Hugging Face's logo

sangkm
/
augmented-go-emotions-plus-other-datasets-fine-tuned-distilroberta

Text Classification
Transformers
TensorBoard
Safetensors
English
roberta
Generated from Trainer
sentiment_analysis
Inference Endpoints
Model card Files Metrics Community
pm commited on
Commit
15a607c
•
1 Parent(s): 036fa6f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +110 -70
README.md CHANGED
@@ -1,70 +1,110 @@
1
- ---
2
- library_name: transformers
3
- license: apache-2.0
4
- base_model: distilbert/distilroberta-base
5
- tags:
6
- - generated_from_trainer
7
- model-index:
8
- - name: augmented-go-emotions-plus-other-datasets-fine-tuned-distilroberta
9
- results: []
10
- ---
11
-
12
- <!-- This model card has been generated automatically according to the information the Trainer had access to. You
13
- should probably proofread and complete it, then remove this comment. -->
14
-
15
- # augmented-go-emotions-plus-other-datasets-fine-tuned-distilroberta
16
-
17
- This model is a fine-tuned version of [distilbert/distilroberta-base](https://huggingface.co/distilbert/distilroberta-base) on the None dataset.
18
- It achieves the following results on the evaluation set:
19
- - Loss: 0.0731
20
- - Micro Precision: 0.7189
21
- - Micro Recall: 0.5774
22
- - Micro F1: 0.6404
23
- - Macro Precision: 0.6049
24
- - Macro Recall: 0.4433
25
- - Macro F1: 0.4898
26
- - Weighted Precision: 0.7004
27
- - Weighted Recall: 0.5774
28
- - Weighted F1: 0.6243
29
- - Hamming Loss: 0.0276
30
-
31
- ## Model description
32
-
33
- More information needed
34
-
35
- ## Intended uses & limitations
36
-
37
- More information needed
38
-
39
- ## Training and evaluation data
40
-
41
- More information needed
42
-
43
- ## Training procedure
44
-
45
- ### Training hyperparameters
46
-
47
- The following hyperparameters were used during training:
48
- - learning_rate: 5e-05
49
- - train_batch_size: 8
50
- - eval_batch_size: 8
51
- - seed: 42
52
- - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
53
- - lr_scheduler_type: linear
54
- - num_epochs: 3.0
55
-
56
- ### Training results
57
-
58
- | Training Loss | Epoch | Step | Validation Loss | Micro Precision | Micro Recall | Micro F1 | Macro Precision | Macro Recall | Macro F1 | Weighted Precision | Weighted Recall | Weighted F1 | Hamming Loss |
59
- |:-------------:|:-----:|:-----:|:---------------:|:---------------:|:------------:|:--------:|:---------------:|:------------:|:--------:|:------------------:|:---------------:|:-----------:|:------------:|
60
- | No log | 1.0 | 11118 | 0.0765 | 0.7647 | 0.5046 | 0.6080 | 0.6047 | 0.3580 | 0.4127 | 0.7321 | 0.5046 | 0.5764 | 0.0277 |
61
- | No log | 2.0 | 22236 | 0.0733 | 0.7309 | 0.5344 | 0.6174 | 0.5791 | 0.4162 | 0.4611 | 0.7105 | 0.5344 | 0.5923 | 0.0282 |
62
- | No log | 3.0 | 33354 | 0.0731 | 0.7189 | 0.5774 | 0.6404 | 0.6049 | 0.4433 | 0.4898 | 0.7004 | 0.5774 | 0.6243 | 0.0276 |
63
-
64
-
65
- ### Framework versions
66
-
67
- - Transformers 4.47.0
68
- - Pytorch 2.3.1+cu121
69
- - Datasets 2.20.0
70
- - Tokenizers 0.21.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: distilbert/distilroberta-base
5
+ tags:
6
+ - generated_from_trainer
7
+ - sentiment_analysis
8
+ model-index:
9
+ - name: augmented-go-emotions-plus-other-datasets-fine-tuned-distilroberta
10
+ results: []
11
+ datasets:
12
+ - google-research-datasets/go_emotions
13
+ language:
14
+ - en
15
+ metrics:
16
+ - f1
17
+ - precision
18
+ - recall
19
+ ---
20
+
21
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
22
+ should probably proofread and complete it, then remove this comment. -->
23
+
24
+ # augmented-go-emotions-plus-other-datasets-fine-tuned-distilroberta
25
+
26
+ This model is a fine-tuned version of [distilbert/distilroberta-base](https://huggingface.co/distilbert/distilroberta-base) on the these datasets:
27
+ - [GoEmotions](https://github.com/google-research/google-research/tree/master/goemotions)
28
+ - [sem_eval_2018_task_1 (English)](https://huggingface.co/datasets/SemEvalWorkshop/sem_eval_2018_task_1)
29
+ - [Emotion Detection from Text - Pashupati Gupta](https://www.kaggle.com/datasets/pashupatigupta/emotion-detection-from-text/data)
30
+ - [Emotions dataset for NLP - praveengovi](https://www.kaggle.com/datasets/praveengovi/emotions-dataset-for-nlp/data)
31
+ It has also been data augmented using TextAttack.
