cardiffnlp commited on
Commit
c3f0011
•
1 Parent(s): c6eb303

Adding tweeteval classifier

Browse files
.ipynb_checkpoints/README-checkpoint.md ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Twitter-roBERTa-base
2
+
3
+ This is a roBERTa-base model trained on ~58M tweets and finetuned for the Sentiment Analysis task at Semeval 2018.
4
+ For full description: [_TweetEval_ benchmark (Findings of EMNLP 2020)](https://arxiv.org/pdf/2010.12421.pdf).
5
+ To evaluate this and other models on Twitter-specific data, please refer to the [Tweeteval official repository](https://github.com/cardiffnlp/tweeteval).
6
+
7
+ ## Example of classification
8
+
9
+ ```python
10
+ from transformers import AutoModelForSequenceClassification
11
+ from transformers import TFAutoModelForSequenceClassification
12
+ from transformers import AutoTokenizer
13
+ import numpy as np
14
+ from scipy.special import softmax
15
+ import csv
16
+ import urllib.request
17
+
18
+ # Tasks:
19
+ # emoji, emotion, hate, irony, offensive, sentiment
20
+ # stance/abortion, stance/atheism, stance/climate, stance/feminist, stance/hillary
21
+
22
+ task='sentiment'
23
+ MODEL = f"cardiffnlp/twitter-roberta-base-{task}"
24
+
25
+ tokenizer = AutoTokenizer.from_pretrained(MODEL)
26
+
27
+ # download label mapping
28
+ labels=[]
29
+ mapping_link = f"https://raw.githubusercontent.com/cardiffnlp/tweeteval/main/datasets/{task}/mapping.txt"
30
+ with urllib.request.urlopen(mapping_link) as f:
31
+ html = f.read().decode('utf-8').split("\n")
32
+ spamreader = csv.reader(html[:-1], delimiter='\t')
33
+ labels = [row[1] for row in spamreader]
34
+
35
+ # PT
36
+ model = AutoModelForSequenceClassification.from_pretrained(MODEL)
37
+ model.save_pretrained(MODEL)
38
+
39
+ text = "Good night 😊"
40
+ encoded_input = tokenizer(text, return_tensors='pt')
41
+ output = model(**encoded_input)
42
+ scores = output[0][0].detach().numpy()
43
+ scores = softmax(scores)
44
+
45
+ # # TF
46
+ # model = TFAutoModelForSequenceClassification.from_pretrained(MODEL)
47
+ # model.save_pretrained(MODEL)
48
+
49
+ # text = "Good night 😊"
50
+ # encoded_input = tokenizer(text, return_tensors='tf')
51
+ # output = model(encoded_input)
52
+ # scores = output[0][0].numpy()
53
+ # scores = softmax(scores)
54
+
55
+ ranking = np.argsort(scores)
56
+ ranking = ranking[::-1]
57
+ for i in range(scores.shape[0]):
58
+ l = labels[ranking[i]]
59
+ s = scores[ranking[i]]
60
+ print(f"{i+1}) {l} {np.round(float(s), 4)}")
61
+
62
+ ```
63
+
64
+ Output:
65
+
66
+ ```
67
+ 1) positive 0.8466
68
+ 2) neutral 0.1458
69
+ 3) negative 0.0076
70
+ ```
README.md ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Twitter-roBERTa-base
2
+
3
+ This is a roBERTa-base model trained on ~58M tweets and finetuned for the Sentiment Analysis task at Semeval 2018.
4
+ For full description: [_TweetEval_ benchmark (Findings of EMNLP 2020)](https://arxiv.org/pdf/2010.12421.pdf).
5
+ To evaluate this and other models on Twitter-specific data, please refer to the [Tweeteval official repository](https://github.com/cardiffnlp/tweeteval).
