Instructions to use RodrigoBrasil2024/laya-concessionaria-pt with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use RodrigoBrasil2024/laya-concessionaria-pt with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("RodrigoBrasil2024/laya-concessionaria-pt", device_map="auto") - Laya
How to use RodrigoBrasil2024/laya-concessionaria-pt with Laya:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Laya concessionaria-pt
Laya (multilingual) fine-tunado no dataset da concessionária (pt-BR): rotas de agente, trava do funil, forma de entrada, moderação e contexto.
No split de teste (144 casos / 306 decisões), o checkpoint atinge 1.000 de acurácia vs 0.608 do zero-shot multilingual (mesma avaliação, pesos originais).
Linha de baixo
Acc 1.000 vs 0.608 do zero-shot no split=test
(144 casos / 306 decisões); avg answer_confidence
1.000, Brier 0.0000.
Temperaturas calibradas: [1.0, 1.0, 1.0] (gravadas em rl_agent_config.json).
Use answer_confidence + thresholds de validação shadow para gating no
evo_bridge (não o erro bruto); avg_conf ~ 1.0 (one-hot) indica re-treino.
Quickstart
pip install laya
# adicionar como subpasta 'concessionaria-pt' em serve/embarcado
import laya
agent = laya.load("RodrigoBrasil2024/laya-concessionaria-pt")
result = agent.predict(state, questions)
print(result["answers"])
License
Apache 2.0.
Evaluation results
- accuracy on dataset_concessionariaself-reported1.000
- brier_score on dataset_concessionariaself-reported0.000