metadata

library_name: setfit
tags:
  - setfit
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
base_model: mental/mental-bert-base-uncased
metrics:
  - accuracy
widget:
  - text: I let myself go, I make no effort to eat, sleep or take care of myself.
  - text: There's no structure in my life, and that makes me even sicker.
  - text: >-
      I'm drifting away from my friends, my family, games that I couldn't
      possibly know anything about.
  - text: >-
      My grandmother's homemade pasta recipe is the best, nothing else compares
      to it.
  - text: >-
      It's frustrating to realize I've made yet another impulsive choice that
      sets me back instead of moving forward.
pipeline_tag: text-classification
inference: true
model-index:
  - name: SetFit with mental/mental-bert-base-uncased
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: Unknown
          type: unknown
          split: test
        metrics:
          - type: accuracy
            value: 0.8275862068965517
            name: Accuracy

SetFit with mental/mental-bert-base-uncased

This is a SetFit model that can be used for Text Classification. This SetFit model uses mental/mental-bert-base-uncased 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: mental/mental-bert-base-uncased
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 tokens
Number of Classes: 8 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
Presence of a loved one	"Living alone feels like a double-edged sword, on one hand, I have my privacy, but on the other hand, there's no one around to support me when I'm feeling like shit." 'I have no one to talk to or no one who understands me.' 'I wish I had a shoulder to lean on when everything is falling apart.'
Previous attempt	'Shit got real after my last attempt, i woke up surrounded by a bunch of medical equipment, wondering what the fuck happened.' "Looking back at my attempts, I can't deny there's a part of me that's glad they didn't work, but there's also this nagging voice wondering if I missed my chance." "In the past, I tried to end it all, but I'm in the same situation now."
Ability to take care of oneself	"I've let go of everything that used to matter to me, my hobbies, my friends, everything seems uninteresting." "It's hard to maintain a routine or assume my responsibilities, I've completely dropped out." "I'm so disoriented, I forget important things and it's making my situation worse."
Ability to hope for change	"I'm just so tired of feeling stuck, every day feels the same, and I can't see any way out, I need help finding hope for a change in my life.," 'I feel so isolated in all of this, like nobody would ever understand what I’m going through.' "I don't know what to do anymore, it feels like my reason to live has been completely drained away, i'm desperately searching for any glimmer of hope to hold onto.,"
Other	"My garden has been thriving this summer, it's so rewarding to see everything in bloom." "I've been trying to be more sustainable, I even started composting my food scraps." "I've been learning to play the piano lately, it's so fulfilling to see my progress."
Suicidal planning	"I don't want to be here anymore, I don't want to endure this, I'm seriously thinking about killing myself." "I've had thoughts of ending my life." "Sometimes, I imagine myself in this life, and I don't see the point of it all."
Ability to control oneself	"I wish I could put my brain on pause, just so I don't act on impulse all the time." 'I try to focus, but my thoughts are too fast and I act before considering the consequences.' "I can't stop myself from falling apart, no matter how hard I try."
Consumption	"Drinking has become my go-to escape from this overwhelming situation, It's like I can't deal with it sober anymore." "The more I struggle with this distressing situation, the more I find solace in drinking, It's becoming a dangerous habit that's tough to break." 'Lately, when shit hits the fan, I find myself reaching for the bottle'

Evaluation

Metrics

Label	Accuracy
all	0.8276

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("richie-ghost/setfit-mental-bert-base-uncased-Suicidal-Topic-Check")
# Run inference
preds = model("There's no structure in my life, and that makes me even sicker.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	7	18.3582	40

Label	Training Sample Count
Suicidal planning	9
Previous attempt	11
Presence of a loved one	8
Other	9
Consumption	6
Ability to take care of oneself	8
Ability to hope for change	7
Ability to control oneself	9

