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: 5 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
6.3.4 Effectiveness of Policy Implementation: Assesses how well policies are executed, supported, and monitored, ensuring that institutions deliver on their commitments and enable positive outcomes.	'nsure effective communication of agricultural related priorities to international partners through formal and non-formal donor coordination meetings. Strengthen capacity of the donor coordinatio' '.2. Development of a guideline for elaboration of climate change adaptation plans for settle- ments and dissemination among local self-government bodies Methodological assis- tance is ensured for lo- cal self-government bodies to plan and im- plement CCA measures in settle- ments Ministry of Territorial Administration and In- frastructures of the RA Urban Development Committee of the RA Ministry of Environ- ment of the RA Ministry of Emer- gency Situations of the RA 2022, 1st trimester Sources not prohibited by legislation (international donor organizations) 8,000 2.3. Development and implementation of the action plan to improve climate projections and early warning system The quality of climate projections is enhanced, and the early warning system is regulated as inputs for public policies for adaptation Ministry of Environment of the RA - 2022, 3rd trimester Sources not prohibited by legislation (international donor organizations) 3,000 2.4. Mapping and development of a database on CC related risks A database on CC re- lated risks is created to inform decision-making and elaboration of development programs Ministry of Environment of the RA - 2022, 1st trimester Sources not prohibited by legislation (inter- national donor organizations) 15,000 2.5. Development of training modules for senior officials, decision-makers and technical staff on CCA in various sectors to drive the NAP processes and implementation of the respective trainings Knowledge and capaci- ties of senior officials, decision-makers and technical officers in sec- toral governmental insti- tutions on strategic CC leadership are increased Ministry of Environment of the RA - 2022, 4th trimester Sources not prohibited by legislation (international donor organizations) 20,000 ' '38. Armenia’s NAP 2021-2025 consists of two sets of implementable measures: 39. The first is a set of cross-sectoral interventions aimed at strengthening the capacity of the country’s institutions to identify, prioritize, plan, attract funding for, and effectively implement adaptation measures, in addition to improving adaptation related public awareness and education at all levels. These were identified by determining those adaptation measures that: 1) were common to more than one key area (i.e., are cross-sectoral); 2) will deliver multiple benefits; and 3) will be beneficial for sectors and marzes in the result of coordinated implementation by key stakeholders. 40. These adaptation options will provide a starting point to focus initial national, regional and cross-sectoral action. 41. The second is a set of adaptation measures, specific to six priority sectors (water, agriculture, energy, settlements, health and tourism) and to two pilot marzes. The resulting SAPs and MAPs will become the blueprints for sectoral and marz adaptation, delineate a detailed 5-year strategic approaches for adaptation within each sector and marz, and will include a portfolio of project concept notes for priority investments in adaptation. Some of the mentioned SAPs and MAPs will be approved by the Government of the RA, while the others will be included in the respective guides to be disseminated among decision-makers and stakeholders. 42. It is anticipated that the adaptation measures presented in the NAP will be implemented or at least initiated during the 2021-2025 period, according to their degree of urgency. It is also clear, however, that their implementation will depend on funding, policy update/introduction etc. 43. The execution of most measures included in the Chapter 9 of this decision relies on the assumption that in addition to national budgetary efforts, the current level of international support for development and CC-oriented projects will be increased, and that additional climate finance for adaptation in the prioritized sectors will be attracted. The execution of the NAP will, nonetheless, require the proactive engagement of the Government and potentially, the allocation of new public resources. It is also assumed that over time, adaptation will become immersed in all new development projects in Armenia. 44. In view of the above considerations, it is the intention, in the coming years, and to the maximum extent possible, that elements of the NAP be integrated into the existing and planned cooperation programs with Armenia’s bilateral and multilateral partners. It should also be noted that the implementation of the mentioned sectoral and marz measures is only the starting point of a more in-depth adaptation process at the sector and marz levels, as it is expected that between 2021 and 2025 the necessary funding can be obtained, not only to initiate the execution of the identified measures, but also for the preparation and implementation of SAPs and MAPs in the sectors and marzes that have not been included in the first cycle of the NAP process. '
1.1. Food Security & Nutrition: Encompasses ensuring everyone’s access to sufficient, safe, and nutritious food, improving overall dietary intake and nutritional well-being.	'Improve systems of monitoring food security Identify criteria, develop Less Favourable Areas, LFA maps, and measures' 'stablish, maintain and replenish public food storage Monitor and prevent food waste and lost Establish close partnership with the partner to ensure synergies with other initiatives, such as school feeding, nutrition education' 'ncrease the production of vital local foods Improve the trade balance for selected commodities where import substitution is economically viable'
