selina09/yt_setfit

SetFit with BAAI/bge-small-en-v1.5

This is a SetFit model that can be used for Text Classification. This SetFit model uses BAAI/bge-small-en-v1.5 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:

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

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: BAAI/bge-small-en-v1.5
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 2 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
0	'Have used Turbo Tax for years Never a problem I m pretty concerned now with the news that many of their users had their returns hacked by people who gained access to Turbo Tax and stole the information Not sure I will use it next year until I research how serious this is was ' 'Can t beat an Apple computer Like P KB best by test ' 'Works for Mac or Pc but not on widows '
1	'Would not install activation code not accepted Returned it ' 'Worth all four of the software programs which are included in this product ' 'The marketing information makes this software look like it should be fabulous lots of useful features that I would love to experiment with However the software just doesn t work I will keep using my very old JASC version of this software instead '

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("selina09/yt_setfit")
# Run inference
preds = model("Works great")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	34.9207	102

Label	Training Sample Count
0	123
1	41

Training Hyperparameters

• batch_size: (32, 32)
• num_epochs: (10, 10)
• max_steps: -1
• sampling_strategy: oversampling
• 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
0.0019	1	0.2503	-
0.0942	50	0.2406	-
0.1883	100	0.2029	-
0.2825	150	0.2207	-
0.3766	200	0.1612	-
0.4708	250	0.0725	-
0.5650	300	0.0163	-
0.6591	350	0.0108	-
0.7533	400	0.0153	-
0.8475	450	0.0486	-
0.9416	500	0.0191	-
1.0358	550	0.0207	-
1.1299	600	0.0148	-
1.2241	650	0.0031	-
1.3183	700	0.001	-
1.4124	750	0.0287	-
1.5066	800	0.0146	-
1.6008	850	0.0147	-
1.6949	900	0.0165	-
1.7891	950	0.0008	-
1.8832	1000	0.0165	-
1.9774	1050	0.0007	-
2.0716	1100	0.0129	-
2.1657	1150	0.0143	-
2.2599	1200	0.0006	-
2.3540	1250	0.0008	-
2.4482	1300	0.0047	-
2.5424	1350	0.0005	-
2.6365	1400	0.0116	-
2.7307	1450	0.0093	-
2.8249	1500	0.0211	-
2.9190	1550	0.0076	-
3.0132	1600	0.0047	-
3.1073	1650	0.0005	-
3.2015	1700	0.0064	-
3.2957	1750	0.014	-
3.3898	1800	0.0479	-
3.4840	1850	0.0005	-
3.5782	1900	0.0045	-
3.6723	1950	0.0188	-
3.7665	2000	0.0004	-
3.8606	2050	0.0122	-
3.9548	2100	0.0004	-
4.0490	2150	0.008	-
4.1431	2200	0.0245	-
4.2373	2250	0.005	-
4.3315	2300	0.0244	-
4.4256	2350	0.0208	-
4.5198	2400	0.0237	-
4.6139	2450	0.0005	-
4.7081	2500	0.0004	-
4.8023	2550	0.02	-
4.8964	2600	0.0004	-
4.9906	2650	0.0067	-
5.0847	2700	0.0099	-
5.1789	2750	0.0138	-
5.2731	2800	0.0192	-
5.3672	2850	0.0217	-
5.4614	2900	0.0056	-
5.5556	2950	0.0003	-
5.6497	3000	0.0052	-
5.7439	3050	0.0123	-
5.8380	3100	0.0136	-
5.9322	3150	0.0221	-
6.0264	3200	0.0235	-
6.1205	3250	0.0144	-
6.2147	3300	0.0174	-
6.3089	3350	0.007	-
6.4030	3400	0.0044	-
6.4972	3450	0.0003	-
6.5913	3500	0.007	-
6.6855	3550	0.0004	-
6.7797	3600	0.0384	-
6.8738	3650	0.0055	-
6.9680	3700	0.0056	-
7.0621	3750	0.0118	-
7.1563	3800	0.0143	-
7.2505	3850	0.0289	-
7.3446	3900	0.0301	-
7.4388	3950	0.0119	-
7.5330	4000	0.012	-
7.6271	4050	0.0138	-
7.7213	4100	0.0148	-
7.8154	4150	0.0003	-
7.9096	4200	0.0268	-
8.0038	4250	0.0131	-
8.0979	4300	0.0237	-
8.1921	4350	0.0004	-
8.2863	4400	0.0211	-
8.3804	4450	0.0092	-
8.4746	4500	0.005	-
8.5687	4550	0.0056	-
8.6629	4600	0.0168	-
8.7571	4650	0.0045	-
8.8512	4700	0.0184	-
8.9454	4750	0.0049	-
9.0395	4800	0.0047	-
9.1337	4850	0.0099	-
9.2279	4900	0.0054	-
9.3220	4950	0.0185	-
9.4162	5000	0.005	-
9.5104	5050	0.0004	-
9.6045	5100	0.013	-
9.6987	5150	0.0002	-
9.7928	5200	0.0187	-
9.8870	5250	0.0003	-
9.9812	5300	0.0081	-

Framework Versions

• Python: 3.10.12
• SetFit: 1.0.3
• Sentence Transformers: 3.0.1
• Transformers: 4.40.2
• PyTorch: 2.4.0+cu121
• Datasets: 2.21.0
• Tokenizers: 0.19.1

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}
}