Epiculous commited on
Commit
ff3da90
1 Parent(s): dc45ebb

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +74 -0
README.md ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ datasets:
4
+ - Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned
5
+ - anthracite-org/stheno-filtered-v1.1
6
+ - PJMixers/hieunguyenminh_roleplay-deduped-ShareGPT
7
+ - Gryphe/Sonnet3.5-Charcard-Roleplay
8
+ - Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned
9
+ - anthracite-org/kalo-opus-instruct-22k-no-refusal
10
+ - anthracite-org/nopm_claude_writing_fixed
11
+ - anthracite-org/kalo_opus_misc_240827
12
+ language:
13
+ - en
14
+ - fr
15
+ - de
16
+ - es
17
+ - it
18
+ - pt
19
+ - ru
20
+ - zh
21
+ - ja
22
+ pipeline_tag: text-generation
23
+ tags:
24
+ - merge
25
+ ---
26
+
27
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/64adfd277b5ff762771e4571/P962FQhRG4I8nbU_DJolY.png)
28
+
29
+ Now for something a bit different, Violet_Twilight-v0.2! This model is a SLERP merge of Azure_Dusk-v0.2 and Crimson_Dawn-v0.2!
30
+
31
+ # Quants!
32
+ [full](https://huggingface.co/Epiculous/Violet_Twilight-v0.2) / [exl2](https://huggingface.co/Epiculous/Violet_Twilight-v0.2-exl2) / <strong>gguf</strong>
33
+
34
+ ## Prompting
35
+ The v0.2 models are trained on ChatML, the prompting structure goes a little something like this:
36
+
37
+ ```
38
+ <|im_start|>user
39
+ Hi there!<|im_end|>
40
+ <|im_start|>assistant
41
+ Nice to meet you!<|im_end|>
42
+ <|im_start|>user
43
+ Can I ask a question?<|im_end|>
44
+ <|im_start|>assistant
45
+ ```
46
+
47
+ ### Context and Instruct
48
+ The v0.2 models are trained on ChatML, please use that Context and Instruct template.
49
+
50
+ ### Current Top Sampler Settings
51
+ [Spicy_Temp](https://files.catbox.moe/9npj0z.json) <br/>
52
+ [Violet_Twilight-Nitral-Special](https://files.catbox.moe/ot54u3.json) <br/>
53
+
54
+ ## Merging
55
+ The following config was used to merge Azure Dusk and Crimson Dawn
56
+ ```yaml
57
+ slices:
58
+ - sources:
59
+ - model: Epiculous/Azure_Dusk-v0.2
60
+ layer_range: [0, 40]
61
+ - model: Epiculous/Crimson_Dawn-V0.2
62
+ layer_range: [0, 40]
63
+ merge_method: slerp
64
+ base_model: Epiculous/Azure_Dusk-v0.2
65
+ parameters:
66
+ t:
67
+ - filter: self_attn
68
+ value: [0, 0.5, 0.3, 0.7, 1]
69
+ - filter: mlp
70
+ value: [1, 0.5, 0.7, 0.3, 0]
71
+ - value: 0.5 # fallback for rest of tensors
72
+ dtype: bfloat16
73
+
74
+ ```