5.2 Resilience Capacities (absorptive, adaptive & transformative): Promotes building skills, diversifying options, strengthening networks, and improving surveillance systems so that communities, ecosystems, and value chains can withstand and recover from disruptions.	'Effective management of laboratory capacities in the areas of food safety, veterinary and phytosan itary Establish a system of productive cooperation between public and private laboratories and' 'mprove plant protection system regulations and enforcement Monitoring of plant quarantine and non-quarantine pests and phytosanitary assessment Develop system for advanced plant protection Develop system of predicting and rapid alert for harmful plant organisms Registration of pesticides (including imported) and creation of a single register; Develop plant protection system using digital technologies and monitoring system for pest and disease control' 'Food safety is one of the most important and urgent problems in Armenia that requires solutions based on modern requirements and standards. The food safety system in the Republic of Armenia does not yet fully guarantee safe and high-quality food for consumers as well as enhanced competitiveness of locally produced food products in export and domestic markets. Compliance with food safety standards will also enhance the overall competitiveness of agriculture, particularly in the export context. In the context of food safety, ensuring the safety of livestock products at the farm level as well as at the level of the last point in this particular value chain - processing (production of dairy and meat products) is of particular importance. For example, currently in Armenia brucellosis is the most important disease that transfers form animals to humans, which is a threat from the food safety point of view, in terms of diseases transferable from milk and dairy products to humans.'
1.2. Diet quality: Focuses on the balance, diversity, and healthfulness of what people eat, aiming to prevent malnutrition and diet-related diseases.	'The objective of ensuring adequate food utilization in Armenia is envisaged to be achieved by involving more nutritious food products in the population’s diet, upgrading 20 the sanitation and food safety standards along with bringing it up to a new level. The following sub-objectives have been stipulated: 1. Providing the population with food that is fully compliant with health standards. improvement of food quality and safety level. 2. Ensuring adequate level of food safety, veterinary and phytosanitary security syste' 'The following challenges have been defined under the pillar of Food utilization in Armenia: 1. High proportion of ready-to-use food wasting and losses; 2. Low level of provision of nutritious food to the population that meets health standards, including: ▪ insufficient level of awareness on healthy food and lifestyle; ▪ inadequate balancing of the food needed for nutrition; 3. Insufficient level of surveillance over food quality and security' ' Mere food availability and accessibility are not enough, people should have access to "safe and nutritious food". The food consumed should supply sufficient energy to empower the consumer to carry out physical activity. Utilization (consumption) of food is characterized by the use of food in compliance with biological and social conditions. Food should be used efficiently to achieve a state of nutritional well-being. This includes the actual quantity and quality of food designed for consumption, as well awareness needed for the right diet choices. 50. Utilization of food also implies factors such as safe drinking water and appropriate sanitary and hygienic conditions to avoid the spread of disease, as well as awareness of food preparation and storage procedures. Consequently, utilization of food contains a set of aspects that depend on the consumer's understanding of what food to choose and how to prepare and store them. Over time, the risks and benefits of human health and welfare grow, which are linked to industrialization, intensification and concentration of production and international trade expansion with longer, more complicated food supply chains. In addition, it is necessary to dramatically improve the scope of the state food surveillance and ensure the level of food security. The strengthening of the food security system will help to improve consumer protection. 51. Recent studies have proved that there is a high prevalence of malnutrition and micronutrient deficiency in Armenia. About 21% of children under the age of 5 are underweight, and 17% are overweight. Child stunting is evidently related to household poverty and poor consumption, as well as poor care and feeding accompanied by low education level of a mother. On the other hand, the prevalence of excess weight is the same across poor and rich households, which indicates the need for greater awareness of healthy food and lifestyle of the population'
6.3.3 Awareness and use of the evidence-based / agrifood systems approach: Encourages long-term, integrated planning for agrifood systems, guided by robust data, stakeholder consensus, and strategic foresight.	"In addition, since many constraints to Armenia's agriculture extend beyond the agricultural sector, this Strategy acknowledges the importance of partnerships and includes a final section with calls to action and ideas for collaboration with other Armenian line ministries and Governmental institutions and initiatives. These include, for example, the Ministries of High Tech, Finance, Environment, Territorial Administration and Infrastructure, Education, Science, Culture and Sport, and other Governmental bodies and programmes such as the Work Armenia initiative. The main key indicators of the Strategy are presented in Table 1" 'The strategy is firmly grounded in global declarations and aspirations, such as the Sustainable Development Goals. It builds on the lessons learned under the Government of Armenia (GoA) Programme of 2019, the GoA Action Programme (2019-2023), and previous strategic Governmental interventions, including the Mid- term Expenditures Framework of RA (2020-2022). Furthermore, the strategy reflects major points of cooperation highlighted in existing cooperation agreements, such as points related to agriculture development, agro-tourism, and agricultural statistics in the CEPA (EU-Armenian Comprehensive and Enhanced Partnership Agreement).' 'Economic policy for ensuring sustainable and accelerated economic growth; \uf0b7 Active social and income policy for vulnerable groups of population (including the poor); \uf0b7 Modernization of governance system, including improved effectiveness of state governance and ensuring accelerated growth of the resource envelope at the disposal of the state.'