6
+
7
+ ## Example of classification
8
+
9
+ ```python
10
+ from transformers import AutoModelForSequenceClassification
11
+ from transformers import TFAutoModelForSequenceClassification
12
+ from transformers import AutoTokenizer
13
+ import numpy as np
14
+ from scipy.special import softmax
15
+ import csv
16
+ import urllib.request
17
+
18
+ # Tasks:
19
+ # emoji, emotion, hate, irony, offensive, sentiment
20
+ # stance/abortion, stance/atheism, stance/climate, stance/feminist, stance/hillary
21
+
22
+ task='sentiment'
23
+ MODEL = f"cardiffnlp/twitter-roberta-base-{task}"
24
+
25
+ tokenizer = AutoTokenizer.from_pretrained(MODEL)
26
+
27
+ # download label mapping
28
+ labels=[]
29
+ mapping_link = f"https://raw.githubusercontent.com/cardiffnlp/tweeteval/main/datasets/{task}/mapping.txt"
30
+ with urllib.request.urlopen(mapping_link) as f:
31
+ html = f.read().decode('utf-8').split("\n")
32
+ spamreader = csv.reader(html[:-1], delimiter='\t')
33
+ labels = [row[1] for row in spamreader]
34
+
35
+ # PT
36
+ model = AutoModelForSequenceClassification.from_pretrained(MODEL)
37
+ model.save_pretrained(MODEL)
38
+
39
+ text = "Good night 😊"
40
+ encoded_input = tokenizer(text, return_tensors='pt')
41
+ output = model(**encoded_input)
42
+ scores = output[0][0].detach().numpy()
43
+ scores = softmax(scores)
44
+
45
+ # # TF
46
+ # model = TFAutoModelForSequenceClassification.from_pretrained(MODEL)
47
+ # model.save_pretrained(MODEL)
48
+
49
+ # text = "Good night 😊"
50
+ # encoded_input = tokenizer(text, return_tensors='tf')
51
+ # output = model(encoded_input)
52
+ # scores = output[0][0].numpy()
53
+ # scores = softmax(scores)
54
+
55
+ ranking = np.argsort(scores)
56
+ ranking = ranking[::-1]
57
+ for i in range(scores.shape[0]):
58
+ l = labels[ranking[i]]
59
+ s = scores[ranking[i]]
60
+ print(f"{i+1}) {l} {np.round(float(s), 4)}")
61
+
62
+ ```
63
+
64
+ Output:
65
+
66
+ ```
67
+ 1) positive 0.8466
68
+ 2) neutral 0.1458
69
+ 3) negative 0.0076
70
+ ```
config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "tweeteval_new/roberta-base-rt-sentiment/",
3
+ "architectures": [
4
+ "RobertaForSequenceClassification"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": 0,
8
+ "eos_token_id": 2,
9
+ "gradient_checkpointing": false,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout_prob": 0.1,
12
+ "hidden_size": 768,
13
+ "id2label": {
14
+ "0": "LABEL_0",
15
+ "1": "LABEL_1",
16
+ "2": "LABEL_2"
17
+ },
18
+ "initializer_range": 0.02,
19
+ "intermediate_size": 3072,
20
+ "label2id": {
21
+ "LABEL_0": 0,
22
+ "LABEL_1": 1,
23
+ "LABEL_2": 2
24
+ },
25
+ "layer_norm_eps": 1e-05,
26
+ "max_position_embeddings": 514,
27
+ "model_type": "roberta",
28
+ "num_attention_heads": 12,
29
+ "num_hidden_layers": 12,
30
+ "pad_token_id": 1,
31
+ "type_vocab_size": 1,
32
+ "vocab_size": 50265
33
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c37a3484c55954cd75b336a85f1e0c023ae874f3a73b05d2418dd04828e293b1
3
+ size 498679497
special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "<s>", "mask_token": "<mask>"}
tf_model.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60edb4641e2b28ee3ecfb272f96cb9a6b0a662f4072eaf294dd8a1fd8b8484f3
3
+ size 501229896
vocab.json ADDED
The diff for this file is too large to render. See raw diff