nghodki/setfit-sre-task-classifier

SetFit with sentence-transformers/paraphrase-mpnet-base-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-mpnet-base-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/paraphrase-mpnet-base-v2
• 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
NONE	'How do I learn to play the guitar?' "What's the longest river in the world?" 'How do I overcome procrastination?'
KUBIE	'What logs should I check to identify container crashes in the qa-soc-svcs namespace?' 'Can you suggest ways to troubleshoot an image pull error in the "kube-public" namespace?' "I'm encountering errors with a pod in the sandbox-6 namespace. Any suggestions on how to debug it?"
aws_iam	'Show me the IAM role details including attached policies.' 'Show me the IAM roles that have the "admin" prefix.' 'How can I get detailed information about a particular IAM role?'
DOC	'How to access ArgoCD on Production?' 'How to run terraform in CDO?' 'How to push images to dockerhub.cisco.com?'
access_management	'Access to prod-aws infrastructure is required urgently for a deployment.' 'Could you provide me access to the dev-aws resources?' 'I require access to the prod-sagemaker instance for machine learning experiments.'

Evaluation

Metrics

Label	Accuracy
all	0.9962

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("How can I reduce stress?")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	8.5408	17

Label	Training Sample Count
aws_iam	20
access_management	20
DOC	18
KUBIE	20
NONE	20

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (4, 4)
• 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
0.0021	1	0.2675	-
0.1042	50	0.1143	-
0.2083	100	0.0578	-
0.3125	150	0.0028	-
0.4167	200	0.0032	-
0.5208	250	0.0007	-
0.625	300	0.0006	-
0.7292	350	0.0004	-
0.8333	400	0.0005	-
0.9375	450	0.0006	-
1.0	480	-	0.0027
1.0417	500	0.0004	-
1.1458	550	0.0002	-
1.25	600	0.0003	-
1.3542	650	0.0002	-
1.4583	700	0.0002	-
1.5625	750	0.0002	-
1.6667	800	0.0002	-
1.7708	850	0.0002	-
1.875	900	0.0002	-
1.9792	950	0.0001	-
2.0	960	-	0.0032
2.0833	1000	0.0001	-
2.1875	1050	0.0002	-
2.2917	1100	0.0001	-
2.3958	1150	0.0002	-
2.5	1200	0.0002	-
2.6042	1250	0.0001	-
2.7083	1300	0.0002	-
2.8125	1350	0.0001	-
2.9167	1400	0.0001	-
3.0	1440	-	0.004
3.0208	1450	0.0001	-
3.125	1500	0.0001	-
3.2292	1550	0.0002	-
3.3333	1600	0.0002	-
3.4375	1650	0.0001	-
3.5417	1700	0.0002	-
3.6458	1750	0.0001	-
3.75	1800	0.0001	-
3.8542	1850	0.0001	-
3.9583	1900	0.0002	-
4.0	1920	-	0.0037

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.9.6
• SetFit: 1.0.3
• Sentence Transformers: 2.7.0
• Transformers: 4.40.1
• PyTorch: 2.1.2
• 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}
}