Tomiwajin/setfit_email_classifier

SetFit with sentence-transformers/all-MiniLM-L6-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/all-MiniLM-L6-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-MiniLM-L6-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 256 tokens
• Number of Classes: 7 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
next-phase	"Next step: Assessment 📋 for the Product Manager role at StartupXYZ Hi Thomas, Thank you again for your interest in the Product Manager position at StartupXYZ. As part of our hiring process, the next step is to complete an assessment that will help us better understand your skills and suitability for the role. Here's what to expect: • Assessment details: Product str" "Next Steps - Front-End Engineer Hey Oluwatomiwa,\nWe're excited to invite you to the next phase of the Front-End Engineer role.\nBefore moving forward, please ensure your location is on the list of accepted locations.\nImportant Notice\n\nIf you currently have or previously had credentials with Outlier or a related platform, please do" 'Coding Assessment - Backend Developer Position Dear Kevin, We appreciate your interest in the Backend Developer role at CloudSync Technologies. As the next step in our selection process, we invite you to complete our technical coding assessment. This assessment has been carefully designed to evaluate your programming skills and problem-solving a'
interview	"Interview for DevOps Engineer at ServerMax Hi Daniel, Thanks again for taking the time to chat with me on the phone! I'm very happy to move you to the next stage of our hiring process — a 45-minute video interview. This interview will include me and my colleague Tom Rodriguez, our Infrastructure Lead. If you'd like to learn a little about hi" "Video interview for Social Media Manager at BuzzMarketing Hi Taylor, Thank you for your application for the Social Media Manager position with BuzzMarketing. We're excited to learn more about you and your qualifications! We would like to invite you to a video interview with Christina Park, our Digital Marketing Director. This will be a chance for us to dis" 'Final round interview for Marketing Director at BrandBoost Hi Michelle, Congratulations on making it to the final round of interviews for the Marketing Director position! We would like to invite you to a final in-person interview with our executive team including CEO Jonathan Miller and CMO Patricia Davis. This will be a chance for us to discuss your strate'
not-job-status-update	"Jobs! Hi Seth,\n\nI found a job you may be interested in: 100% REMOTE - Senior Fullstack Engineer\n\nIf you'd like to apply, let me know in a quick response with your updated resume. Full job details below.\n\nf you are a Senior Software Engineer with Python and React experience, please read on!\n\nWe headquarter" 'Oluwatomiwa, you have new application updates this week Check out the status of your applications on LinkedIn Check out the status of your applications on LinkedIn Here are the latest updates from the past week Junior Software Engineer Fortune 500 · Plano, TX (On-site) No longer accepting applications Software Quality Assurance Engineer ChronicCar' 'Junior Software Engineer role at AmeriNat: you would be a great fit! Hey! Check out the latest industry content about career advice, salary negotiations, and interview tips, among other topics. Explore now! Jobs for you Jobs for you We‚Äôre on a mission to connect you with a dream job. To help us refine this list, search for more jobs AmeriNat 4.1 ‚òÖ Junior Softwar'
not-job-related	'Welcome to Idealist! Four actions you can take right now to get started Hi Oluwatomiwa, My name is Ami, and I’m the founder and executive director of Idealist. We started Idealist in the summer of 1995—on one old computer and with no full-time staff—to help connect people around the world with opportunities to do' 'New arrivals are here SHEIN Shop at SHEIN for the latest trends! Shop at SHEIN for the latest trends! Unsubscribe
applied	'Thank you for applying! Dear Name,\n\nThank you for your interest in a career at Delta Dental of Iowa. We have received your application for Software Development Intern.\nIn the event that we wish to arrange a personal interview, we will contact you. Again, thank you for your interest in employment at Delta Dental of Iowa.' 'Thank You For Applying! Dear Name,\nThank you for applying! Your application will be taken into careful consideration and you will hear back from us after we review your application.\n\n\nBest Regards,\n\nBracco Human Resources Team' 'Thank you for applying to Passes Name,\n\nThanks for applying to Passes. Your application has been received and we will review it right away.\n\nIf your application seems like a good fit for the position we will contact you soon.\n\nRegards,\nPasses\n\n** Please note: Do not reply to this email. This email is sent from an unattended mailbox'
offer	"Congratulations - You're Our New Management Consultant! Dear Diana Brown, Congratulations! StrategyConsult Partners is excited to call you our new Management Consultant. We'll focus on wrapping up a few more formalities, including the successful completion of your background check and client reference verification, and aim to get you settled into your ne" 'Full-Time Employment Offer Dear Brandon Taylor, ArchitectureMax is offering to extend your current employment status from contractor to full-time employee, as of June 1st, 2024. If you choose to accept our offer, please review the terms and conditions of your new employment contract below: Position: You will be working as a S' 'Employment Offer - Product Manager Position Michael Chen 456 Innovation Drive, San Francisco, CA 94105 Re: Employment Offer Dear Michael: On behalf of ProductMax, Inc. (the "Company"), it is my pleasure to offer you employment with the Company in the role set forth below. The purpose of this letter is to summarize the initial terms of your em'
rejected	'Thanks for your time Thank you for your interest in the Software Engineer position at Lantana Consulting Group in Vermont, United States. Unfortunately, we will not be moving forward with your application, but we appreciate your time and interest in Lantana Consulting Group.\n\nRegards,\n\nLantana Consulting Group' "Thanks for your time Hello Name,\n\nThank you very much for your interest in our Software Engineer - React/Redux opening. We've had a chance to discuss your background and qualifications with the hiring manager and unfortunately, we have decided to pursue other candidates who appear to match our requirements more closely" "Thanks for your interest in Supernova Technology, Name Hi Name,\nThank you for your interest in Supernova Technology. After reviewing your background and experience, we’ve decided not to move forward with your application at this time.\n\nWe truly appreciate the time and effort you put into the process, and we hope you don't mind if we reach out in the fut"

