PEFT
Safetensors
English
lora
adapter
agent-safety
trajectory-monitoring
stepshield
conformal-prediction
Instructions to use DoronP/stepshield-nemotron-mini-4b-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use DoronP/stepshield-nemotron-mini-4b-lora with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("nvidia/Nemotron-Mini-4B-Instruct") model = PeftModel.from_pretrained(base_model, "DoronP/stepshield-nemotron-mini-4b-lora") - Notebooks
- Google Colab
- Kaggle
File size: 892 Bytes
9daa81e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 | NOTICE
This repository distributes a LoRA adapter (adapter_model.safetensors,
adapter_config.json) — a Derivative Model of the NVIDIA Model
"Nemotron-Mini-4B-Instruct" (c) NVIDIA Corporation, distributed under and
accompanied by the NVIDIA Community Model License (see
nvidia-community-model-license-aug2024.pdf, as required by Section 1.2 of
that Agreement).
Modification notice (Section 2.2.1): the distributed weights are LoRA
adapter deltas trained by Doron Podoleanu on the StepShield benchmark's
train split. They modify the behavior of the base NVIDIA Model and are not
original NVIDIA weights. NVIDIA does not endorse this work; no trademark
license is granted or implied.
The adapter weights are licensed under the NVIDIA Community Model License,
not an open-source license. Production use is subject to the license's
NVIDIA NIM / AI Enterprise conditions (see the model card).
|