ShiWarai/CVC-Panda

SetFit with google/embeddinggemma-300M

This is a SetFit model that can be used for Text Classification. This SetFit model uses google/embeddinggemma-300M 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: google/embeddinggemma-300M
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 2048 tokens
• Number of Classes: 14 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
help	'помощь' 'помоги' 'помогите'
silence	'тишина' 'молчи' 'молчите'
bind	'привяжи робота' 'привяжи панду' 'привяжи робота 1'
unbind	'отвяжи робота' 'отвяжи панду' 'отвяжите робота'
report_command	'исправить команду' 'исправь команду' 'исправьте команду'
give_paw	'лапу' 'дай лапу' 'дать лапу'
stand_at_attention	'равняйсь' 'равняйся' 'равняться'
dismiss	'отставить' 'отставь' 'встать'
lie_down	'лежать' 'лечь' 'ложиться'
rotate	'кувыркнуться' 'кувыркнись' 'кувыркаться'
run	'бежать' 'беги' 'бегать'
stop_running	'остановиться' 'остановись' 'останавливаться'
reconnect_joystick	'подключить джойстик' 'подключи джойстик' 'подключать джойстик'
unknown	'привет' 'как дела' 'что происходит'

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("tmpb84tfylb/panda_commands")
# Run inference
preds = model("часто вращается")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	2.3808	7

Label	Training Sample Count
bind	55
dismiss	160
give_paw	104
help	22
lie_down	172
reconnect_joystick	135
report_command	50
rotate	137
run	106
silence	27
stand_at_attention	88
stop_running	135
unbind	37
unknown	479

Training Hyperparameters

• batch_size: (256, 256)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• body_learning_rate: (2e-05, 2e-05)
• head_learning_rate: 2e-05
• 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
0.0037	1	0.2375	-
0.1873	50	0.0728	-
0.3745	100	0.009	-
0.5618	150	0.005	-
0.7491	200	0.0038	-
0.9363	250	0.0028	-

Framework Versions

• Python: 3.11.14
• SetFit: 1.1.3
• Sentence Transformers: 5.2.2
• Transformers: 4.57.6
• PyTorch: 2.9.1+cu128
• Datasets: 4.5.0
• 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}
}