NaveenKumar96/intent-setfit-deberta-v1

SetFit

This is a SetFit model that can be used for Text Classification. 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
• 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
INFORMATION	'We lose a lot of time on the streets, easy laptop.' 'Okay, Lucid Gap 1.2, nice work. Right, eight laps to go once you cross the line.' "Well done, Daniel. That's the checkered flag. P7, buddy. P7 is a good start. Well defended at the end. That was great driving last few laps. Cheers, guys. Good start. Obviously, I'll learn a bit from today. We'll keep getting better, but not a bad first weekend. Congrats."
PROBLEM	'I think I have still some damage. Balance is quite a bit off. Understood. We do see from data.' 'So that floor damage looks like it picked up at turn 8, so right hand side.' 'Okay, so switch off the engine please Lewis.'
ORDER	'Okay, mode race and focus on turn two preparation, turn two and three for this lap.' 'Okay, that last lap, same pace as signs. And mode seven. Mode seven.' 'Yuki, we are rear left limited, so watch we spin and we can push in the high speed. Push in the high speed.'
WARNING	'Be careful, a lot of people in the pit lane, so obviously they will move, but watch yourself as well.' "The only concern I have, if the safety car comes out and we have to pit for that hard tyre and I can't get any temperature into it, we're in trouble. Yeah, copy Lewis, we're not concerned, we think everyone will be in the same boat, if not worse." "So, Max, for info, you've been given a 10 second penalty for forcing Lando off track at turn 4. So, head down. 10. That's quite impressive. That was a lot of whinging. A lot."
QUESTION	"Do you want to flap adjust Nico? Not really, but it's new sticky tires. Let's maybe try. Which way would you go? I would take off, yeah. I would try less, I guess, and see what it feels like. Okay, copy. Take like half percent down. Turn 10 feels pretty low in grip, like tailwind, I think. The rear is pretty unhappy there. Yes, on the tailwind into 10. Also, 1 and 6 and 7 are tailwind. Okay, so we've taken off 0.5. We'll get a feel. One more grid. Go to the grid after this. Close the radio rear, it's too loud." "That's it, mate. You are world champion. World champion. What a mate. I'm so proud. Lando, this is Zach from McLaren. Is this the world champion hotline? Yeah. You did it! You did it! Arthur! Woo! Thank you, guys. Oh my god. You made a kid's dream come true. Thank you so much. I love you guys. Thanks for everything. You deserve it. I love you, Mum. I love you, Dad. Thanks for everything. I'm not crying." "And Lando, do you think you can get past? Otherwise, what about plan B? Remember R switch, R switch. Yeah, if you've got the goblins, go for it."

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("Don't go too crazy on the brake warm up.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	27.2063	386

Label	Training Sample Count
INFORMATION	32
PROBLEM	32
ORDER	32
WARNING	32
QUESTION	32

Training Hyperparameters

• batch_size: (8, 8)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• 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
0.0013	1	0.1755	-
0.0625	50	0.3393	-
0.125	100	0.3649	-
0.1875	150	0.4159	-
0.25	200	0.3289	-
0.3125	250	0.3171	-
0.375	300	0.2952	-
0.4375	350	0.3004	-
0.5	400	0.3111	-
0.5625	450	0.3039	-
0.625	500	0.2702	-
0.6875	550	0.2891	-
0.75	600	0.2793	-
0.8125	650	0.2652	-
0.875	700	0.2761	-
0.9375	750	0.2603	-
1.0	800	0.2543	0.5

Framework Versions

• Python: 3.12.13
• SetFit: 1.1.3
• Sentence Transformers: 5.4.1
• Transformers: 5.0.0
• PyTorch: 2.10.0+cu128
• Datasets: 5.0.0
• Tokenizers: 0.22.2

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}
}