migueladarlo
commited on
Commit
•
77f6485
1
Parent(s):
466bbd5
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,74 @@
|
|
1 |
---
|
2 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
license: mit # Example: apache-2.0 or any license from https://huggingface.co/docs/hub/model-repos#list-of-license-identifiers
|
5 |
+
tags:
|
6 |
+
- text # Example: audio
|
7 |
+
- Twitter
|
8 |
+
datasets:
|
9 |
+
- CLPsych 2015 # Example: common_voice. Use dataset id from https://hf.co/datasets
|
10 |
+
metrics:
|
11 |
+
- accuracy, f1, precision, recall, AUC # Example: wer. Use metric id from https://hf.co/metrics
|
12 |
+
|
13 |
+
model-index:
|
14 |
+
- name: distilbert-depression-mixed
|
15 |
+
results: []
|
16 |
---
|
17 |
+
|
18 |
+
# distilbert-depression-mixed
|
19 |
+
|
20 |
+
This model is a fine-tuned version of [base-uncased](https://huggingface.co/distilbert-base-uncased) trained on CLPsych 2015 and a scraped dataset, and evaluated on a scraped dataset from Twitter.
|
21 |
+
It achieves the following results on the evaluation set:
|
22 |
+
- Evaluation Loss: 0.71
|
23 |
+
- Accuracy: 0.63
|
24 |
+
- F1: 0.59
|
25 |
+
- Precision: 0.66
|
26 |
+
- Recall: 0.53
|
27 |
+
- AUC: 0.63
|
28 |
+
|
29 |
+
|
30 |
+
## Intended uses & limitations
|
31 |
+
|
32 |
+
Feed a corpus of tweets to the model to generate label if input is indicative of depression or not. Label 1 is depression, Label 0 is not.
|
33 |
+
|
34 |
+
Limitation: All token sequences longer than 512 are automatically truncated. Also, training and test data may be contaminated with mislabeled users.
|
35 |
+
|
36 |
+
### How to use
|
37 |
+
|
38 |
+
You can use this model directly with a pipeline for sentiment analysis:
|
39 |
+
|
40 |
+
```python
|
41 |
+
>>> from transformers import DistilBertTokenizerFast, AutoTokenizer
|
42 |
+
>>> tokenizer = AutoTokenizer.from_pretrained('distilbert-base-uncased')
|
43 |
+
>>> from transformers import DistilBertForSequenceClassification
|
44 |
+
>>> model = DistilBertForSequenceClassification.from_pretrained(r"distilbert-depression-mixed")
|
45 |
+
>>> from transformers import pipeline
|
46 |
+
>>> classifier = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
|
47 |
+
>>> tokenizer_kwargs = {'padding':True,'truncation':True,'max_length':512}
|
48 |
+
>>> result=classifier('pain peko',**tokenizer_kwargs) #For truncation to apply in the pipeline
|
49 |
+
|
50 |
+
|
51 |
+
[{'label': 'LABEL_1', 'score': 0.5048992037773132}]
|
52 |
+
```
|
53 |
+
|
54 |
+
Otherwise, download the files and specify within the pipeline the path to the folder that contains the config.json, pytorch_model.bin, and training_args.bin
|
55 |
+
|
56 |
+
## Training hyperparameters
|
57 |
+
|
58 |
+
The following hyperparameters were used during training:
|
59 |
+
- learning_rate: 4.19e-05
|
60 |
+
- train_batch_size: 16
|
61 |
+
- eval_batch_size: 16
|
62 |
+
- weight_decay: 0.06
|
63 |
+
- num_epochs: 5.0
|
64 |
+
|
65 |
+
## Training results
|
66 |
+
|
67 |
+
|
68 |
+
| Epoch | Training Loss | Validation Loss | Accuracy | F1 | Precision | Recall | AUC |
|
69 |
+
|:-----:|:-------------:|:---------------:|:--------:|:--------:|:---------:|:--------:|:--------:|
|
70 |
+
| 1.0 | 0.68 | 0.66 | 0.61 | 0.54 | 0.60 | 0.50 | 0.60 |
|
71 |
+
| 2.0 | 0.65 | 0.65 | 0.63 | 0.49 | 0.70 | 0.37 | 0.62 |
|
72 |
+
| 3.0 | 0.53 | 0.63 | 0.66 | 0.58 | 0.69 | 0.50 | 0.65 |
|
73 |
+
| 4.0 | 0.39 | 0.66 | 0.67 | 0.61 | 0.69 | 0.54 | 0.67 |
|
74 |
+
| 5.0 | 0.27 | 0.72 | 0.65 | 0.61 | 0.63 | 0.60 | 0.64 |
|