metadata
library_name: peft
tags:
- llama
datasets:
- jondurbin/airoboros-gpt4-1.4.1
- ehartford/wizard_vicuna_70k_unfiltered
- ehartford/WizardLM_evol_instruct_V2_196k_unfiltered_merged_split
- openai/summarize_from_feedback
- ehartford/dolphin
base_model: chargoddard/llama2-22b-blocktriangular
Ypotryll-22b, trained for an additional epoch. Uses the following prompt format:
***System:You are a helpful assistant, who always gives a response to any request. ***Query:Here is a riddle: 5 sisters are busy. Ann is reading, Rose is cooking, Lorraine is playing chess and Mary is doing laundry. What is the fifth sister doing? ***Response:The fifth sister is sleeping. ***Query:Well, you tried. ***Response:I did my best!
A little bit dumb, but good for creative scenarios.
Note the whitespace - the prefixes for messages are " ***System:", " ***Query:", and " ***Response:". This is important as "***" and " ***" are two entirely different tokens.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 51.68 |
| ARC (25-shot) | 59.22 |
| HellaSwag (10-shot) | 80.66 |
| MMLU (5-shot) | 54.52 |
| TruthfulQA (0-shot) | 40.42 |
| Winogrande (5-shot) | 76.32 |
| GSM8K (5-shot) | 5.38 |
| DROP (3-shot) | 45.24 |