lrodrigues commited on
Commit
c39c03b
1 Parent(s): 7b1992c
Files changed (8) hide show
  1. README.md +52 -0
  2. config.json +48 -0
  3. flax_model.msgpack +3 -0
  4. merges.txt +0 -0
  5. pytorch_model.bin +3 -0
  6. tokenizer.json +0 -0
  7. tokenizer_config.json +3 -0
  8. vocab.json +0 -0
README.md ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets:
3
+ - multi_nli
4
+ language: en
5
+ license: mit
6
+ pipeline_tag: zero-shot-classification
7
+ tags:
8
+ - bart
9
+ - zero-shot-classification
10
+ ---
11
+ # Bart large model for NLI-based Zero Shot Text Classification
12
+
13
+ This model uses [bart-large](https://huggingface.co/facebook/bart-large).
14
+
15
+ ## Training Data
16
+ This model was trained on the [MultiNLI (MNLI)](https://huggingface.co/datasets/multi_nli) dataset in the manner originally described in [Yin et al. 2019](https://arxiv.org/abs/1909.00161).
17
+
18
+ It can be used to predict whether a topic label can be assigned to a given sequence, whether or not the label has been seen before.
19
+
20
+ ## Usage and Performance
21
+ The trained model can be used like this:
22
+ ```python
23
+ from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
24
+
25
+ # Load model & tokenizer
26
+ bart_model = AutoModelForSequenceClassification.from_pretrained('navteca/bart-large-mnli')
27
+ bart_tokenizer = AutoTokenizer.from_pretrained('navteca/bart-large-mnli')
28
+
29
+ # Get predictions
30
+ nlp = pipeline('zero-shot-classification', model=bart_model, tokenizer=bart_tokenizer)
31
+
32
+ sequence = 'One day I will see the world.'
33
+ candidate_labels = ['cooking', 'dancing', 'travel']
34
+
35
+ result = nlp(sequence, candidate_labels, multi_label=True)
36
+
37
+ print(result)
38
+
39
+ #{
40
+ # "sequence": "One day I will see the world.",
41
+ # "labels": [
42
+ # "travel",
43
+ # "dancing",
44
+ # "cooking"
45
+ # ],
46
+ # "scores": [
47
+ # 0.9941897988319397,
48
+ # 0.0060537424869835,
49
+ # 0.0020010927692056
50
+ # ]
51
+ #}
52
+ ```
config.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_num_labels": 3,
3
+ "activation_dropout": 0,
4
+ "activation_function": "gelu",
5
+ "add_final_layer_norm": false,
6
+ "architectures": [
7
+ "BartForSequenceClassification"
8
+ ],
9
+ "attention_dropout": 0,
10
+ "bos_token_id": 0,
11
+ "classif_dropout": 0,
12
+ "classifier_dropout": 0,
13
+ "d_model": 1024,
14
+ "decoder_attention_heads": 16,
15
+ "decoder_ffn_dim": 4096,
16
+ "decoder_layerdrop": 0,
17
+ "decoder_layers": 12,
18
+ "decoder_start_token_id": 2,
19
+ "dropout": 0.1,
20
+ "encoder_attention_heads": 16,
21
+ "encoder_ffn_dim": 4096,
22
+ "encoder_layerdrop": 0,
23
+ "encoder_layers": 12,
24
+ "eos_token_id": 2,
25
+ "forced_eos_token_id": 2,
26
+ "gradient_checkpointing": false,
27
+ "id2label": {
28
+ "0": "contradiction",
29
+ "1": "neutral",
30
+ "2": "entailment"
31
+ },
32
+ "init_std": 0.02,
33
+ "is_encoder_decoder": true,
34
+ "label2id": {
35
+ "contradiction": 0,
36
+ "neutral": 1,
37
+ "entailment": 2
38
+ },
39
+ "max_position_embeddings": 1024,
40
+ "model_type": "bart",
41
+ "normalize_before": false,
42
+ "num_hidden_layers": 12,
43
+ "output_past": false,
44
+ "pad_token_id": 1,
45
+ "scale_embedding": false,
46
+ "use_cache": true,
47
+ "vocab_size": 50265
48
+ }
flax_model.msgpack ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e07d1ae73ae1c1267fd174a3b21c73b0d77bad288f8ed17fb685f79c419a897
3
+ size 1629394629
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:ce253627f98f9db22af6a86efee6e905f001f7d8dc02dd14a8b4b4710c302b17
3
+ size 1629486723
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "model_max_length": 1024
3
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff