RichardErkhov
commited on
Commit
•
db622c6
1
Parent(s):
4dc579f
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,334 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
CosmicBun-8B - GGUF
|
11 |
+
- Model creator: https://huggingface.co/aloobun/
|
12 |
+
- Original model: https://huggingface.co/aloobun/CosmicBun-8B/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [CosmicBun-8B.Q2_K.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q2_K.gguf) | Q2_K | 2.96GB |
|
18 |
+
| [CosmicBun-8B.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.IQ3_XS.gguf) | IQ3_XS | 3.28GB |
|
19 |
+
| [CosmicBun-8B.IQ3_S.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.IQ3_S.gguf) | IQ3_S | 3.43GB |
|
20 |
+
| [CosmicBun-8B.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q3_K_S.gguf) | Q3_K_S | 3.41GB |
|
21 |
+
| [CosmicBun-8B.IQ3_M.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.IQ3_M.gguf) | IQ3_M | 3.52GB |
|
22 |
+
| [CosmicBun-8B.Q3_K.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q3_K.gguf) | Q3_K | 3.74GB |
|
23 |
+
| [CosmicBun-8B.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q3_K_M.gguf) | Q3_K_M | 3.74GB |
|
24 |
+
| [CosmicBun-8B.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q3_K_L.gguf) | Q3_K_L | 4.03GB |
|
25 |
+
| [CosmicBun-8B.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.IQ4_XS.gguf) | IQ4_XS | 4.18GB |
|
26 |
+
| [CosmicBun-8B.Q4_0.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q4_0.gguf) | Q4_0 | 4.34GB |
|
27 |
+
| [CosmicBun-8B.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.IQ4_NL.gguf) | IQ4_NL | 4.38GB |
|
28 |
+
| [CosmicBun-8B.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q4_K_S.gguf) | Q4_K_S | 4.37GB |
|
29 |
+
| [CosmicBun-8B.Q4_K.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q4_K.gguf) | Q4_K | 4.58GB |
|
30 |
+
| [CosmicBun-8B.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q4_K_M.gguf) | Q4_K_M | 4.58GB |
|
31 |
+
| [CosmicBun-8B.Q4_1.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q4_1.gguf) | Q4_1 | 4.78GB |
|
32 |
+
| [CosmicBun-8B.Q5_0.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q5_0.gguf) | Q5_0 | 5.21GB |
|
33 |
+
| [CosmicBun-8B.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q5_K_S.gguf) | Q5_K_S | 5.21GB |
|
34 |
+
| [CosmicBun-8B.Q5_K.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q5_K.gguf) | Q5_K | 5.34GB |
|
35 |
+
| [CosmicBun-8B.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q5_K_M.gguf) | Q5_K_M | 5.34GB |
|
36 |
+
| [CosmicBun-8B.Q5_1.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q5_1.gguf) | Q5_1 | 5.65GB |
|
37 |
+
| [CosmicBun-8B.Q6_K.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q6_K.gguf) | Q6_K | 6.14GB |
|
38 |
+
| [CosmicBun-8B.Q8_0.gguf](https://huggingface.co/RichardErkhov/aloobun_-_CosmicBun-8B-gguf/blob/main/CosmicBun-8B.Q8_0.gguf) | Q8_0 | 7.95GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
license: mit
|
46 |
+
library_name: transformers
|
47 |
+
tags:
|
48 |
+
- mergekit
|
49 |
+
- merge
|
50 |
+
- math
|
51 |
+
- llama3
|
52 |
+
- physics
|
53 |
+
- chemistry
|
54 |
+
- biology
|
55 |
+
- dolphin
|
56 |
+
base_model:
|
57 |
+
- cognitivecomputations/dolphin-2.9-llama3-8b
|
58 |
+
- Weyaxi/Einstein-v6.1-Llama3-8B
|
59 |
+
- Locutusque/llama-3-neural-chat-v1-8b
|
60 |
+
model-index:
|
61 |
+
- name: CosmicBun-8B
|
62 |
+
results:
|
63 |
+
- task:
|
64 |
+
type: text-generation
|
65 |
+
name: Text Generation
|
66 |
+
dataset:
|
67 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
68 |
+
type: ai2_arc
|
69 |
+
config: ARC-Challenge
|
70 |
+
split: test
|
71 |
+
args:
|
72 |
+
num_few_shot: 25
|
73 |
+
metrics:
|
74 |
+
- type: acc_norm
|
75 |
+
value: 61.86
|
76 |
+
name: normalized accuracy
|
77 |
+
source:
|
78 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/CosmicBun-8B
|
79 |
+
name: Open LLM Leaderboard
|
80 |
+
- task:
|
81 |
+
type: text-generation
|
82 |
+
name: Text Generation
|
83 |
+
dataset:
|
84 |
+
name: HellaSwag (10-Shot)
|
85 |
+
type: hellaswag
|
86 |
+
split: validation
|
87 |
+
args:
|
88 |
+
num_few_shot: 10
|
89 |
+
metrics:
|
90 |
+
- type: acc_norm
|
91 |
+
value: 84.29
|
92 |
+
name: normalized accuracy
|
93 |
+
source:
|
94 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/CosmicBun-8B
|
95 |
+
name: Open LLM Leaderboard
|
96 |
+
- task:
|
97 |
+
type: text-generation
|
98 |
+
name: Text Generation
|
99 |
+
dataset:
|
100 |
+
name: MMLU (5-Shot)
|
101 |
+
type: cais/mmlu
|
102 |
+
config: all
|
103 |
+
split: test
|
104 |
+
args:
|
105 |
+
num_few_shot: 5
|
106 |
+
metrics:
|
107 |
+
- type: acc
|
108 |
+
value: 65.53
|
109 |
+
name: accuracy
|
110 |
+
source:
|
111 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/CosmicBun-8B
|
112 |
+
name: Open LLM Leaderboard
|
113 |
+
- task:
|
114 |
+
type: text-generation
|
115 |
+
name: Text Generation
|
116 |
+
dataset:
|
117 |
+
name: TruthfulQA (0-shot)
|
118 |
+
type: truthful_qa
|
119 |
+
config: multiple_choice
|
120 |
+
split: validation
|
121 |
+
args:
|
122 |
+
num_few_shot: 0
|
123 |
+
metrics:
|
124 |
+
- type: mc2
|
125 |
+
value: 54.08
|
126 |
+
source:
|
127 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/CosmicBun-8B
|
128 |
+
name: Open LLM Leaderboard
|
129 |
+
- task:
|
130 |
+
type: text-generation
|
131 |
+
name: Text Generation
|
132 |
+
dataset:
|
133 |
+
name: Winogrande (5-shot)
|
134 |
+
type: winogrande
|
135 |
+
config: winogrande_xl
|
136 |
+
split: validation
|
137 |
+
args:
|
138 |
+
num_few_shot: 5
|
139 |
+
metrics:
|
140 |
+
- type: acc
|
141 |
+
value: 78.85
|
142 |
+
name: accuracy
|
143 |
+
source:
|
144 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/CosmicBun-8B
|
145 |
+
name: Open LLM Leaderboard
|
146 |
+
- task:
|
147 |
+
type: text-generation
|
148 |
+
name: Text Generation
|
149 |
+
dataset:
|
150 |
+
name: GSM8k (5-shot)
|
151 |
+
type: gsm8k
|
152 |
+
config: main
|
153 |
+
split: test
|
154 |
+
args:
|
155 |
+
num_few_shot: 5
|
156 |
+
metrics:
|
157 |
+
- type: acc
|
158 |
+
value: 68.23
|
159 |
+
name: accuracy
|
160 |
+
source:
|
161 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/CosmicBun-8B
|
162 |
+
name: Open LLM Leaderboard
|
163 |
+
---
|
164 |
+
# model
|
165 |
+
|
166 |
+
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
167 |
+
|
168 |
+
### Merge Method
|
169 |
+
|
170 |
+
This model was merged using the [DARE](https://arxiv.org/abs/2311.03099) [TIES](https://arxiv.org/abs/2306.01708) merge method using [Locutusque/llama-3-neural-chat-v1-8b](https://huggingface.co/Locutusque/llama-3-neural-chat-v1-8b) as a base.
|
171 |
+
|
172 |
+
### Models Merged
|
173 |
+
|
174 |
+
The following models were included in the merge:
|
175 |
+
* [cognitivecomputations/dolphin-2.9-llama3-8b](https://huggingface.co/cognitivecomputations/dolphin-2.9-llama3-8b)
|
176 |
+
* [Weyaxi/Einstein-v6.1-Llama3-8B](https://huggingface.co/Weyaxi/Einstein-v6.1-Llama3-8B)
|
177 |
+
|
178 |
+
### Configuration
|
179 |
+
|
180 |
+
The following YAML configuration was used to produce this model:
|
181 |
+
|
182 |
+
```yaml
|
183 |
+
base_model: Locutusque/llama-3-neural-chat-v1-8b
|
184 |
+
dtype: bfloat16
|
185 |
+
merge_method: dare_ties
|
186 |
+
parameters:
|
187 |
+
int8_mask: 1.0
|
188 |
+
normalize: 0.0
|
189 |
+
slices:
|
190 |
+
- sources:
|
191 |
+
- layer_range: [0, 4]
|
192 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
193 |
+
parameters:
|
194 |
+
density: 1.0
|
195 |
+
weight: 0.6
|
196 |
+
- layer_range: [0, 4]
|
197 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
198 |
+
parameters:
|
199 |
+
density: 0.6
|
200 |
+
weight: 0.5
|
201 |
+
- layer_range: [0, 4]
|
202 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
203 |
+
parameters:
|
204 |
+
density: 1.0
|
205 |
+
weight: 0.5
|
206 |
+
- sources:
|
207 |
+
- layer_range: [4, 8]
|
208 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
