80 lines
2.1 KiB
Markdown
80 lines
2.1 KiB
Markdown
---
|
|
license: apache-2.0
|
|
tags:
|
|
- merge
|
|
---
|
|
|
|
|
|
<center><h1 style="font-size: 45px">⭐ UPDATE ⭐</h1></center>
|
|
|
|
Use this instead:
|
|
|
|
https://hf.co/Weyaxi/OpenHermes-2.5-neural-chat-v3-3-Slerp
|
|
|
|
|
|
|
|
# OpenHermes-2.5-neural-chat-v3-2-Slerp
|
|
|
|
This is the model for OpenHermes-2.5-neural-chat-v3-2-Slerp. I used [mergekit](https://github.com/cg123/mergekit) to merge models.
|
|
|
|
# Prompt Templates
|
|
|
|
You can use these prompt templates, but I recommend using ChatML.
|
|
|
|
### ChatML [(OpenHermes-2.5-Mistral-7B)](https://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B):
|
|
|
|
```
|
|
<|im_start|>system
|
|
{system}<|im_end|>
|
|
<|im_start|>user
|
|
{user}<|im_end|>
|
|
<|im_start|>assistant
|
|
{asistant}<|im_end|>
|
|
```
|
|
|
|
### [neural-chat-7b-v3-2](https://huggingface.co/Intel/neural-chat-7b-v3-2):
|
|
|
|
```
|
|
### System:
|
|
{system}
|
|
### User:
|
|
{user}
|
|
### Assistant:
|
|
```
|
|
|
|
# Yaml Config to reproduce
|
|
|
|
```yaml
|
|
|
|
slices:
|
|
- sources:
|
|
- model: teknium/OpenHermes-2.5-Mistral-7B
|
|
layer_range: [0, 32]
|
|
- model: Intel/neural-chat-7b-v3-2
|
|
layer_range: [0, 32]
|
|
merge_method: slerp
|
|
base_model: mistralai/Mistral-7B-v0.1
|
|
parameters:
|
|
t:
|
|
- filter: self_attn
|
|
value: [0, 0.5, 0.3, 0.7, 1]
|
|
- filter: mlp
|
|
value: [1, 0.5, 0.7, 0.3, 0]
|
|
- value: 0.5 # fallback for rest of tensors
|
|
dtype: float16
|
|
|
|
```
|
|
|
|
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
|
|
|
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_PulsarAI__OpenHermes-2.5-neural-chat-v3-2-Slerp)
|
|
|
|
| Metric | Value |
|
|
|-----------------------|---------------------------|
|
|
| Avg. | 70.2 |
|
|
| ARC (25-shot) | 67.49 |
|
|
| HellaSwag (10-shot) | 85.42 |
|
|
| MMLU (5-shot) | 64.13 |
|
|
| TruthfulQA (0-shot) | 61.05 |
|
|
| Winogrande (5-shot) | 80.3 |
|
|
| GSM8K (5-shot) | 63.08 | |