SetFit with sentence-transformers/paraphrase-mpnet-base-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-mpnet-base-v2 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification.

The model has been trained using an efficient few-shot learning technique that involves:

Fine-tuning a Sentence Transformer with contrastive learning.
Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

Model Type: SetFit
Sentence Transformer body: sentence-transformers/paraphrase-mpnet-base-v2
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 tokens
Number of Classes: 119 classes

Model Sources

Repository: SetFit on GitHub
Paper: Efficient Few-Shot Learning Without Prompts
Blogpost: SetFit: Efficient Few-Shot Learning Without Prompts

Model Labels

Label	Examples
79	'peony middle notes' 'lemon middle notes' 'coconut middle notes'
86	'no print/no pattern' 'two tone' 'diagonal stripe'
37	'eel skin leather' 'metal' 'raffia'
82	'collarless' 'peaked lapel' 'front keyhole'
95	'standard toe' 'wide toe' 'extra wide toe'
83	'indoor' 'hike' 'beach'
107	'surplice' 'messenger bag' 'camera bag'
19	'mary jane' 'zip around wallet' 'tongue buckle'
102	'slits at knee' 'slits above hips' 'front slit at hem'
35	'tie' 'gem embellishment' 'caged'
18	'rolo chain' 'cord bracelet' 'figaro'
65	'wheat protein' 'rosemary ingredient' 'pea protein'
68	'bath towel' 'art print' 'reusable bottle'
40	'polyfill' 'silk fill' 'feather fill'
50	'palm grip' 'carpenter hook' 'storm flap'
113	'wide waistband' 'elastic inset' 'belt loops'
75	'glass' 'acrylic' 'opal'
11	'foam cups' 'wire' 'molded cups'
38	'dual layer fabric' '2 way stretch' '4 way stretch'
63	'light support' 'medium supprt' 'high support'
44	'face' 'hand' 'neck/dècolletage'
115	'soy wax' 'paraffin wax'
42	'regular' 'tailored' 'fitted'
97	'king' 'euro' 'standard'
70	'wrist length' 'above thigh' 'below bust'
34	'feminine' 'religious' 'boho'
10	'slim' 'regular'
15	'6-10 oz' '11-20 oz'
77	'rose gold metal' 'gold plated' 'alloy'
43	'contrast inner lining' 'simple seaming' 'princess seams'
7	'neroli base notes' 'amber base notes' 'musk base notes'
17	'spot clean' 'dry clean' 'microwave safe'
8	'nourishing' 'firming' 'soothing/healing'
103	'lugged soles' 'non marking soles'
26	'wall control' 'switch control'
99	'fitted sleeves' 'fitted sleeve' 'structured sleeves'
33	'rim' 'feet' '5 panel construction'
64	'mineral oil free' 'propylene glycol free' 'paraffin free'
96	'double strap' 'spaghetti straps' 'thin straps'
1	'shoulder back' 'full coverage' 'low back'
62	'rustic' 'coastal' 'scandinavian'
39	'metallic' 'swiss dot' 'base layer'
60	'halloween' 'christmas holiday'
92	'seamless' 'mid rise waist seam' 'flat seam'
114	'ultra high rise' 'mid rise' 'high waisted'
105	'top handle' 'detachable straps' 'chain strap'
90	'floral' 'psychedelic print' 'paisley'
91	'night' 'day'
45	'serum formulation' 'cream/creme' 'solid'
59	'strong hold' 'flexible hold'
46	'leather' 'fresh aquatic' 'green aromatic'
21	'matte' 'metallic' 'olive'
69	'cinnamon key notes' 'violet key notes' 'pepper key notes'
101	'dropped shoulder' 'puff shoulder' 'flutter sleeve'
61	'summer' 'everyday' 'indoor'
104	'wedding guest' 'bridal' 'halloween'
32	'indigo wash' 'acid wash' 'stonewash'
51	'still life graphic' 'sports graphic' 'star wars'
48	'beige' 'black' 'rose gold frame'
87	'medium pile' 'low pile'
22	'bright' 'pastel' 'light'
41	'matte finish' 'shiny finish'
93	'no buckle' 'geometric shape' 'straight silhouette'
71	'polarized' 'color tinted' 'mirrored'
2	'split back' 'racer back' 'open back'
89	'round stitch pocket' 'seam pocket' 'kangaroo pocket'
20	'removable hoodie' 'packable hood collar' 'hooded'
52	'thick' 'medium thick'
55	'amber head notes' 'lime head notes' 'musk head notes'
58	'back curved hem' 'twist hem' 'ribbed hem'
118	'light wood' 'medium wood'
25	'gifts for him' 'apres ski' 'cozy'
109	'closed toe' 'square toe' 'round toe'
30	'extended cuffs' 'storm cuffs' 'elastic cuff'
24	'ingrown hairs' 'frizz' 'redness'
9	'high cut' 'string bikini'
94	'opaque' 'sheer'
16	'2 card slot' 'card slots'
78	'gothcore' 'vanilla girl' 'dyed out'
4	'layered' 'bangle' 'cuff'
23	'parfum' 'eau de toilette'
111	'delicate' 'statement'
12	'flat brim' 'curved brim' 'fold over brim'
98	'dry' 'acne prone' 'mature'
57	'stacked heel' 'kitten heel' 'cone heel'
67	'id slot' 'interior pocket' 'interior zipper pocket'
31	'light wash' 'medium wash' 'colored'
85	'detailed stitching pant' 'simple seaming'
116	'knotted' 'percale' 'waffle weave'
88	'shag' 'cut pile'
74	'study hall' 'y2k' 'enchanted'
72	'fur' 'fleece' 'mesh'
108	'animal' 'love'
73	'unlined' 'fully lined' 'partially lined'
13	'wide brim' 'medium brim'
76	'bpa free material' 'scratch resistant material'
54	'straight handle' 'curved handle'
100	'rolled up sleeves' '3/4 sleeve' 'bracelet length'
84	'manual open' 'auto open'
14	'wide' 'medium'
27	'superhero' 'disney'
49	'half rim' 'full rim'
29	'tall crown' 'short crown'
106	'low stretch' 'non stretch'
112	'mid vamp' 'high vamp'
66	'large interior' 'medium interior' 'small interior'
53	'all hair types' 'damaged/dry hair'
117	'light weight' 'mid weight'
81	'low cut' 'mid chest neckline' 'open front'
5	'thin band' 'soft band elastic' 'elastic band'
28	'flat top crown' 'round crown' 'no crown'
56	'ultra high heel' 'mid heel' 'high heel'
110	'relaxed' 'tailored'
47	'uplifting' 'bold'
3	'changing pad' 'bottle pocket'
0	'squeeze dispenser' 'dropper'
80	'wall mount' 'ceiling mount'
6	'medium' 'wide'
36	'exterior pocket' 'exterior snap pocket'

