metadata
datasets:
- stingning/ultrachat
UltraLM-65b
This is UltraLM-65b delta weights, a chat language model trained upon UltraChat
Model Details
Model Description
The model is fine-tuned based on LLaMA-65b with a multi-turn chat-format template as below
User: instruction 1
Assistant: response 1<eos_token>
User: instruction 2
Assistant: response 2<eos_token>
...
- License: UltraLM is based on LLaMA and should be used under LLaMA's model license.
- Finetuned from model: LLaMA-65b
- Finetuned on data: UltraChat
Model Sources
Uses
To use this model, you need to recover the full model from the delta weights and perform inference following the template below:
[Optional]User: system prompt
User: user input
Assistant:
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 58.99 |
| ARC (25-shot) | 67.06 |
| HellaSwag (10-shot) | 84.98 |
| MMLU (5-shot) | 63.48 |
| TruthfulQA (0-shot) | 53.51 |
| Winogrande (5-shot) | 81.14 |
| GSM8K (5-shot) | 32.75 |
| DROP (3-shot) | 30.0 |