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("faodl/setfit-paraphrase-mpnet-base-v2-5ClassesDesc-10augmented")
# Run inference
preds = model("Since the development of the first National Nutrition Strategy of Timor-Leste in 2004, there have been several emerging global, regional and national initiatives to accelerate improvements in nutritional status. ")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	6	93.0804	1014

Label	Training Sample Count
6.3.4 Effectiveness of Policy Implementation: Assesses how well policies are executed, supported, and monitored, ensuring that institutions deliver on their commitments and enable positive outcomes.	32
1.1. Food Security & Nutrition: Encompasses ensuring everyone’s access to sufficient, safe, and nutritious food, improving overall dietary intake and nutritional well-being.	72
5.2 Resilience Capacities (absorptive, adaptive & transformative): Promotes building skills, diversifying options, strengthening networks, and improving surveillance systems so that communities, ecosystems, and value chains can withstand and recover from disruptions.	27
1.2. Diet quality: Focuses on the balance, diversity, and healthfulness of what people eat, aiming to prevent malnutrition and diet-related diseases.	28
6.3.3 Awareness and use of the evidence-based / agrifood systems approach: Encourages long-term, integrated planning for agrifood systems, guided by robust data, stakeholder consensus, and strategic foresight.	40

Training Hyperparameters

batch_size: (8, 8)
num_epochs: (1, 1)
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
l2_weight: 0.01
seed: 42
eval_max_steps: -1
load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0003	1	0.2267	-
0.0132	50	0.2304	-
0.0264	100	0.2207	-
0.0396	150	0.1975	-
0.0528	200	0.1701	-
0.0661	250	0.1499	-
0.0793	300	0.1411	-
0.0925	350	0.113	-
0.1057	400	0.1	-
0.1189	450	0.0741	-
0.1321	500	0.0898	-
0.1453	550	0.0665	-
0.1585	600	0.0582	-
0.1717	650	0.0537	-
0.1849	700	0.0337	-
0.1982	750	0.0443	-
0.2114	800	0.0345	-
0.2246	850	0.0408	-
0.2378	900	0.0354	-
0.2510	950	0.0332	-
0.2642	1000	0.0326	-
0.2774	1050	0.0299	-
0.2906	1100	0.0285	-
0.3038	1150	0.0359	-
0.3170	1200	0.0355	-
0.3303	1250	0.035	-
0.3435	1300	0.0257	-
0.3567	1350	0.0188	-
0.3699	1400	0.0303	-
0.3831	1450	0.0226	-
0.3963	1500	0.0322	-
0.4095	1550	0.0235	-
0.4227	1600	0.0192	-
0.4359	1650	0.0303	-
0.4491	1700	0.033	-
0.4624	1750	0.0209	-
0.4756	1800	0.0218	-
0.4888	1850	0.0225	-
0.5020	1900	0.0236	-
0.5152	1950	0.0228	-
0.5284	2000	0.019	-
0.5416	2050	0.019	-
0.5548	2100	0.0116	-
0.5680	2150	0.0209	-
0.5812	2200	0.016	-
0.5945	2250	0.0234	-
0.6077	2300	0.0165	-
0.6209	2350	0.0159	-
0.6341	2400	0.0172	-
0.6473	2450	0.0208	-
0.6605	2500	0.0264	-
0.6737	2550	0.0267	-
0.6869	2600	0.0285	-
0.7001	2650	0.0195	-
0.7133	2700	0.0253	-
0.7266	2750	0.0159	-
0.7398	2800	0.0284	-
0.7530	2850	0.0216	-
0.7662	2900	0.0179	-
0.7794	2950	0.0193	-
0.7926	3000	0.0159	-
0.8058	3050	0.0254	-
0.8190	3100	0.0209	-
0.8322	3150	0.0242	-
0.8454	3200	0.0221	-
0.8587	3250	0.016	-
0.8719	3300	0.0191	-
0.8851	3350	0.0218	-
0.8983	3400	0.0194	-
0.9115	3450	0.0168	-
0.9247	3500	0.0274	-
0.9379	3550	0.0202	-
0.9511	3600	0.0226	-
0.9643	3650	0.0251	-
0.9775	3700	0.0264	-
0.9908	3750	0.018	-
1.0	3785	-	0.2535

Framework Versions

Python: 3.11.11
SetFit: 1.1.1
Sentence Transformers: 3.4.1
Transformers: 4.50.0
PyTorch: 2.6.0+cu124
Datasets: 3.5.0
Tokenizers: 0.21.1

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}
}

faodl
/

setfit-paraphrase-mpnet-base-v2-5ClassesDesc-10augmented