Evaluation

Metrics

Label	Accuracy
all	0.5493

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("kaustubhgap/kaustubh_setfit_1iteration")
# Run inference
preds = model("tube")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	1.7047	6

Label	Training Sample Count
0	2
1	5
2	12
3	2
4	6
5	3
6	2
7	12
8	16
9	2
10	2
11	11
12	4
13	2
14	2
15	2
16	2
17	6
18	9
19	63
20	8
21	31
22	6
23	2
24	13
25	5
26	2
27	2
28	3
29	2
30	13
31	3
32	7
33	22
34	12
35	102
36	2
37	119
38	34
39	32
40	6
41	2
42	13
43	17
44	5
45	10
46	6
47	2
48	10
49	2
50	91
51	13
52	2
53	2
54	2
55	12
56	4
57	7
58	17
59	2
60	2
61	7
62	9
63	3
64	14
65	53
66	3
67	6
68	41
69	41
70	33
71	5
72	5
73	4
74	7
75	49
76	2
77	23
78	11
79	12
80	2
81	5
82	33
83	33
84	2
85	2
86	17
87	2
88	2
89	10
90	29
91	2
92	8
93	21
94	2
95	3
96	5
97	10
98	5
99	6
100	6
101	12
102	13
103	2
104	10
105	28
106	2
107	321
108	2
109	10
110	2
111	2
112	2
113	15
114	4
115	2
116	5
117	2
118	2

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 10
body_learning_rate: (2e-05, 1e-05)
head_learning_rate: 0.01
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: False
use_amp: False
warmup_proportion: 0.1
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0004	1	0.2895	-
0.0225	50	0.2059	-
0.0449	100	0.1794	-
0.0674	150	0.1994	-
0.0898	200	0.2708	-
0.1123	250	0.1355	-
0.1347	300	0.0695	-
0.1572	350	0.117	-
0.1796	400	0.0601	-
0.2021	450	0.0873	-
0.2245	500	0.07	-
0.2470	550	0.0805	-
0.2694	600	0.0204	-
0.2919	650	0.1059	-
0.3143	700	0.1178	-
0.3368	750	0.1804	-
0.3592	800	0.0979	-
0.3817	850	0.1597	-
0.4041	900	0.1215	-
0.4266	950	0.0188	-
0.4490	1000	0.0738	-
0.4715	1050	0.0635	-
0.4939	1100	0.1439	-
0.5164	1150	0.0684	-
0.5388	1200	0.0732	-
0.5613	1250	0.0401	-
0.5837	1300	0.1223	-
0.6062	1350	0.1044	-
0.6286	1400	0.0717	-
0.6511	1450	0.0413	-
0.6736	1500	0.0544	-
0.6960	1550	0.1419	-
0.7185	1600	0.0284	-
0.7409	1650	0.0484	-
0.7634	1700	0.0049	-
0.7858	1750	0.0229	-
0.8083	1800	0.0739	-
0.8307	1850	0.0371	-
0.8532	1900	0.0213	-
0.8756	1950	0.0753	-
0.8981	2000	0.0359	-
0.9205	2050	0.0232	-
0.9430	2100	0.0507	-
0.9654	2150	0.0258	-
0.9879	2200	0.0606	-
1.0	2227	-	0.2105

