EuriskoMobility/trained_SetFit_follow-up_model_taline

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

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/all-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:

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: sentence-transformers/all-mpnet-base-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 384 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
1	' Who was the Germany national team captain during the 2006 World cup? Michael Ballack was the Germany national team captrain during the 2006 world cup How old was he? ' ' Who was the Germany national team captain during the 2006 World cup? Michael Ballack was the Germany national team captrain during the 2006 world cup Who won it back then? ' ' How old was Ronaldo when he moved to Real Madrid? Ronaldo moved to Real Madrid after leaving Inter when he was 25 years old. What year did he leave? '
0	' Which ocean surrounds Antarctica? The ocean that surrounds Antarctica is the Southern Ocean. What challenges do scientists face when conducting research in Antarctica? ' ' Name a country in Oceania. A country in Oceania is Australia. What are some of the popular tourist attractions in Oceania? ' " What's the significance of the Suez Canal? The Suez Canal holds great importance as a crucial Egyptian waterway that links the Mediterranean Sea to the Red Sea. It plays a pivotal role in enhancing maritime trade and transportation between Europe and Asia, providing ships with a shorter and safer route compared to the arduous journey around the southern tip of Africa. How has the Suez Canal impacted global trade? "

Evaluation

Metrics

Label	Accuracy
all	0.9348

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("setfit_model_id")
# Run inference
preds = model("<Question> What is the highest grossing movie at the box office? </Question> <Answer> The highest-grossing movie at the box office is Avatar. </Answer> <Question> How much money did the movie make? </Question>")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	14	44.4406	221

Label	Training Sample Count
0	240
1	248

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (3, 3)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• body_learning_rate: (2e-05, 2e-05)
• head_learning_rate: 2e-05
• loss: CosineSimilarityLoss
• distance_metric: cosine_distance
• margin: 0.25
• end_to_end: False
• use_amp: False
• warmup_proportion: 0.1
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0008	1	0.5762	-
0.0410	50	0.2742	-
0.0820	100	0.2188	-
0.1230	150	0.0586	-
0.1639	200	0.0194	-
0.2049	250	0.0028	-
0.2459	300	0.0004	-
0.2869	350	0.0003	-
0.3279	400	0.0002	-
0.3689	450	0.0001	-
0.4098	500	0.0001	-
0.4508	550	0.0001	-
0.4918	600	0.0001	-
0.5328	650	0.0006	-
0.5738	700	0.0001	-
0.6148	750	0.0001	-
0.6557	800	0.0001	-
0.6967	850	0.0001	-
0.7377	900	0.0001	-
0.7787	950	0.0001	-
0.8197	1000	0.0001	-
0.8607	1050	0.0001	-
0.9016	1100	0.0001	-
0.9426	1150	0.0001	-
0.9836	1200	0.0	-
0.0008	1	0.0	-
0.0410	50	0.0	-
0.0820	100	0.0003	-
0.1230	150	0.0005	-
0.1639	200	0.0013	-
0.2049	250	0.0008	-
0.2459	300	0.0	-
0.2869	350	0.0	-
0.3279	400	0.0	-
0.3689	450	0.0	-
0.4098	500	0.0	-
0.4508	550	0.0	-
0.4918	600	0.0	-
0.5328	650	0.0	-
0.5738	700	0.0	-
0.6148	750	0.0	-
0.6557	800	0.008	-
0.6967	850	0.0285	-
0.7377	900	0.012	-
0.7787	950	0.0073	-
0.8197	1000	0.0013	-
0.8607	1050	0.0	-
0.9016	1100	0.0	-
0.9426	1150	0.0	-
0.9836	1200	0.0013	-
1.0246	1250	0.0013	-
1.0656	1300	0.0	-
1.1066	1350	0.0	-
1.1475	1400	0.0	-
1.1885	1450	0.0	-
1.2295	1500	0.0	-
1.2705	1550	0.0	-
1.3115	1600	0.0	-
1.3525	1650	0.0022	-
1.3934	1700	0.0	-
1.4344	1750	0.0	-
1.4754	1800	0.0	-
1.5164	1850	0.0013	-
1.5574	1900	0.0	-
1.5984	1950	0.0	-
1.6393	2000	0.0	-
1.6803	2050	0.0	-
1.7213	2100	0.0	-
1.7623	2150	0.0	-
1.8033	2200	0.0	-
1.8443	2250	0.0048	-
1.8852	2300	0.0023	-
1.9262	2350	0.0049	-
1.9672	2400	0.0012	-
2.0082	2450	0.0	-
2.0492	2500	0.0	-
2.0902	2550	0.0	-
2.1311	2600	0.0	-
2.1721	2650	0.0	-
2.2131	2700	0.0	-
2.2541	2750	0.0	-
2.2951	2800	0.0	-
2.3361	2850	0.0	-
2.3770	2900	0.0	-
2.4180	2950	0.0	-
2.4590	3000	0.0	-
2.5	3050	0.0	-
2.5410	3100	0.0	-
2.5820	3150	0.0	-
2.6230	3200	0.0	-
2.6639	3250	0.0	-
2.7049	3300	0.0	-
2.7459	3350	0.0	-
2.7869	3400	0.0	-
2.8279	3450	0.0	-
2.8689	3500	0.0	-
2.9098	3550	0.0007	-
2.9508	3600	0.0	-
2.9918	3650	0.0	-

Framework Versions

• Python: 3.10.12
• SetFit: 1.1.0
• Sentence Transformers: 3.2.1
• Transformers: 4.42.2
• PyTorch: 2.5.1+cu121
• Datasets: 3.1.0
• 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}
}