Update README.md
Browse files
README.md
CHANGED
|
@@ -113,53 +113,4 @@ model-index:
|
|
| 113 |
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=antiven0m/finch
|
| 114 |
name: Open LLM Leaderboard
|
| 115 |
---
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
# **Finch 7b Merge**
|
| 120 |
-
## A SLERP merge of my two current fav 7B models
|
| 121 |
-
|
| 122 |
-
[macadeliccc/WestLake-7B-v2-laser-truthy-dpo](https://huggingface.co/macadeliccc/WestLake-7B-v2-laser-truthy-dpo) & [SanjiWatsuki/Kunoichi-DPO-v2-7B](https://huggingface.co/SanjiWatsuki/Kunoichi-DPO-v2-7B)
|
| 123 |
-
|
| 124 |
-
- [6bpw EXL2 Quant](https://huggingface.co/antiven0m/finch-6bpw-exl2)
|
| 125 |
-
- [GGUF Quants](https://huggingface.co/antiven0m/finch-gguf)
|
| 126 |
-
|
| 127 |
-
## Settings
|
| 128 |
-
I reccomend using the **ChatML** format. As for samplers, I reccomend the following:
|
| 129 |
-
```yml
|
| 130 |
-
Temperature: 1.2
|
| 131 |
-
Min P: 0.2
|
| 132 |
-
Smoothing Factor: 0.2
|
| 133 |
-
```
|
| 134 |
-
## Mergekit Config
|
| 135 |
-
```yaml
|
| 136 |
-
base_model: macadeliccc/WestLake-7B-v2-laser-truthy-dpo
|
| 137 |
-
dtype: float16
|
| 138 |
-
merge_method: slerp
|
| 139 |
-
parameters:
|
| 140 |
-
t:
|
| 141 |
-
- filter: self_attn
|
| 142 |
-
value: [0.0, 0.5, 0.3, 0.7, 1.0]
|
| 143 |
-
- filter: mlp
|
| 144 |
-
value: [1.0, 0.5, 0.7, 0.3, 0.0]
|
| 145 |
-
- value: 0.5
|
| 146 |
-
slices:
|
| 147 |
-
- sources:
|
| 148 |
-
- layer_range: [0, 32]
|
| 149 |
-
model: macadeliccc/WestLake-7B-v2-laser-truthy-dpo
|
| 150 |
-
- layer_range: [0, 32]
|
| 151 |
-
model: SanjiWatsuki/Kunoichi-DPO-v2-7B
|
| 152 |
-
```
|
| 153 |
-
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
| 154 |
-
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_antiven0m__finch)
|
| 155 |
-
|
| 156 |
-
| Metric |Value|
|
| 157 |
-
|---------------------------------|----:|
|
| 158 |
-
|Avg. |73.78|
|
| 159 |
-
|AI2 Reasoning Challenge (25-Shot)|71.59|
|
| 160 |
-
|HellaSwag (10-Shot) |87.87|
|
| 161 |
-
|MMLU (5-Shot) |64.81|
|
| 162 |
-
|TruthfulQA (0-shot) |67.96|
|
| 163 |
-
|Winogrande (5-shot) |84.14|
|
| 164 |
-
|GSM8k (5-shot) |66.34|
|
| 165 |
-
|
|
|
|
| 113 |
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=antiven0m/finch
|
| 114 |
name: Open LLM Leaderboard
|
| 115 |
---
|
| 116 |
+
<head> <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.0.0-beta3/css/all.min.css"> </head> <style> body { font-family: "Helvetica Neue", Arial, sans-serif; background: radial-gradient(circle, #ffb347, #ffa92d, #ff9f14, #ff9500, #f08b00); color: #fff; line-height: 1.6; } .container { max-width: 800px; margin: 0 auto; padding: 40px; background-color: rgba(255, 255, 255, 0.1); border-radius: 10px; box-shadow: 0 0 20px rgba(0, 0, 0, 0.2); backdrop-filter: blur(10px); } .header { text-align: center; margin-bottom: 40px; } .title { font-size: 48px; font-weight: bold; text-transform: uppercase; letter-spacing: 2px; color: #fff; text-shadow: 2px 2px 4px rgba(0, 0, 0, 0.5); margin-bottom: 10px; } .subtitle { font-size: 24px; font-style: italic; color: #e6f7ff; text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.3); margin-bottom: 20px; } .gif { text-align: center; margin-bottom: 40px; } .gif img { max-width: 100%; height: auto; border-radius: 10px; box-shadow: 0 0 20px rgba(0, 0, 0, 0.3); } .info-section { margin-bottom: 40px; } .section-title { font-size: 32px; font-weight: bold; color: #e6f7ff; text-shadow: 2px 2px 4px rgba(0, 0, 0, 0.5); margin-bottom: 20px; position: relative; padding-left: 30px; } .section-title::before { content: ""; position: absolute; left: 0; top: 50%; transform: translateY(-50%); width: 20px; height: 20px; background-color: #e6f7ff; border-radius: 50%; box-shadow: 0 0 10px rgba(0, 0, 0, 0.3); } .info-item { background-color: rgba(255, 255, 255, 0.1); padding: 20px; border-radius: 