metadata

library_name: setfit
tags:
  - setfit
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
metrics:
  - accuracy
  - f1
  - precision
  - recall
widget:
  - text: |2+
        I'll pay $1,000 if anyone can find a published study that ChatGPT confirms merely attempts to refute the OPV AIDS theory without desperately resorting to a pathetic strawman.

  - text: >-
      my disappointment is immeasurable and my day is ruined. any idea if they
      will ever fix it or is it just permanent? i feel like just wow man just
      freaking wow
  - text: >-
      The stuff chatgpt gives is entirely too scripted *and* impractical, which
      is what I'm trying to avoid :/
  - text: >-
      my experience with product product and brand: it's amazing and not a bit
      scary. despite the articles about product's personality, my experience
      shows the opposite: it's useful, friendly, and truly amazing technology.
  - text: >-
      product is a massive energy hog. have a bunch of tabs open and your
      computer will come to a crawl. also, ad blocking is terrible on product
      company ads) because product apparently has a "whitelist" of ads that it
      refuses to be blocked. company is way better
pipeline_tag: text-classification
inference: true
base_model: BAAI/bge-small-en-v1.5
model-index:
  - name: SetFit with BAAI/bge-small-en-v1.5
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: Unknown
          type: unknown
          split: test
        metrics:
          - type: accuracy
            value: 0.5192307692307693
            name: Accuracy
          - type: f1
            value:
              - 0.2641509433962264
              - 0.1553398058252427
              - 0.6593406593406593
            name: F1
          - type: precision
            value:
              - 0.1590909090909091
              - 0.09090909090909091
              - 0.9375
            name: Precision
          - type: recall
            value:
              - 0.7777777777777778
              - 0.5333333333333333
              - 0.5084745762711864
            name: Recall

SetFit with BAAI/bge-small-en-v1.5

This is a SetFit model that can be used for Text Classification. This SetFit model uses BAAI/bge-small-en-v1.5 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: BAAI/bge-small-en-v1.5
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 tokens
Number of Classes: 3 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
peak	" I used Word on Microsoft 10 on my laptop to type up my manuscript, and when I uploaded it onto KDP, it was automatically formatted perfectly for Kindle E-book. I didn't need to make any adjustments (thankfully)." 'feeling myself getting obsessed with/addicted to ChatGPT and the entire generative AI universe and its evolution. \n\ndelightful to have another really big, seemingly biggest yet tech to go deep on and obsess over and think about implications of for the foreseeable future' '1/2 obsidian translate amazing plugin currently in beta. it can translate text in to multiple languages using multiple services. i just hooked it up to a free product translation account, and i am stunned by its accuracy. tft'
pit	"Looks like I got a new Microsoft 365 update last night. Now when I go to Options or Print, I crash. It's happening on multiple files. Probably other issues too, but haven't experimented much beyond that. Windows 11 and, obviously, the most up-to-date PPT. Fortunately I don't need PowerPoint right now - except to answer questions here - so I guess I'll just stick it out to see what happens before I do a repair/reinstall. Update: Quick repair didn't work. Full repair that I believe is a full reinstall didn't work." 'my disappointment is immeasurable and my day is ruined. any idea if they will ever fix it or is it just permanent? i feel like just wow man just freaking wow' 'between 100 pages of the packet devoted to some crumbly looking old house and the powerpoint about the importance of the military industrial complex, this meeting has me feeling hostile.'
neither	" Elevate your game with these mind-blowing ChatGPT prompts! \n\nWhether you're diving into knowledge, refining your skills, or making decisions, let be your guide to excellence. \n\nReady to unlock the power of AI? \n\n " "As an alternative you can always use Ask Sage ( Basically the gov version of ChatGPT and allowed to be used for CUI. It's what I use on NMCI and I've never had any problems!" " I'll pay $1,000 if anyone can find a published study that ChatGPT confirms merely attempts to refute the OPV AIDS theory without desperately resorting to a pathetic strawman.\n\n"

Evaluation

Metrics

Label	Accuracy	F1	Precision	Recall
all	0.5192	[0.2641509433962264, 0.1553398058252427, 0.6593406593406593]	[0.1590909090909091, 0.09090909090909091, 0.9375]	[0.7777777777777778, 0.5333333333333333, 0.5084745762711864]

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("tjmooney98/725_test_model")
# Run inference
preds = model("The stuff chatgpt gives is entirely too scripted *and* impractical, which is what I'm trying to avoid :/")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	18	38.0667	91

Label	Training Sample Count
pit	5
peak	5
neither	5

Training Hyperparameters

batch_size: (5, 5)
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
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0333	1	0.1809	-

Framework Versions

Python: 3.10.12
SetFit: 1.0.3
Sentence Transformers: 2.5.1
Transformers: 4.38.1
PyTorch: 2.1.0+cu121
Datasets: 2.18.0
Tokenizers: 0.15.2

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}
}