Framework Versions

Python: 3.10.12
SetFit: 1.0.3
Sentence Transformers: 3.0.1
Transformers: 4.36.1
PyTorch: 2.0.1+cu118
Datasets: 2.20.0
Tokenizers: 0.15.0

Citation

BibTeX

@article{https://doi.org/10.48550/arxiv.2209.11055,
    doi = {10.48550/ARXIV.2209.11055},
    url = {https://arxiv.org/abs/2209.11055},
    author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
    keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Efficient Few-Shot Learning Without Prompts},
    publisher = {arXiv},
    year = {2022},
    copyright = {Creative Commons Attribution 4.0 International}
}

Downloads last month: -

Safetensors

Model size

0.1B params

Tensor type

F32

Model tree for kaustubhgap/kaustubh_setfit_1iteration

Base model

sentence-transformers/paraphrase-mpnet-base-v2

Finetuned

(320)

this model

Paper for kaustubhgap/kaustubh_setfit_1iteration

Efficient Few-Shot Learning Without Prompts

Paper • 2209.11055 • Published Sep 22, 2022 • 4

Evaluation results

Accuracy on Unknown
test set self-reported

0.549

Label	Training Sample Count
0	2
1	5
2	12
3	2
4	6
5	3
6	2
7	12
8	16
9	2
10	2
11	11
12	4
13	2
14	2
15	2
16	2
17	6
18	9
19	63
20	8
21	31
22	6
23	2
24	13
25	5
26	2
27	2
28	3
29	2
30	13
31	3
32	7
33	22
34	12
35	102
36	2
37	119
38	34
39	32
40	6
41	2
42	13
43	17
44	5
45	10
46	6
47	2
48	10
49	2
50	91
51	13
52	2
53	2
54	2
55	12
56	4
57	7
58	17
59	2
60	2
61	7
62	9
63	3
64	14
65	53
66	3
67	6
68	41
69	41
70	33
71	5
72	5
73	4
74	7
75	49
76	2
77	23
78	11
79	12
80	2
81	5
82	33
83	33
84	2
85	2
86	17
87	2
88	2
89	10
90	29
91	2
92	8
93	21
94	2
95	3
96	5
97	10
98	5
99	6
100	6
101	12
102	13
103	2
104	10
105	28
106	2
107	321
108	2
109	10
110	2
111	2
112	2
113	15
114	4
115	2
116	5
117	2
118	2

Label	Training Sample Count
0	2
1	5
2	12
3	2
4	6
5	3
6	2
7	12
8	16
9	2
10	2
11	11
12	4
13	2
14	2
15	2
16	2
17	6
18	9
19	63
20	8
21	31
22	6
23	2
24	13
25	5
26	2
27	2
28	3
29	2
30	13
31	3
32	7
33	22
34	12
35	102
36	2
37	119
38	34
39	32
40	6
41	2
42	13
43	17
44	5
45	10
46	6
47	2
48	10
49	2
50	91
51	13
52	2
53	2
54	2
55	12
56	4
57	7
58	17
59	2
60	2
61	7
62	9
63	3
64	14
65	53
66	3
67	6
68	41
69	41
70	33
71	5
72	5
73	4
74	7
75	49
76	2
77	23
78	11
79	12
80	2
81	5
82	33
83	33
84	2
85	2
86	17
87	2
88	2
89	10
90	29
91	2
92	8
93	21
94	2
95	3
96	5
97	10
98	5
99	6
100	6
101	12
102	13
103	2
104	10
105	28
106	2
107	321
108	2
109	10
110	2
111	2
112	2
113	15
114	4
115	2
116	5
117	2
118	2

Label	Training Sample Count
0	2
1	5
2	12
3	2
4	6
5	3
6	2
7	12
8	16
9	2
10	2
11	11
12	4
13	2
14	2
15	2
16	2
17	6
18	9
19	63
20	8
21	31
22	6
23	2
24	13
25	5
26	2
27	2
28	3
29	2
30	13
31	3
32	7
33	22
34	12
35	102
36	2
37	119
38	34
39	32
40	6
41	2
42	13
43	17
44	5
45	10
46	6
47	2
48	10
49	2
50	91
51	13
52	2
53	2
54	2
55	12
56	4
57	7
58	17
59	2
60	2
61	7
62	9
63	3
64	14
65	53
66	3
67	6
68	41
69	41
70	33
71	5
72	5
73	4
74	7
75	49
76	2
77	23
78	11
79	12
80	2
81	5
82	33
83	33
84	2
85	2
86	17
87	2
88	2
89	10
90	29
91	2
92	8
93	21
94	2
95	3
96	5
97	10
98	5
99	6
100	6
101	12
102	13
103	2
104	10
105	28
106	2
107	321
108	2
109	10
110	2
111	2
112	2
113	15
114	4
115	2
116	5
117	2
118	2