209 |
+
parameters:
|
210 |
+
density: 0.8
|
211 |
+
weight: 0.1
|
212 |
+
- layer_range: [4, 8]
|
213 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
214 |
+
parameters:
|
215 |
+
density: 1.0
|
216 |
+
weight: 0.2
|
217 |
+
- layer_range: [4, 8]
|
218 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
219 |
+
parameters:
|
220 |
+
density: 1.0
|
221 |
+
weight: 0.7
|
222 |
+
- sources:
|
223 |
+
- layer_range: [8, 12]
|
224 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
225 |
+
parameters:
|
226 |
+
density: 0.7
|
227 |
+
weight: 0.1
|
228 |
+
- layer_range: [8, 12]
|
229 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
230 |
+
parameters:
|
231 |
+
density: 0.7
|
232 |
+
weight: 0.2
|
233 |
+
- layer_range: [8, 12]
|
234 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
235 |
+
parameters:
|
236 |
+
density: 0.7
|
237 |
+
weight: 0.6
|
238 |
+
- sources:
|
239 |
+
- layer_range: [12, 16]
|
240 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
241 |
+
parameters:
|
242 |
+
density: 0.9
|
243 |
+
weight: 0.2
|
244 |
+
- layer_range: [12, 16]
|
245 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
246 |
+
parameters:
|
247 |
+
density: 0.6
|
248 |
+
weight: 0.6
|
249 |
+
- layer_range: [12, 16]
|
250 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
251 |
+
parameters:
|
252 |
+
density: 0.7
|
253 |
+
weight: 0.3
|
254 |
+
- sources:
|
255 |
+
- layer_range: [16, 20]
|
256 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
257 |
+
parameters:
|
258 |
+
density: 1.0
|
259 |
+
weight: 0.2
|
260 |
+
- layer_range: [16, 20]
|
261 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
262 |
+
parameters:
|
263 |
+
density: 1.0
|
264 |
+
weight: 0.2
|
265 |
+
- layer_range: [16, 20]
|
266 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
267 |
+
parameters:
|
268 |
+
density: 0.9
|
269 |
+
weight: 0.4
|
270 |
+
- sources:
|
271 |
+
- layer_range: [20, 24]
|
272 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
273 |
+
parameters:
|
274 |
+
density: 0.7
|
275 |
+
weight: 0.2
|
276 |
+
- layer_range: [20, 24]
|
277 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
278 |
+
parameters:
|
279 |
+
density: 0.9
|
280 |
+
weight: 0.3
|
281 |
+
- layer_range: [20, 24]
|
282 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
283 |
+
parameters:
|
284 |
+
density: 1.0
|
285 |
+
weight: 0.4
|
286 |
+
- sources:
|
287 |
+
- layer_range: [24, 28]
|
288 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
289 |
+
parameters:
|
290 |
+
density: 1.0
|
291 |
+
weight: 0.4
|
292 |
+
- layer_range: [24, 28]
|
293 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
294 |
+
parameters:
|
295 |
+
density: 0.8
|
296 |
+
weight: 0.2
|
297 |
+
- layer_range: [24, 28]
|
298 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
299 |
+
parameters:
|
300 |
+
density: 0.9
|
301 |
+
weight: 0.4
|
302 |
+
- sources:
|
303 |
+
- layer_range: [28, 32]
|
304 |
+
model: cognitivecomputations/dolphin-2.9-llama3-8b
|
305 |
+
parameters:
|
306 |
+
density: 1.0
|
307 |
+
weight: 0.3
|
308 |
+
- layer_range: [28, 32]
|
309 |
+
model: Weyaxi/Einstein-v6.1-Llama3-8B
|
310 |
+
parameters:
|
311 |
+
density: 0.9
|
312 |
+
weight: 0.2
|
313 |
+
- layer_range: [28, 32]
|
314 |
+
model: Locutusque/llama-3-neural-chat-v1-8b
|
315 |
+
parameters:
|
316 |
+
density: 1.0
|
317 |
+
weight: 0.3
|
318 |
+
|
319 |
+
```
|
320 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
321 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_aloobun__CosmicBun-8B)
|
322 |
+
|
323 |
+
| Metric |Value|
|
324 |
+
|---------------------------------|----:|
|
325 |
+
|Avg. |68.81|
|
326 |
+
|AI2 Reasoning Challenge (25-Shot)|61.86|
|
327 |
+
|HellaSwag (10-Shot) |84.29|
|
328 |
+
|MMLU (5-Shot) |65.53|
|
329 |
+
|TruthfulQA (0-shot) |54.08|
|
330 |
+
|Winogrande (5-shot) |78.85|
|
331 |
+
|GSM8k (5-shot) |68.23|
|
332 |
+
|
333 |
+
|
334 |
+
|