Hugging Face's logo

RichardErkhov
/
llmixer_-_BigWeave-v27-95b-gguf

GGUF
Model card Files
xet
Community
RichardErkhov commited on
Commit
9fb355c
·
verified ·
1 Parent(s): f4d252d

uploaded readme

Browse files
Files changed (1) hide show
  1. README.md +108 -0
README.md ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Quantization made by Richard Erkhov.
2
+
3
+ [Github](https://github.com/RichardErkhov)
4
+
5
+ [Discord](https://discord.gg/pvy7H8DZMG)
6
+
7
+ [Request more models](https://github.com/RichardErkhov/quant_request)
8
+
9
+
10
+ BigWeave-v27-95b - GGUF
11
+ - Model creator: https://huggingface.co/llmixer/
12
+ - Original model: https://huggingface.co/llmixer/BigWeave-v27-95b/
13
+
14
+
15
+ | Name | Quant method | Size |
16
+ | ---- | ---- | ---- |
17
+ | [BigWeave-v27-95b.Q2_K.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/blob/main/BigWeave-v27-95b.Q2_K.gguf) | Q2_K | 32.98GB |
18
+ | [BigWeave-v27-95b.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/blob/main/BigWeave-v27-95b.IQ3_XS.gguf) | IQ3_XS | 36.69GB |
19
+ | [BigWeave-v27-95b.IQ3_S.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | IQ3_S | 38.78GB |
20
+ | [BigWeave-v27-95b.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q3_K_S | 38.66GB |
21
+ | [BigWeave-v27-95b.IQ3_M.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | IQ3_M | 40.12GB |
22
+ | [BigWeave-v27-95b.Q3_K.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q3_K | 43.16GB |
23
+ | [BigWeave-v27-95b.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q3_K_M | 43.16GB |
24
+ | [BigWeave-v27-95b.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q3_K_L | 47.01GB |
25
+ | [BigWeave-v27-95b.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | IQ4_XS | 48.37GB |
26
+ | [BigWeave-v27-95b.Q4_0.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q4_0 | 50.55GB |
27
+ | [BigWeave-v27-95b.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | IQ4_NL | 51.04GB |
28
+ | [BigWeave-v27-95b.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q4_K_S | 50.94GB |
29
+ | [BigWeave-v27-95b.Q4_K.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q4_K | 53.82GB |
30
+ | [BigWeave-v27-95b.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q4_K_M | 53.82GB |
31
+ | [BigWeave-v27-95b.Q4_1.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q4_1 | 56.14GB |
32
+ | [BigWeave-v27-95b.Q5_0.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q5_0 | 61.74GB |
33
+ | [BigWeave-v27-95b.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q5_K_S | 61.74GB |
34
+ | [BigWeave-v27-95b.Q5_K.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q5_K | 63.42GB |
35
+ | [BigWeave-v27-95b.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q5_K_M | 63.42GB |
36
+ | [BigWeave-v27-95b.Q5_1.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q5_1 | 67.33GB |
37
+ | [BigWeave-v27-95b.Q6_K.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q6_K | 73.62GB |
38
+ | [BigWeave-v27-95b.Q8_0.gguf](https://huggingface.co/RichardErkhov/llmixer_-_BigWeave-v27-95b-gguf/tree/main/) | Q8_0 | 95.35GB |
39
+
40
+
41
+
42
+
43
+ Original model description:
44
+ ---
45
+ base_model:
46
+ - 152334H/miqu-1-70b-sf
47
+ license: unknown
48
+ language:
49
+ - en
50
+ pipeline_tag: text-generation
51
+ tags:
52
+ - merge
53
+ - frankenmerge
54
+ - 95b
55
+ ---
56
+ # BigWeave v27 95b
57
+
58
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/65a6db055c58475cf9e6def1/4CbbAN-X7ZWj702JrcCGH.png" width=600>
59
+
60
+ The BigWeave models aim to experimentally identify merge settings for increasing model performance. The version number merely tracks various attempts and is not a quality indicator. Only results demonstrating good performance are retained and shared.
61
+
62
+ # Prompting Format
63
+ Chatml, Mistral, Vicuna.
64
+
65
+ # Merge process
66
+ This is a self-merge of 152334H/miqu-1-70b-sf. The 30 most important layers (according to exl2 measurements) are duplicated with 50% overlap.
67
+
68
+ Merge configuration:
69
+ ```
70
+ slices:
71
+ - sources:
72
+ - model: 152334H/miqu-1-70b-sf
73
+ layer_range: [0,40]
74
+ - sources:
75
+ - model: 152334H/miqu-1-70b-sf
76
+ layer_range: [34,45] # dup 34-44
77
+ - sources:
78
+ - model: 152334H/miqu-1-70b-sf
79
+ layer_range: [40,52]
80
+ - sources:
81
+ - model: 152334H/miqu-1-70b-sf
82
+ layer_range: [51,53] # dup 51-52
83
+ - sources:
84
+ - model: 152334H/miqu-1-70b-sf
85
+ layer_range: [52,55]
86
+ - sources:
87
+ - model: 152334H/miqu-1-70b-sf
88
+ layer_range: [54,56] # dup 54-55
89
+ - sources:
90
+ - model: 152334H/miqu-1-70b-sf
91
+ layer_range: [55,59]
92
+ - sources:
93
+ - model: 152334H/miqu-1-70b-sf
94
+ layer_range: [58,60] # dup 58-59
95
+ - sources:
96
+ - model: 152334H/miqu-1-70b-sf
97
+ layer_range: [59,72]
98
+ - sources:
99
+ - model: 152334H/miqu-1-70b-sf
100
+ layer_range: [64,79] # dup 64-78
101
+ - sources:
102
+ - model: 152334H/miqu-1-70b-sf
103
+ layer_range: [72,80]
104
+ merge_method: passthrough
105
+ dtype: float16
106
+
107
+ ```
108
+