--- library_name: setfit tags: - setfit - sentence-transformers - text-classification - generated_from_setfit_trainer datasets: - Kevinger/hub-report-dataset metrics: - accuracy widget: - text: 'A 16-acre property once home to the long-shuttered Foxborough State Hospital will soon provide housing for 141 low-income senior households. Walnut Street, an affordable housing project being developed by the Affordable Housing Services Collaborative and Onyx, will turn land that has been vacant for decades into much-needed affordable housing. “Housing is empowering. No matter our age, it is a comfort not to worry about whether we can afford a place,” Onyx CEO Chanda Smart said at a press conference Thursday. “Senior housing for the town of Foxborough means that seniors who worked and raised their families here in Foxborough still have the opportunity to remain here.” Foxborough State Hospital opened in 1889 as the Massachusetts Hospital for Dipsomaniacs and Inebriates for treatment of alcoholism, according to the National Park Service, and was later converted to a standard psychiatric hospital. It closed in 1975, and parts of the property have already been redeveloped over the years. The Foxborough Housing Authority first began working on the project back in 2011. The land was transferred to the agency from the state in 2017 to be used for affordable housing. Acting Town Manager Paige Duncan told MassLive that the town held a number of community meetings to decide what to build on the property. “It was controversial, but what came out was a clear support for senior housing,” she said. “We really tried to address the needs of the community and we came up with a project that was sensitive to the area. We didn’t want a big block of buildings that towered over the neighborhood.” After that, she said, there was overwhelming support for the project. The permits were filed in February and approved by April, an almost unheard-of timeline. The finished project will provide 141 new apartments for residents age 55 and over. Of those, 35 will be reserved for people making 30% or less of the area median income, and 85 will be for those making 60% AMI. Foxborough residents will be given preference for 70% of the units. A second phase of the project once this one is complete will add approximately 60 more units. Greg Spiers, chairman of the Housing Authority, said the new senior housing was badly needed, noting there are about 5,500 elderly and disabled people on public housing waiting lists in Massachusetts. “With 195 of those on that list Foxborough residents, that 70% local preference for first-time rentals is one of our goals,” he said. “The need is so great for affordable housing in our area and the entire state.” Housing and Livable Communities Secretary Ed Augustus praised the town for its dedication to creating more affordable housing, even though more than 10% of its total housing units qualify as affordable. The 10% threshold is the state requirement to stop projects being filed under Chapter 40B, a law which allows affordable housing developments to bypass certain local permitting requirements. “You know that that is just an arbitrary number, but the real needs are significantly more than that,” Augustus said. “We need more communities to take note of what Foxborough is doing.” Lt. Gov. Kim Driscoll said the project is a good example of the use of surplus state land for housing. Gov. Maura Healey’s housing bond bill filed in October included a proposed $30 million that would support similar projects to use underutilized state property for housing. Healey also issued an executive order requesting state agencies to conduct an audit of their property to find land any surplus land suitable for this purpose. “Converting state-owned land to another entity can be a little bit of a torturous pathway. We know that building all the resources you need takes time,” Driscoll said Thursday. “How do we leverage the cost of land, which is one of the reasons housing is so expensive, to build the type of housing we need, but do it in a shorter timetable? That’s what this (project) is all about.” The project has received more than $25 million in state and federal funding, including through American Rescue Plan Act rental funds and state and federal Low Income Housing Tax Credits. Work on the site has not yet started.' - text: '“I was on my co-op last year for, like, a straight year, so coming back to campus feels kind of nerve-wracking,” said Jasmine Rodriguez, 21. “But I feel more experienced than I did in my first year. I had a lot of anxiety in my first year, but now it’s been really chill.” As about a dozen Northeastern University students went around a conference table talking about their college experiences, voices were soft and answers halting, at least initially. Gradually, though, the students at this check-in meeting last fall began to open up and speak candidly about the challenges and adjustments of college life. Advertisement The students were