | 0.0    | 0.9843   |
| 0.0492        | 2.94  | 50   | 0.0355          | 0.2576    | 0.3676 | 0.3029 | 0.9863   |
| 0.0258        | 4.41  | 75   | 0.0224          | 0.6       | 0.6811 | 0.6380 | 0.9933   |
| 0.013         | 5.88  | 100  | 0.0141          | 0.8267    | 0.9027 | 0.8630 | 0.9969   |
| 0.0031        | 7.35  | 125  | 0.0162          | 0.8218    | 0.8973 | 0.8579 | 0.9971   |
| 0.0028        | 8.82  | 150  | 0.0187          | 0.8449    | 0.8541 | 0.8495 | 0.9961   |
| 0.0024        | 10.29 | 175  | 0.0154          | 0.8267    | 0.9027 | 0.8630 | 0.9965   |
| 0.0014        | 11.76 | 200  | 0.0159          | 0.8221    | 0.9243 | 0.8702 | 0.9966   |
| 0.0013        | 13.24 | 225  | 0.0179          | 0.8579    | 0.8811 | 0.8693 | 0.9971   |
| 0.0009        | 14.71 | 250  | 0.0165          | 0.8807    | 0.8378 | 0.8587 | 0.9964   |
| 0.0005        | 16.18 | 275  | 0.0184          | 0.8549    | 0.8919 | 0.8730 | 0.9966   |
| 0.0003        | 17.65 | 300  | 0.0188          | 0.8777    | 0.8919 | 0.8847 | 0.9967   |
| 0.0002        | 19.12 | 325  | 0.0195          | 0.8474    | 0.8703 | 0.8587 | 0.9964   |
| 0.0002        | 20.59 | 350  | 0.0192          | 0.8836    | 0.9027 | 0.8930 | 0.9969   |
| 0.0003        | 22.06 | 375  | 0.0191          | 0.8889    | 0.9081 | 0.8984 | 0.9969   |
| 0.0002        | 23.53 | 400  | 0.0194          | 0.8877    | 0.8973 | 0.8925 | 0.9968   |


### Framework versions

- Transformers 4.27.4
- Pytorch 2.0.0+cu118
- Datasets 2.11.0
- Tokenizers 0.13.2