Training Hyperparameters

batch_size: (16, 16)
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: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0041	1	0.3127	-
0.2041	50	0.1378	-
0.4082	100	0.0519	-
0.6122	150	0.0043	-
0.8163	200	0.0014	-
1.0	245	-	0.0717
1.0204	250	0.0008	-
1.2245	300	0.0006	-
1.4286	350	0.0006	-
1.6327	400	0.0003	-
1.8367	450	0.0005	-
2.0	490	-	0.0693
2.0408	500	0.0005	-
2.2449	550	0.0006	-
2.4490	600	0.0005	-
2.6531	650	0.0003	-
2.8571	700	0.0003	-
3.0	735	-	0.0698
0.0041	1	0.0003	-
0.2041	50	0.0006	-
0.4082	100	0.0004	-
0.6122	150	0.001	-
0.8163	200	0.0002	-
1.0	245	-	0.0633
1.0204	250	0.0002	-
1.2245	300	0.0	-
1.4286	350	0.0001	-
1.6327	400	0.0001	-
1.8367	450	0.0001	-
2.0	490	-	0.0598
2.0408	500	0.0001	-
2.2449	550	0.0001	-
2.4490	600	0.0001	-
2.6531	650	0.0001	-
2.8571	700	0.0001	-
3.0	735	-	0.0585
3.0612	750	0.0001	-
3.2653	800	0.0001	-
3.4694	850	0.0001	-
3.6735	900	0.0001	-
3.8776	950	0.0	-
4.0	980	-	0.0582
4.0816	1000	0.0001	-
4.2857	1050	0.0	-
4.4898	1100	0.0	-
4.6939	1150	0.0	-
4.8980	1200	0.0	-
5.0	1225	-	0.0583
5.1020	1250	0.0	-
5.3061	1300	0.0	-
5.5102	1350	0.0	-
5.7143	1400	0.0	-
5.9184	1450	0.0	-
6.0	1470	-	0.0561
0.0041	1	0.0	-
0.2041	50	0.0	-
0.4082	100	0.0001	-
0.6122	150	0.0002	-
0.8163	200	0.0002	-
1.0	245	-	0.0699
1.0204	250	0.0001	-
1.2245	300	0.0001	-
1.4286	350	0.0	-
1.6327	400	0.0	-
1.8367	450	0.0	-
2.0	490	-	0.0653
2.0408	500	0.0001	-
2.2449	550	0.0	-
2.4490	600	0.0	-
2.6531	650	0.0001	-
2.8571	700	0.0001	-
3.0	735	-	0.0651
3.0612	750	0.0	-
3.2653	800	0.0	-
3.4694	850	0.0	-
3.6735	900	0.0	-
3.8776	950	0.0001	-
4.0	980	-	0.0634
4.0816	1000	0.0	-
4.2857	1050	0.0	-
4.4898	1100	0.0	-
4.6939	1150	0.0	-
4.8980	1200	0.0	-
5.0	1225	-	0.0654
5.1020	1250	0.0	-
5.3061	1300	0.0	-
5.5102	1350	0.0	-
5.7143	1400	0.0	-
5.9184	1450	0.0	-
6.0	1470	-	0.0627
6.1224	1500	0.0	-
6.3265	1550	0.0	-
6.5306	1600	0.0	-
6.7347	1650	0.0	-
6.9388	1700	0.0	-
7.0	1715	-	0.0648
7.1429	1750	0.0	-
7.3469	1800	0.0	-
7.5510	1850	0.0	-
7.7551	1900	0.0	-
7.9592	1950	0.0	-
8.0	1960	-	0.0636
8.1633	2000	0.0	-
8.3673	2050	0.0	-
8.5714	2100	0.0	-
8.7755	2150	0.0	-
8.9796	2200	0.0	-
9.0	2205	-	0.0648
9.1837	2250	0.0	-
9.3878	2300	0.0	-
9.5918	2350	0.0	-
9.7959	2400	0.0	-
10.0	2450	0.0	0.0643

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
SetFit: 1.0.3
Sentence Transformers: 2.7.0
Transformers: 4.40.0
PyTorch: 2.2.1+cu121
Datasets: 2.19.0
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}
}