Ezi commited on
Commit
9eb0223
•
1 Parent(s): af0f99b

Model Card

Browse files

Hi!:👋
This PR has a some additional information for the model card, based on the format we are using as part of our effort to standardise model cards at Hugging Face. Feel free to merge if you are ok with the changes! (cc

@Marissa



@Meg



@Nazneen

)

Files changed (1) hide show
  1. README.md +67 -9
README.md CHANGED
@@ -7,20 +7,55 @@ datasets:
7
 
8
  # DistilBERT base uncased finetuned SST-2
9
 
10
- This model is a fine-tune checkpoint of [DistilBERT-base-uncased](https://huggingface.co/distilbert-base-uncased), fine-tuned on SST-2.
 
 
 
 
 
 
 
 
11
  This model reaches an accuracy of 91.3 on the dev set (for comparison, Bert bert-base-uncased version reaches an accuracy of 92.7).
 
 
 
 
 
 
 
12
 
13
- For more details about DistilBERT, we encourage users to check out [this model card](https://huggingface.co/distilbert-base-uncased).
14
 
15
- # Fine-tuning hyper-parameters
 
 
 
 
16
 
17
- - learning_rate = 1e-5
18
- - batch_size = 32
19
- - warmup = 600
20
- - max_seq_length = 128
21
- - num_train_epochs = 3.0
 
 
 
 
22
 
23
- # Bias
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
  Based on a few experimentations, we observed that this model could produce biased predictions that target underrepresented populations.
26
 
@@ -29,3 +64,26 @@ For instance, for sentences like `This film was filmed in COUNTRY`, this binary
29
  <img src="https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english/resolve/main/map.jpeg" alt="Map of positive probabilities per country." width="500"/>
30
 
31
  We strongly advise users to thoroughly probe these aspects on their use-cases in order to evaluate the risks of this model. We recommend looking at the following bias evaluation datasets as a place to start: [WinoBias](https://huggingface.co/datasets/wino_bias), [WinoGender](https://huggingface.co/datasets/super_glue), [Stereoset](https://huggingface.co/datasets/stereoset).
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7
 
8
  # DistilBERT base uncased finetuned SST-2
9
 
10
+ ## Table of Contents
11
+ - [Model Details](#model-details)
12
+ - [How to Get Started With the Model](#how-to-get-started-with-the-model)
13
+ - [Uses](#uses)
14
+ - [Risks, Limitations and Biases](#risks-limitations-and-biases)
15
+ - [Training](#training)
16
+
17
+ ## Model Details
18
+ **Model Description:** This model is a fine-tune checkpoint of [DistilBERT-base-uncased](https://huggingface.co/distilbert-base-uncased), fine-tuned on SST-2.
19
  This model reaches an accuracy of 91.3 on the dev set (for comparison, Bert bert-base-uncased version reaches an accuracy of 92.7).
20
+ - **Developed by:** Hugging Face
21
+ - **Model Type:** Text Classification
22
+ - **Language(s):** English
23
+ - **License:** Apache-2.0
24
+ - **Parent Model:** For more details about DistilBERT, we encourage users to check out [this model card](https://huggingface.co/distilbert-base-uncased).
25
+ - **Resources for more information:**
26
+ - [Model Documentation](https://huggingface.co/docs/transformers/main/en/model_doc/distilbert#transformers.DistilBertForSequenceClassification)
27
 
28
+ ## How to Get Started With the Model
29
 
30
+ Example of single-label classification:
31
+ ​​
32
+ ```python
33
+ import torch
34
+ from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
35
 
36
+ tokenizer = DistilBertTokenizer.from_pretrained("distilbert-base-uncased")
37
+ model = DistilBertForSequenceClassification.from_pretrained("distilbert-base-uncased")
38
+
39
+ inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
40
+ with torch.no_grad():
41
+ logits = model(**inputs).logits
42
+
43
+ predicted_class_id = logits.argmax().item()
44
+ model.config.id2label[predicted_class_id]
45
 
46
+ ```
47
+
48
+ ## Uses
49
+
50
+ #### Direct Use
51
+
52
+ This model can be used for topic classification. You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the model hub to look for fine-tuned versions on a task that interests you.
53
+
54
+ #### Misuse and Out-of-scope Use
55
+ The model should not be used to intentionally create hostile or alienating environments for people. In addition, the model was not trained to be factual or true representations of people or events, and therefore using the model to generate such content is out-of-scope for the abilities of this model.
56
+
57
+
58
+ ## Risks, Limitations and Biases
59
 
60
  Based on a few experimentations, we observed that this model could produce biased predictions that target underrepresented populations.
61
 
 
64
  <img src="https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english/resolve/main/map.jpeg" alt="Map of positive probabilities per country." width="500"/>
65
 
66
  We strongly advise users to thoroughly probe these aspects on their use-cases in order to evaluate the risks of this model. We recommend looking at the following bias evaluation datasets as a place to start: [WinoBias](https://huggingface.co/datasets/wino_bias), [WinoGender](https://huggingface.co/datasets/super_glue), [Stereoset](https://huggingface.co/datasets/stereoset).
67
+
68
+
69
+
70
+ # Training
71
+
72
+
73
+ #### Training Data
74
+
75
+
76
+ The authors use the following Stanford Sentiment Treebank([sst2](https://huggingface.co/datasets/sst2)) corpora for the model.
77
+
78
+ #### Training Procedure
79
+
80
+ ###### Fine-tuning hyper-parameters
81
+
82
+
83
+ - learning_rate = 1e-5
84
+ - batch_size = 32
85
+ - warmup = 600
86
+ - max_seq_length = 128
87
+ - num_train_epochs = 3.0
88
+
89
+