10px; box-shadow: 0 0 10px rgba(0, 0, 0, 0.2); margin-bottom: 20px; } .info-item h3 { font-size: 24px; font-weight: bold; color: #e6f7ff; text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.3); margin-bottom: 10px; } .info-item p { font-size: 18px; color: #fff; text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.3); line-height: 1.4; } .info-item pre { background-color: rgba(0, 0, 0, 0.2); padding: 20px; border-radius: 10px; font-family: monospace; font-size: 16px; color: #fff; text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.3); overflow-x: auto; } .info-item a { color: #e6f7ff; text-decoration: none; border-bottom: 1px dashed #e6f7ff; transition: border-bottom 0.3s ease; } .info-item a:hover { border-bottom: 1px solid #e6f7ff; } .info-item table { width: 100%; border-collapse: collapse; box-shadow: 0 0 10px rgba(0, 0, 0, 0.2); } .info-item th, .info-item td { padding: 10px; text-align: left; border: 1px solid rgba(255, 255, 255, 0.2); } .info-item th { background-color: rgba(0, 0, 0, 0.2); font-weight: bold; color: #fff; text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.3); } .info-item td { color: #e6f7ff; text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.3); } </style> <div class="container"> <div class="header"> <h1 class="title">Finch 7B Merge</h1> <p class="subtitle">A SLERP merge of two powerful 7B language models</p> </div> <div class="gif"> <img src="https://i.imgur.com/Da14544.gif" alt="Finch GIF"> </div> <div class="info-section"> <h2 class="section-title">Description</h2> <div class="info-item"> <p>Finch is a 7B language model created by merging <a href="https://huggingface.co/macadeliccc/WestLake-7B-v2-laser-truthy-dpo">macadeliccc/WestLake-7B-v2-laser-truthy-dpo</a> and <a href="https://huggingface.co/SanjiWatsuki/Kunoichi-DPO-v2-7B">SanjiWatsuki/Kunoichi-DPO-v2-7B</a> using the SLERP method.</p> </div> </div> <div class="info-section"> <h2 class="section-title">Quantized Models</h2> <div class="info-item"> <p>Quantized versions of Finch are available:</p> <ul> <li><a href="https://huggingface.co/antiven0m/finch-6bpw-exl2">6bpw EXL2 Quant</a></li> <li><a href="https://huggingface.co/antiven0m/finch-gguf">GGUF Quants</a></li> </ul> </div> </div> <div class="info-section"> <h2 class="section-title">Recommended Settings</h2> <div class="info-item"> <p>For best results, use the <b>ChatML</b> format with the following sampler settings:</p> <pre>Temperature: 1.2 Min P: 0.2 Smoothing Factor: 0.2</pre> </div> </div> <div class="info-section"> <h2 class="section-title">Mergekit Configuration</h2> <div class="info-item"> <pre>base_model: macadeliccc/WestLake-7B-v2-laser-truthy-dpo dtype: float16 merge_method: slerp parameters: t: - filter: self_attn value: [0.0, 0.5, 0.3, 0.7, 1.0] - filter: mlp value: [1.0, 0.5, 0.7, 0.3, 0.0] - value: 0.5 slices: - sources: - layer_range: [0, 32] model: macadeliccc/WestLake-7B-v2-laser-truthy-dpo - layer_range: [0, 32] model: SanjiWatsuki/Kunoichi-DPO-v2-7B</pre> </div> </div> <div class="info-section"> <h2 class="section-title">Evaluation Results</h2> <div class="info-item"> <p>Finch's performance on the Open LLM Leaderboard:</p> <table> <tr><th>Metric</th><th>Value</th></tr> <tr><td>Avg.</td><td>73.78</td></tr> <tr><td>AI2 Reasoning Challenge (25-Shot)</td><td>71.59</td></tr> <tr><td>HellaSwag (10-Shot)</td><td>87.87</td></tr> <tr><td>MMLU (5-Shot)</td><td>64.81</td></tr> <tr><td>TruthfulQA (0-shot)</td><td>67.96</td></tr> <tr><td>Winogrande (5-shot)</td><td>84.14</td></tr> <tr><td>GSM8k (5-shot)</td><td>66.34</td></tr> </table> <p>Detailed results: <a href="https://huggingface.co/datasets/open-llm-leaderboard/details_antiven0m__finch">https://huggingface.co/datasets/open-llm-leaderboard/details_antiven0m__finch</a></p> </div> </div> </div>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|