Instructions to use SOTAagi2030/PocketIntent-Pilot with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SOTAagi2030/PocketIntent-Pilot with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="SOTAagi2030/PocketIntent-Pilot")# Load model directly from transformers import AutoTokenizer, AutoModelForSequenceClassification tokenizer = AutoTokenizer.from_pretrained("SOTAagi2030/PocketIntent-Pilot") model = AutoModelForSequenceClassification.from_pretrained("SOTAagi2030/PocketIntent-Pilot", device_map="auto") - Notebooks
- Google Colab
- Kaggle
PocketIntent Pilot
PocketIntent is an offline classifier for triaging mobile support messages.
Release decision
Selection rule: choose the lowest validation latency among runs with macro F1 >= 0.900.
- Selected run:
run_sable - Validation macro F1: 0.906
- Validation latency: 11.2 ms
- Supported intents: billing, connectivity, device_setup, account_access
Pilot evidence
- Device latency measurements:
reports/latency_by_device.csv - Confusion matrix:
reports/confusion_matrix.png
This pilot artifact is intended for offline evaluation only.
- Downloads last month
- 25