Black, Latino, and Asian American and ranged from first-years to seniors, mostly from neighborhoods across Boston; the majority were the first generation of their family to attend college. Most were their high schools’ valedictorians — hardworking, smart students who excelled despite lacking the advantages of many peers. That’s where The Valedictorian Project came in. The Boston-based nonprofit was founded in 2020 in response to the Boston Globe’s award-winning 2019 investigative series, The Valedictorians Project, which found that the city’s best and brightest public school students often encounter major obstacles to their academic and professional goals. (The Globe is not involved with the organization.) The Valedictorian Project matches participating high school graduates with peer mentors close to their age and a senior mentor who is an experienced professional in their intended line of work. It also provides a $500 stipend for books and other necessities, and supplemental support through partnerships with other organizations to help students navigate their new lives on campus and choose career paths. “Many of our mentors are first-gen college students themselves,” cofounder and executive director Amy McDermott said in an interview. “Many navigated very similar personal backgrounds to our mentees. I hear often in our mentor interviews, they want to be that person that they wish they had in navigating college.” Advertisement This academic year marks a milestone for the organization, as its first cohort of college freshmen are now seniors. McDermott said the organization began by inviting Boston valedictorians to participate in its first year, then added students from Lawrence in year two, Brockton and Worcester in 2022, and Chelsea last spring. Jasmine Rodriguez took part in a roundtable discussion at Northeastern University for students participating in The Valedictorian Project. Jonathan Wiggs/Globe Staff Mentor John Marley, 30, of Taunton, said the organization helps level the playing field for young people who don’t come from privileged backgrounds. “Students from wealthier families have always had these mentorship relationships, always had these connections, and those things are just unseen,” said Marley, an attorney whose family came to the United States from Jamaica when he was 5. “Unfairly or not ... it’s always advantaged a particular group and class of students over another. And I think they do a good job addressing that.” This academic year, The Valedictorian Project is supporting 140 students, of whom about three-quarters are first-generation college students and roughly 85 percent are people of color, according to McDermott. Besides Northeastern, students in the program attend Boston University, Harvard, MIT, Tufts, Brown, Yale, Stanford, and other colleges around the country, she said. As a student of color at an expensive private university, Rodriguez said, “You have to physically go out and try to find people that look like you. And I feel like for everyone else, it’s very easy. They find them in their classes. But it’s like, in my classes there’ll be like one other Black or Hispanic person.” Advertisement Rodriguez, a Dorchester native majoring in communications and sociology, recently spent a year as a social media co-op for an organization that supports domestic violence victims. She is drawn to work that will help others, she said, because she saw people in need in her neighborhood and her own family as she grew up. “I saw a lot of people that look like me struggle and go through a lot of things,” she said. “My mom is an immigrant. … We grew up on Section 8 [housing assistance]; we grew up on food stamps and stuff like that.” Ciana Omnis participated in a Northeastern University roundtable discussion for students participating in The Valedictorian Project. Jonathan Wiggs/Globe Staff Ciana Omnis, 20, a third-year industrial engineering major who grew up in Florida, moved to Dorchester at age 14, and was the 2021 valedictorian at Brighton High School. She is the eldest of three children, so she can’t lean on older siblings for advice, she said. Her father, a truck driver who immigrated to the United States from Haiti, didn’t complete high school, she said, while her mother, a health care administrator, completed an associate’s degree but doesn’t yet have her bachelor’s. “I’ve met a lot of people in college who have parents who have done four-year degrees or whatnot, or even other kinds of higher education, so they’re able to get advice from their parents,” Omnis said. “For me, it’s been a bit harder, because I have to kind of figure out certain things on my own.” Advertisement Her mentors help fill that gap, she said, and the program helps her “meet other people who have the same background as me.” After they met through a Valedictorian Project event, John Le, who was the 2022 valedictorian at East Boston High School, became friends with Connor Lashley, the 2022 valedictorian at Jeremiah E. Burke High