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("Thanks for your time Thank you for applying to the Backend Developer position at YinzCam, Inc..
Unfortunately, YinzCam, Inc. has moved to the next step in their hiring process, and your application was not selected at this time.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	14	55.2121	288

Label	Training Sample Count
applied	40
interview	45
next-phase	35
not-job-related	55
not-job-status-update	41
offer	36
rejected	45

Training Hyperparameters

• batch_size: (16, 2)
• num_epochs: (1, 16)
• 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: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0002	1	0.3397	-
0.0106	50	0.2699	-
0.0212	100	0.2293	-
0.0319	150	0.1907	-
0.0425	200	0.1685	-
0.0531	250	0.1174	-
0.0637	300	0.078	-
0.0743	350	0.0524	-
0.0849	400	0.0319	-
0.0956	450	0.0113	-
0.1062	500	0.0073	-
0.1168	550	0.0051	-
0.1274	600	0.0038	-
0.1380	650	0.0029	-
0.1487	700	0.0023	-
0.1593	750	0.0021	-
0.1699	800	0.0017	-
0.1805	850	0.0017	-
0.1911	900	0.0015	-
0.2017	950	0.0012	-
0.2124	1000	0.0011	-
0.2230	1050	0.0011	-
0.2336	1100	0.001	-
0.2442	1150	0.001	-
0.2548	1200	0.0009	-
0.2654	1250	0.0008	-
0.2761	1300	0.0008	-
0.2867	1350	0.0007	-
0.2973	1400	0.0007	-
0.3079	1450	0.0006	-
0.3185	1500	0.0006	-
0.3292	1550	0.0006	-
0.3398	1600	0.0006	-
0.3504	1650	0.0006	-
0.3610	1700	0.0005	-
0.3716	1750	0.0005	-
0.3822	1800	0.0005	-
0.3929	1850	0.0005	-
0.4035	1900	0.0004	-
0.4141	1950	0.0004	-
0.4247	2000	0.0004	-
0.4353	2050	0.0004	-
0.4460	2100	0.0004	-
0.4566	2150	0.0004	-
0.4672	2200	0.0004	-
0.4778	2250	0.0004	-
0.4884	2300	0.0003	-
0.4990	2350	0.0003	-
0.5097	2400	0.0003	-
0.5203	2450	0.0003	-
0.5309	2500	0.0003	-
0.5415	2550	0.0003	-
0.5521	2600	0.0003	-
0.5628	2650	0.0003	-
0.5734	2700	0.0003	-
0.5840	2750	0.0002	-
0.5946	2800	0.0002	-
0.6052	2850	0.0003	-
0.6158	2900	0.0002	-
0.6265	2950	0.0002	-
0.6371	3000	0.0002	-
0.6477	3050	0.0002	-
0.6583	3100	0.0002	-
0.6689	3150	0.0002	-
0.6795	3200	0.0002	-
0.6902	3250	0.0002	-
0.7008	3300	0.0002	-
0.7114	3350	0.0002	-
0.7220	3400	0.0002	-
0.7326	3450	0.0002	-
0.7433	3500	0.0002	-
0.7539	3550	0.0002	-
0.7645	3600	0.0002	-
0.7751	3650	0.0002	-
0.7857	3700	0.0002	-
0.7963	3750	0.0002	-
0.8070	3800	0.0002	-
0.8176	3850	0.0002	-
0.8282	3900	0.0002	-
0.8388	3950	0.0002	-
0.8494	4000	0.0002	-
0.8601	4050	0.0002	-
0.8707	4100	0.0002	-
0.8813	4150	0.0002	-
0.8919	4200	0.0002	-
0.9025	4250	0.0002	-
0.9131	4300	0.0002	-
0.9238	4350	0.0002	-
0.9344	4400	0.0002	-
0.9450	4450	0.0001	-
0.9556	4500	0.0002	-
0.9662	4550	0.0001	-
0.9769	4600	0.0002	-
0.9875	4650	0.0001	-
0.9981	4700	0.0002	-

Framework Versions

• Python: 3.11.13
• SetFit: 1.1.3
• Sentence Transformers: 5.1.0
• Transformers: 4.56.1
• PyTorch: 2.2.2
• Datasets: 4.0.0
• Tokenizers: 0.22.0

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}
}