---
title: Llama-3-8b-rm-mixture
canonical_url: "https://www.modelscope.cn/models/AI-ModelScope/Llama-3-8b-rm-mixture"
md_url: "https://www.modelscope.cn/models/AI-ModelScope/Llama-3-8b-rm-mixture.md"
repository: AI-ModelScope/Llama-3-8b-rm-mixture
last_updated: 2025-02-14
model_type:
  - llama
architectures:
  - RewardModel
parameters: 7.5B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 905
stars: 0
---

# Llama-3-8b-rm-mixture

> Llama-3-8b-rm-mixture - AI-ModelScope 在 ModelScope 开源的模型。The Llama3-8b-based Reward Model was trained using OpenRLHF and a combination of datasets available at https://huggingface.co/datasets/OpenLLMAI/preferencedatasetmixture2andsafepku.

- **Repository**: AI-ModelScope/Llama-3-8b-rm-mixture
- **Parameters**: 7.5B
- **Downloads**: 905
- **Stars**: 0
- **Last updated**: 2025-02-14

Source: https://www.modelscope.cn/models/AI-ModelScope/Llama-3-8b-rm-mixture

---

The Llama3-8b-based Reward Model was trained using OpenRLHF and a combination of datasets available at https://huggingface.co/datasets/OpenLLMAI/preference_dataset_mixture2_and_safe_pku.

Base model: https://huggingface.co/OpenRLHF/Llama-3-8b-sft-mixture

```
Cosine Scheduler
Learning Rate: 9e-6
Warmup Ratio: 0.03
Batch Size: 256
Epoch: 1
```