aashish1904
commited on
Commit
•
4ca5ba4
1
Parent(s):
46689eb
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
---
|
3 |
+
|
4 |
+
base_model:
|
5 |
+
- nothingiisreal/L3.1-8B-Celeste-V1.5
|
6 |
+
- Sao10K/Llama-3.1-8B-Stheno-v3.4
|
7 |
+
- Sao10K/L3.1-8B-Niitama-v1.1
|
8 |
+
- arcee-ai/Llama-3.1-SuperNova-Lite
|
9 |
+
- akjindal53244/Llama-3.1-Storm-8B
|
10 |
+
- arcee-ai/Llama-Spark
|
11 |
+
- grimjim/Llama-3-Instruct-abliteration-LoRA-8B
|
12 |
+
- crestf411/sunfall-peft
|
13 |
+
- v000000/L3.1-Celestial-Stone-2x8B
|
14 |
+
library_name: transformers
|
15 |
+
tags:
|
16 |
+
- merge
|
17 |
+
- llama
|
18 |
+
- mixtral
|
19 |
+
- dpo
|
20 |
+
|
21 |
+
---
|
22 |
+
|
23 |
+
[![QuantFactory Banner](https://lh7-rt.googleusercontent.com/docsz/AD_4nXeiuCm7c8lEwEJuRey9kiVZsRn2W-b4pWlu3-X534V3YmVuVc2ZL-NXg2RkzSOOS2JXGHutDuyyNAUtdJI65jGTo8jT9Y99tMi4H4MqL44Uc5QKG77B0d6-JfIkZHFaUA71-RtjyYZWVIhqsNZcx8-OMaA?key=xt3VSDoCbmTY7o-cwwOFwQ)](https://hf.co/QuantFactory)
|
24 |
+
|
25 |
+
|
26 |
+
# QuantFactory/L3.1-Celestial-Stone-2x8B-DPO-GGUF
|
27 |
+
This is quantized version of [v000000/L3.1-Celestial-Stone-2x8B-DPO](https://huggingface.co/v000000/L3.1-Celestial-Stone-2x8B-DPO) created using llama.cpp
|
28 |
+
|
29 |
+
# Original Model Card
|
30 |
+
|
31 |
+
|
32 |
+
> [!WARNING]
|
33 |
+
> **Sampler:**<br>
|
34 |
+
> Likes a low temperature due to the MoE architecture. I use 0.3 personally.
|
35 |
+
|
36 |
+
# Llama-3.1-Celestial-Stone-2x8B-DPO (BF16)
|
37 |
+
|
38 |
+
* *DPO Trained, Mixture of Experts (14B).*
|
39 |
+
|
40 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64f74b6e6389380c77562762/lyRa7z5maTqAaa43sxC2J.png)
|
41 |
+
|
42 |
+
* <b>Direct Preference Optimization run</b>
|
43 |
+
|
44 |
+
----> [Q6_K](https://huggingface.co/v000000/L3.1-Celestial-Stone-2x8B-DPO-Q6_K-GGUF)
|
45 |
+
|
46 |
+
---------------------------------------------------------------------------------
|
47 |
+
|
48 |
+
[L3.1-Celestial-Stone-2x8B](https://huggingface.co/v000000/L3.1-Celestial-Stone-2x8B) Finetuned on Nvidia A100.
|
49 |
+
|
50 |
+
0.5 Epoch completed of dataset [jondurbin/gutenberg-dpo-v0.1](https://huggingface.co/datasets/jondurbin/gutenberg-dpo-v0.1) with learning_rate=8e-6
|
51 |
+
|
52 |
+
Result seems pretty good. More compliant and verbose, less sloppy and safety aligned.
|
53 |
+
|
54 |
+
------------------------------------------------------------------------------
|
55 |
+
|
56 |
+
*The first expert* is Instruct 405B distillation/RP vector merge <b>(Supernova-Lite, Niitama1.1, Storm)</b>
|
57 |
+
|
58 |
+
*The second expert* is ERP/Reddit data merge <b>(Celeste1.5, Stheno3.4, Storm)</b>
|
59 |
+
|
60 |
+
-------------------------------------------------------------------------------
|
61 |
+
|
62 |
+
*The base model* is <b>Sao10k/L3.1-Stheno-3.4</b> with the <b>Sunfall LoRa 0.6.1</b> to make it understand SillyTavern prompts and storywriting better.
|
63 |
+
|
64 |
+
-------------------------------------------------------------------------------
|
65 |
+
|
66 |
+
*Resultant merge finetuned* on [jondurbin/gutenberg-dpo-v0.1](https://huggingface.co/datasets/jondurbin/gutenberg-dpo-v0.1).
|
67 |
+
|
68 |
+
# Prompt Template:
|
69 |
+
```bash
|
70 |
+
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
|
71 |
+
|
72 |
+
{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
|
73 |
+
|
74 |
+
{input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
|
75 |
+
|
76 |
+
{output}<|eot_id|>
|
77 |
+
|
78 |
+
```
|