Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,61 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- AuriAetherwiing/Allura
|
| 5 |
+
- anthracite-org/kalo-opus-instruct-22k-no-refusal
|
| 6 |
+
- Nopm/Opus_WritingStruct
|
| 7 |
+
- Gryphe/Sonnet3.5-SlimOrcaDedupCleaned
|
| 8 |
+
- Gryphe/Sonnet3.5-Charcard-Roleplay
|
| 9 |
+
- Gryphe/ChatGPT-4o-Writing-Prompts
|
| 10 |
+
- AuriAetherwiing/shortstories_synthlabels
|
| 11 |
+
- Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned
|
| 12 |
+
- Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned
|
| 13 |
+
base_model:
|
| 14 |
+
- Qwen/Qwen2.5-7B
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
**EVA Qwen2.5 7B**
|
| 18 |
+
|
| 19 |
+
<p>
|
| 20 |
+
A RP/storywriting specialist model, full-parameter finetune of Qwen2.5-7B on mixture of synthetic and natural data.<br>
|
| 21 |
+
A continuation of nothingiisreal's Celeste 1.x series, made to improve stability and versatility, without losing unique, diverse writing style of Celeste.<br>
|
| 22 |
+
</p>
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
<p>
|
| 26 |
+
Prompt format is ChatML.<br>
|
| 27 |
+
<h3>Recommended sampler values:</h3>
|
| 28 |
+
|
| 29 |
+
- Temperature: 0.87
|
| 30 |
+
- Top-P: 0.81
|
| 31 |
+
- Repetition Penalty: 1.05
|
| 32 |
+
|
| 33 |
+
<h3>Recommended SillyTavern presets (via CalamitousFelicitousness):</h3>
|
| 34 |
+
|
| 35 |
+
- [Context](https://huggingface.co/EVA-UNIT-01/EVA-Yi-1.5-9B-32K-V1/blob/main/%5BChatML%5D%20Roleplay-v1.9%20Context.json)
|
| 36 |
+
- [Instruct and System Prompt](https://huggingface.co/EVA-UNIT-01/EVA-Yi-1.5-9B-32K-V1/blob/main/%5BChatML%5D%20Roleplay-v1.9%20Instruct.json)
|
| 37 |
+
</p>
|
| 38 |
+
|
| 39 |
+
<p>
|
| 40 |
+
<br>
|
| 41 |
+
<h3>
|
| 42 |
+
Training data:
|
| 43 |
+
</h3>
|
| 44 |
+
<ul>
|
| 45 |
+
<li>Celeste 70B 0.1 data mixture minus Opus Instruct subset. See that model's <a href=https://huggingface.co/nothingiisreal/L3.1-70B-Celeste-V0.1-BF16>card</a> for details.</li>
|
| 46 |
+
<li>Kalomaze's Opus_Instruct_25k dataset, filtered for refusals.</li>
|
| 47 |
+
<li>A subset (1k rows) of ChatGPT-4o-WritingPrompts by Gryphe</li>
|
| 48 |
+
<li>A subset (2k rows) of Sonnet3.5-Charcards-Roleplay by Gryphe</li>
|
| 49 |
+
<li>A cleaned subset (~3k rows) of shortstories_synthlabels by Auri</li>
|
| 50 |
+
<li>Synthstruct and SynthRP datasets by Epiculous</li>
|
| 51 |
+
</ul>
|
| 52 |
+
<h3>
|
| 53 |
+
Hardware used:
|
| 54 |
+
</h3>
|
| 55 |
+
<ul><li>4xA6000 for 5 hours.</li></ul><br>
|
| 56 |
+
</p>
|
| 57 |
+
Model was trained by Kearm and Auri.
|
| 58 |
+
<h4>Special thanks:</h4><ul>
|
| 59 |
+
<li>to Gryphe, Lemmy, Kalomaze, Nopm and Epiculous for the data</li>
|
| 60 |
+
<li>to Alpindale helping with FFT config for Qwen2.5</li>
|
| 61 |
+
<li>and to InfermaticAI's community for their continued support for our endeavors</li></ul>
|