School in Dorchester. “One of the issues is socializing, like making a friend group, because from my experience, from each class you kind of like meet people there, but if you’re not in the same major, you might not be able to maintain a relationship with them,” said Le, 20. The Valedictorian Project, he added, “has really been helpful to meet people at Northeastern and ... find people with similar interests.” Lashley, 19, said his mentors have helped him learn how to network with others in his field and steered him toward scholarship opportunities, and he can count on their support whenever he needs it. “They’re pretty much available the same day if stuff comes up,” he said. Connor Lashley (left) and John Le took part in a roundtable discussion at Northeastern University for students participating in The Valedictorian Project. Jonathan Wiggs/Globe Staff Jeremy C. Fox can be reached at jeremy.fox@globe.com. Follow him @jeremycfox.' - text: 'LEVERETT — Dakin Humane Society announced Wednesday that it has sold its former animal shelter at 63 Montague Road in Leverett to Better Together Dog Rescue. The news release didn’t include a sales price for the 3,480-square-foot building on 5 acres of land. But records at the Franklin County Registry of Deeds show the sale was for $575,000.' - text: 'Joan Acocella, a cultural critic whose elegant, erudite essays about dance and literature appeared in The New Yorker and The New York Review of Books for more than four decades, died on Sunday at her home in Manhattan. She was 78. Her son, Bartholomew Acocella, said the cause was cancer. Ms. Acocella (pronounced ack-ah-CHELL-uh) wrote deeply about dancers and choreographers, including Mikhail Baryshnikov, Suzanne Farrell and George Balanchine. She scrutinized the vicissitudes of the New York City Ballet as well as the feats of the ballroom-dancing pros and celebrity oafs of the popular TV series “Dancing With the Stars.” She was The New Yorker’s dance critic from 1998 to 2019 and freelanced for The Review for 33 years. Her final articles for The Review were a two-part commentary in May on the biography “Mr. B: George Balanchine’s 20th Century,” by Jennifer Homans, her successor as The New Yorker’s dance critic. “What she wrote for us,” Emily Greenhouse, the editor of The Review, said in an email, “was often mischievous and always delicious — on crotch shots and cuss words, on Neapolitan hand gestures and Isadora Duncan’s emphasis on the solar plexus.”' - text: 'StreetsblogMASS relies on the generous support of readers like you. Help us meet our year-end fundraising goals – give today! Last week, the labor union that represents most Boston police officers ratified a new contract that will introduce a number of reforms – including one that will start allowing civilians to take unwanted traffic detail shifts at construction sites. Under the former contract, Boston Police officers were the only people allowed to direct traffic for events and at construction sites. And they got paid extremely handsomely to do so: Boston police working as flaggers take home $60 an hour. In spite of that lucrative pay, Boston has a lot of construction sites, and fewer and fewer people who want to wear a police uniform. Boston City Councilor Kendra Lara told StreetsblogMASS earlier this year that over 40 percent of requests for police details at construction sites were going unfilled. The new labor contract removes a key barrier to reforming this system. But there is still a city ordinance on the books that requires at least one Boston Police officer at every city construction site "to protect the safety and general welfare of the public and to preserve the free circulation of traffic." A press spokesperson for Mayor Michelle Wu told StreetsblogMASS last week that their office is aware of the ordinance and has "identified multiple legal paths to implementing the new collective bargaining agreement." Old rules created absurd delays for street projects Councilor Lara also told StreetsblogMASS that many privately-run construction sites will simply ignore the law and do their work without a flagger if nobody responds to their requests for a detail. But construction firms who are sticklers for the rules can end up waiting months before a cop shows up to let them get their work done. That''s what happened earlier this year in Oak Square, where the MBTA waited a full year for a police detail to show up so that they could paint some new crosswalks on Washington Street in Oak Square. Neighbors report that those crosswalks finally got painted in August – after a year-long wait. New contract hikes pay, allows civilian flaggers For all these reasons, allowing civilian flaggers at construction sites had been one of the city''s key points of negotiation for a new collective bargaining agreement with its police union. Police details will still be required at "high-priority" events and construction sites, which involve major streets, busy intersections, or major events that anticipate