Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: llama2
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
### Daddy Dave's stamp of approval 👍
|
| 8 |
+
|
| 9 |
+
4-bit GPTQ quants of the writer version of [Sao10K](https://huggingface.co/Sao10K)'s fantastic [SthenoWriter model (Stheno model collection link)](https://huggingface.co/collections/Sao10K/stheno-6536a20823c9d18c09288fb1) model.
|
| 10 |
+
|
| 11 |
+
The main branch contains 4-bit groupsize of 128.
|
| 12 |
+
|
| 13 |
+
The other branches contain groupsizes of 64 and 32 with act order.
|
| 14 |
+
|
| 15 |
+
Original card below!
|
| 16 |
+
|
| 17 |
+
***
|
| 18 |
+
|
| 19 |
+
<img src="https://c4.wallpaperflare.com/wallpaper/309/535/658/anime-anime-girls-fate-series-fate-grand-order-stheno-fate-grand-order-hd-wallpaper-preview.jpg" style="width: 70%; min-width: 300px; display: block; margin: auto;">
|
| 20 |
+
|
| 21 |
+
A Stheno-1.8 Variant focused on writing.
|
| 22 |
+
|
| 23 |
+
Stheno-1.8 + Storywriter, mixed with Holodeck + Spring Dragon qLoRA. End Result is mixed with One More Experimental Literature-based LoRA.
|
| 24 |
+
|
| 25 |
+
Re-Reviewed... it's not bad, honestly.
|
| 26 |
+
|
| 27 |
+
Support me [here](https://ko-fi.com/sao10k) :)
|
| 28 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
| 29 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Sao10K__SthenoWriter-L2-13B)
|
| 30 |
+
|
| 31 |
+
| Metric | Value |
|
| 32 |
+
|-----------------------|---------------------------|
|
| 33 |
+
| Avg. | 48.35 |
|
| 34 |
+
| ARC (25-shot) | 62.29 |
|
| 35 |
+
| HellaSwag (10-shot) | 83.28 |
|
| 36 |
+
| MMLU (5-shot) | 56.14 |
|
| 37 |
+
| TruthfulQA (0-shot) | 44.72 |
|
| 38 |
+
| Winogrande (5-shot) | 74.35 |
|
| 39 |
+
| GSM8K (5-shot) | 11.22 |
|
| 40 |
+
| DROP (3-shot) | 6.48 |
|