32
+
33
+ It achieves the following results on the evaluation set:
34
+ - Loss: 0.0731
35
+ - Micro Precision: 0.7189
36
+ - Micro Recall: 0.5774
37
+ - Micro F1: 0.6404
38
+ - Macro Precision: 0.6049
39
+ - Macro Recall: 0.4433
40
+ - Macro F1: 0.4898
41
+ - Weighted Precision: 0.7004
42
+ - Weighted Recall: 0.5774
43
+ - Weighted F1: 0.6243
44
+ - Hamming Loss: 0.0276
45
+
46
+
47
+ ### Training hyperparameters
48
+
49
+ The following hyperparameters were used during training:
50
+ - learning_rate: 5e-05
51
+ - train_batch_size: 8
52
+ - eval_batch_size: 8
53
+ - seed: 42
54
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
55
+ - lr_scheduler_type: linear
56
+ - num_epochs: 3.0
57
+
58
+ ### Training results
59
+
60
+ | Training Loss | Epoch | Step | Validation Loss | Micro Precision | Micro Recall | Micro F1 | Macro Precision | Macro Recall | Macro F1 | Weighted Precision | Weighted Recall | Weighted F1 | Hamming Loss |
61
+ |:-------------:|:-----:|:-----:|:---------------:|:---------------:|:------------:|:--------:|:---------------:|:------------:|:--------:|:------------------:|:---------------:|:-----------:|:------------:|
62
+ | No log | 1.0 | 11118 | 0.0765 | 0.7647 | 0.5046 | 0.6080 | 0.6047 | 0.3580 | 0.4127 | 0.7321 | 0.5046 | 0.5764 | 0.0277 |
63
+ | No log | 2.0 | 22236 | 0.0733 | 0.7309 | 0.5344 | 0.6174 | 0.5791 | 0.4162 | 0.4611 | 0.7105 | 0.5344 | 0.5923 | 0.0282 |
64
+ | No log | 3.0 | 33354 | 0.0731 | 0.7189 | 0.5774 | 0.6404 | 0.6049 | 0.4433 | 0.4898 | 0.7004 | 0.5774 | 0.6243 | 0.0276 |
65
+
66
+
67
+ ### Test results
68
+ Threshold = 0.5
69
+ | Label | Precision | Recall | F1-Score | Support |
70
+ |------------------|-----------|--------|----------|---------|
71
+ | admiration | 0.65 | 0.70 | 0.67 | 504 |
72
+ | amusement | 0.72 | 0.88 | 0.79 | 264 |
73
+ | anger | 0.79 | 0.69 | 0.73 | 1585 |
74
+ | annoyance | 0.45 | 0.12 | 0.19 | 320 |
75
+ | approval | 0.63 | 0.27 | 0.38 | 351 |
76
+ | caring | 0.44 | 0.36 | 0.40 | 135 |
77
+ | confusion | 0.44 | 0.39 | 0.41 | 153 |
78
+ | curiosity | 0.52 | 0.36 | 0.43 | 284 |
79
+ | desire | 0.50 | 0.37 | 0.43 | 83 |
80
+ | disappointment | 0.35 | 0.19 | 0.25 | 151 |
81
+ | disapproval | 0.49 | 0.31 | 0.38 | 267 |
82
+ | disgust | 0.72 | 0.62 | 0.66 | 1222 |
83
+ | embarrassment | 0.68 | 0.35 | 0.46 | 37 |
84
+ | excitement | 0.46 | 0.43 | 0.44 | 103 |
85
+ | fear | 0.82 | 0.73 | 0.77 | 787 |
86
+ | gratitude | 0.93 | 0.89 | 0.91 | 352 |
87
+ | grief | 0.00 | 0.00 | 0.00 | 6 |
88
+ | joy | 0.85 | 0.78 | 0.81 | 2298 |
89
+ | love | 0.70 | 0.60 | 0.65 | 1305 |
90
+ | nervousness | 0.44 | 0.17 | 0.25 | 23 |
91
+ | optimism | 0.70 | 0.56 | 0.62 | 1329 |
92
+ | pride | 0.00 | 0.00 | 0.00 | 16 |
93
+ | realization | 0.36 | 0.17 | 0.23 | 145 |
94
+ | relief | 0.28 | 0.22 | 0.24 | 160 |
95
+ | remorse | 0.59 | 0.80 | 0.68 | 56 |
96
+ | sadness | 0.78 | 0.66 | 0.71 | 2212 |
97
+ | surprise | 0.63 | 0.29 | 0.40 | 572 |
98
+ | neutral | 0.70 | 0.52 | 0.60 | 2668 |
99
+ | **Micro Avg** | 0.73 | 0.59 | 0.65 | 17388 |
100
+ | **Macro Avg** | 0.56 | 0.44 | 0.48 | 17388 |
101
+ | **Weighted Avg** | 0.72 | 0.59 | 0.64 | 17388 |
102
+ | **Samples Avg** | 0.63 | 0.60 | 0.60 | 17388 |
103
+
104
+
105
+ ### Framework versions
106
+
107
+ - Transformers 4.47.0
108
+ - Pytorch 2.3.1+cu121
109
+ - Datasets 2.20.0
110
+ - Tokenizers 0.21.0