over 5,000 attendees. The new contract would also pay cops who work those high-priority details "the highest overtime rate of the most senior officer." At other worksites, such as those along quiet neighborhood streets, Boston Police would still get the right of first refusal to fill traffic details. But if no Boston Police are interested, the work can be offered to other non-BPD certified officers, including campus police and retired Boston cops. If people with those qualifications still aren''t interested, construction contractors can then offer the job to civilian workers. The agreement further specifies that anyone directing traffic in those lower-priority sites will earn $60 per hour. The new agreement will also ban cops from double-booking their shifts, which allowed some to get paid twice for the same period of time when one detail ended early. Incredibly, the police department is still using a labor-intensive paper-based system to assign details in each police district. The new agreement will allow for a citywide electronic scheduling system.' pipeline_tag: text-classification inference: false base_model: sentence-transformers/paraphrase-mpnet-base-v2 model-index: - name: SetFit with sentence-transformers/paraphrase-mpnet-base-v2 results: - task: type: text-classification name: Text Classification dataset: name: Kevinger/hub-report-dataset type: Kevinger/hub-report-dataset split: test metrics: - type: accuracy value: 0.6529242569511026 name: Accuracy --- # SetFit with sentence-transformers/paraphrase-mpnet-base-v2 This is a [SetFit](https://github.com/huggingface/setfit) model trained on the [Kevinger/hub-report-dataset](https://huggingface.co/datasets/Kevinger/hub-report-dataset) dataset that can be used for Text Classification. This SetFit model uses [sentence-transformers/paraphrase-mpnet-base-v2](https://huggingface.co/sentence-transformers/paraphrase-mpnet-base-v2) as the Sentence Transformer embedding model. A OneVsRestClassifier 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](https://www.sbert.net) 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/paraphrase-mpnet-base-v2](https://huggingface.co/sentence-transformers/paraphrase-mpnet-base-v2) - **Classification head:** a OneVsRestClassifier instance - **Maximum Sequence Length:** 512 tokens - **Training Dataset:** [Kevinger/hub-report-dataset](https://huggingface.co/datasets/Kevinger/hub-report-dataset) ### Model Sources - **Repository:** [SetFit on GitHub](https://github.com/huggingface/setfit) - **Paper:** [Efficient Few-Shot Learning Without Prompts](https://arxiv.org/abs/2209.11055) - **Blogpost:** [SetFit: Efficient Few-Shot Learning Without Prompts](https://huggingface.co/blog/setfit) ## Evaluation ### Metrics | Label | Accuracy | |:--------|:---------| | **all** | 0.6529 | ## Uses ### Direct Use for Inference First install the SetFit library: ```bash pip install setfit ``` Then you can load this model and run inference. ```python from setfit import SetFitModel # Download from the 🤗 Hub model = SetFitModel.from_pretrained("Kevinger/setfit-hub-multilabel-example") # Run inference preds = model("LEVERETT — Dakin Humane Society announced Wednesday that it has sold its former animal shelter at 63 Montague Road in Leverett to Better Together Dog Rescue. The news release didn’t include a sales price for the 3,480-square-foot building on 5 acres of land. But records at the Franklin County Registry of Deeds show the sale was for $575,000.") ``` ## Training Details ### Training Set Metrics | Training set | Min | Median | Max | |:-------------|:----|:---------|:-----| | Word count | 53 | 386.3906 | 2161 | ### Training Hyperparameters - batch_size: (8, 8) - num_epochs: (1, 1) - max_steps: -1 - sampling_strategy: oversampling - num_iterations: 75 - 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 - seed: 42 - eval_max_steps: -1 - load_best_model_at_end: False ### Training Results | Epoch | Step | Training Loss | Validation Loss | |:------:|:----:|:-------------:|:---------------:| | 0.0008 | 1 | 0.1304 | - | | 0.0417 | 50 | 0.1596 | - | | 0.0833 | 100 | 0.132 | - | | 0.125 | 150 | 0.0064 | - | | 0.1667 | 200 | 0.0017 | - | | 0.2083 | 250 | 0.0004 | - | | 0.25 | 300 | 0.0001 | - | | 0.2917 | 350 | 0.0002 | - | | 0.3333 | 400 | 0.0003 | - | | 0.375 | 450 | 0.0002 | - | | 0.4167 | 500 | 0.0001 | - | | 0.4583 | 550 | 0.0002 | - | | 0.5 | 600 | 0.0002 | - | | 0.5417 | 650 | 0.0002 | - | | 0.5833 | 700 | 0.0001 | - | | 0.625 | 750 | 0.0001 | - | | 0.6667 | 800 | 0.0001 | - | | 0.7083 | 850 | 0.0001 | - | | 0.75 | 900 | 0.0 | - | | 0.7917 | 950 | 0.0001 | - | | 0.8333 | 1000 | 0.0001 | - | | 0.875 | 1050 | 0.0001 | - | | 0.9167 | 1100 | 0.0001 | - | | 0.9583 | 1150 | 0.0 | - | | 1.0 | 1200 | 0.0001 | - | ### Framework Versions - Python: 3.10.12 - SetFit: 1.0.3 - Sentence Transformers: 2.3.1 - Transformers: 4.35.2 - PyTorch: 2.1.0+cu121 - Datasets: 2.17.0 - Tokenizers: 0.15.2 ## Citation ### BibTeX ```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} } ```