Safetensors
llama
File size: 2,201 Bytes
ef8e597
 
 
 
8caa530
ef8e597
 
 
 
 
 
 
 
 
9948900
 
e77b004
8237ceb
e77b004
 
e19c9ab
09ee31d
e77b004
28e06c0
e77b004
9948900
 
057b631
9948900
9104b9a
 
9948900
5595bd2
9948900
 
 
057b631
 
 
a4083ad
057b631
 
 
 
 
a4083ad
057b631
 
 
 
a4083ad
057b631
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
---
license: apache-2.0
datasets:
- kalomaze/Opus_Instruct_25k
- allura-org/Celeste-1.x-data-mixture
base_model:
- AuriAetherwiing/Yi-1.5-9B-32K-tokfix
---

**EVA Yi 1.5 9B v1**
  
<p>
  A RP/storywriting focused model, full-parameter finetune of Yi-1.5-9B-32K on mixture of synthetic and natural data.<br>
  A continuation of nothingiisreal's Celeste 1.x series, made to improve stability and versatility, without losing unique, diverse writing style of Celeste.
</p>

<p>
  <h3>Quants: (GGUF is not recommended, lcpp breaks tokenizer fix)</h3>
  <ul>
    <li><a href=https://huggingface.co./bartowski/EVA-Yi-1.5-9B-32K-V1-GGUF>IMatrix GGUF by bartowski</a></li>
    <li><a href=https://huggingface.co./mradermacher/EVA-Yi-1.5-9B-32K-V1-GGUF>Static GGUF by Mradermacher</a></li>
    <li><a href=https://huggingface.co./bartowski/EVA-Yi-1.5-9B-32K-V1-exl2>EXL2 by bartowski</a></li>
  </ul>
   We recommend using original BFloat16 weights, quantization seems to affect Yi significantly more than other model architectures. 
</p>
<p>
  Prompt format is ChatML.<br>
  <h3>Recommended sampler values:</h3>
  
  -  Temperature: 1
  -  Min-P: 0.05
  
  <h3>Recommended SillyTavern presets (via CalamitousFelicitousness):</h3>

  - [Context](https://huggingface.co./EVA-UNIT-01/EVA-Yi-1.5-9B-32K-V1/blob/main/%5BChatML%5D%20Roleplay-v1.9%20Context.json)
  - [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)
</p>

<p>
  <br>
  <h3>
    Training data:
  </h3>
    <ul>
       <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>
      <li>Kalomaze's Opus_Instruct_25k dataset, filtered for refusals.</li></ul>
  <h3>
     Hardware used:
  </h3>
      <ul><li>4x3090Ti for 5 days.</li></ul><br>
</p>
  Model was trained by Kearm and Auri.
  <h4>Special thanks:</h4><ul>
  <li>to Lemmy, Gryphe, Kalomaze and Nopm for the data</li>
  <li>to ALK, Fizz and CalamitousFelicitousness for Yi tokenizer fix</li>
  <li>and to InfermaticAI's community for their continued support for our endeavors</li></ul>