RichardErkhov
commited on
Commit
•
3a2a4d7
1
Parent(s):
bb08bb9
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,166 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
Buttocks-7B-v1.1 - bnb 8bits
|
11 |
+
- Model creator: https://huggingface.co/TeeZee/
|
12 |
+
- Original model: https://huggingface.co/TeeZee/Buttocks-7B-v1.1/
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
|
17 |
+
Original model description:
|
18 |
+
---
|
19 |
+
license: cc-by-nc-4.0
|
20 |
+
tags:
|
21 |
+
- not-for-all-audiences
|
22 |
+
- merge
|
23 |
+
model-index:
|
24 |
+
- name: Buttocks-7B-v1.1
|
25 |
+
results:
|
26 |
+
- task:
|
27 |
+
type: text-generation
|
28 |
+
name: Text Generation
|
29 |
+
dataset:
|
30 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
31 |
+
type: ai2_arc
|
32 |
+
config: ARC-Challenge
|
33 |
+
split: test
|
34 |
+
args:
|
35 |
+
num_few_shot: 25
|
36 |
+
metrics:
|
37 |
+
- type: acc_norm
|
38 |
+
value: 54.61
|
39 |
+
name: normalized accuracy
|
40 |
+
source:
|
41 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
|
42 |
+
name: Open LLM Leaderboard
|
43 |
+
- task:
|
44 |
+
type: text-generation
|
45 |
+
name: Text Generation
|
46 |
+
dataset:
|
47 |
+
name: HellaSwag (10-Shot)
|
48 |
+
type: hellaswag
|
49 |
+
split: validation
|
50 |
+
args:
|
51 |
+
num_few_shot: 10
|
52 |
+
metrics:
|
53 |
+
- type: acc_norm
|
54 |
+
value: 75.61
|
55 |
+
name: normalized accuracy
|
56 |
+
source:
|
57 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
|
58 |
+
name: Open LLM Leaderboard
|
59 |
+
- task:
|
60 |
+
type: text-generation
|
61 |
+
name: Text Generation
|
62 |
+
dataset:
|
63 |
+
name: MMLU (5-Shot)
|
64 |
+
type: cais/mmlu
|
65 |
+
config: all
|
66 |
+
split: test
|
67 |
+
args:
|
68 |
+
num_few_shot: 5
|
69 |
+
metrics:
|
70 |
+
- type: acc
|
71 |
+
value: 50.22
|
72 |
+
name: accuracy
|
73 |
+
source:
|
74 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
|
75 |
+
name: Open LLM Leaderboard
|
76 |
+
- task:
|
77 |
+
type: text-generation
|
78 |
+
name: Text Generation
|
79 |
+
dataset:
|
80 |
+
name: TruthfulQA (0-shot)
|
81 |
+
type: truthful_qa
|
82 |
+
config: multiple_choice
|
83 |
+
split: validation
|
84 |
+
args:
|
85 |
+
num_few_shot: 0
|
86 |
+
metrics:
|
87 |
+
- type: mc2
|
88 |
+
value: 44.72
|
89 |
+
source:
|
90 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
|
91 |
+
name: Open LLM Leaderboard
|
92 |
+
- task:
|
93 |
+
type: text-generation
|
94 |
+
name: Text Generation
|
95 |
+
dataset:
|
96 |
+
name: Winogrande (5-shot)
|
97 |
+
type: winogrande
|
98 |
+
config: winogrande_xl
|
99 |
+
split: validation
|
100 |
+
args:
|
101 |
+
num_few_shot: 5
|
102 |
+
metrics:
|
103 |
+
- type: acc
|
104 |
+
value: 68.9
|
105 |
+
name: accuracy
|
106 |
+
source:
|
107 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
|
108 |
+
name: Open LLM Leaderboard
|
109 |
+
- task:
|
110 |
+
type: text-generation
|
111 |
+
name: Text Generation
|
112 |
+
dataset:
|
113 |
+
name: GSM8k (5-shot)
|
114 |
+
type: gsm8k
|
115 |
+
config: main
|
116 |
+
split: test
|
117 |
+
args:
|
118 |
+
num_few_shot: 5
|
119 |
+
metrics:
|
120 |
+
- type: acc
|
121 |
+
value: 5.76
|
122 |
+
name: accuracy
|
123 |
+
source:
|
124 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
|
125 |
+
name: Open LLM Leaderboard
|
126 |
+
---
|
127 |
+
|
128 |
+
### Buttocks 7B v1.1 ###
|
129 |
+
|
130 |
+
An experiment that has gone very, very wrong.
|
131 |
+
|
132 |
+
### Model details ###
|
133 |
+
|
134 |
+
- Recreation of the original recipe for [Undi95/Toppy-M-7B](https://huggingface.co/Undi95/Toppy-M-7B), but instead of final merge done by mergekit, [MergeMoster](https://github.com/Gryphe/MergeMonster/) was used with extended RPG preset.
|
135 |
+
- recipe in [mergekit-config](https://huggingface.co/TeeZee/Toppy-7B-remake-mergemonster-SLERP-v1.0/resolve/main/toppy-slerp-merge-config.yml), stepsAA, BB, CC are the original models with LORAS as per Toppy M 7B sauce.
|
136 |
+
- LERP merge method was used
|
137 |
+
|
138 |
+
### Results ###
|
139 |
+
|
140 |
+
- in simple terms this model is totally unhinged
|
141 |
+
- it always produces sequences similar to fever dreams or drug trips
|
142 |
+
- on a good day it can produce scenarios similar to old Monty Python sketches
|
143 |
+
- models shows incredible affinity to words like 'ass', 'buttocks', 'farts', prompting with those single words will probably
|
144 |
+
produce a whole story revolving around those topics.
|
145 |
+
|
146 |
+
### Possible uses ###
|
147 |
+
|
148 |
+
- to generate dream sequence in a story
|
149 |
+
- to make the boring model more unpredictable by merging at low weights with this monster
|
150 |
+
- to take a break, connect Silly Tavern to this model and get a few ROTFLs observing how every story deteriorates into pure craziness
|
151 |
+
- research on LLM hallucinations
|
152 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
153 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_TeeZee__Buttocks-7B-v1.1)
|
154 |
+
|
155 |
+
| Metric |Value|
|
156 |
+
|---------------------------------|----:|
|
157 |
+
|Avg. |49.97|
|
158 |
+
|AI2 Reasoning Challenge (25-Shot)|54.61|
|
159 |
+
|HellaSwag (10-Shot) |75.61|
|
160 |
+
|MMLU (5-Shot) |50.22|
|
161 |
+
|TruthfulQA (0-shot) |44.72|
|
162 |
+
|Winogrande (5-shot) |68.90|
|
163 |
+
|GSM8k (5-shot) | 5.76|
|
164 |
+
|
165 |
+
|
166 |
+
|