Commit
•
2d5b216
1
Parent(s):
b108ae7
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,255 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
SOLAR-tail-10.7B-Merge-v1.0 - GGUF
|
11 |
+
- Model creator: https://huggingface.co/PracticeLLM/
|
12 |
+
- Original model: https://huggingface.co/PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q2_K.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q2_K.gguf) | Q2_K | 3.73GB |
|
18 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.IQ3_XS.gguf) | IQ3_XS | 4.14GB |
|
19 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.IQ3_S.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.IQ3_S.gguf) | IQ3_S | 4.37GB |
|
20 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q3_K_S.gguf) | Q3_K_S | 4.34GB |
|
21 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.IQ3_M.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.IQ3_M.gguf) | IQ3_M | 4.51GB |
|
22 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q3_K.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q3_K.gguf) | Q3_K | 4.84GB |
|
23 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q3_K_M.gguf) | Q3_K_M | 4.84GB |
|
24 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q3_K_L.gguf) | Q3_K_L | 5.26GB |
|
25 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.IQ4_XS.gguf) | IQ4_XS | 5.43GB |
|
26 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q4_0.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q4_0.gguf) | Q4_0 | 5.66GB |
|
27 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.IQ4_NL.gguf) | IQ4_NL | 5.72GB |
|
28 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q4_K_S.gguf) | Q4_K_S | 5.7GB |
|
29 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q4_K.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q4_K.gguf) | Q4_K | 6.02GB |
|
30 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q4_K_M.gguf) | Q4_K_M | 6.02GB |
|
31 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q4_1.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q4_1.gguf) | Q4_1 | 6.27GB |
|
32 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q5_0.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q5_0.gguf) | Q5_0 | 6.89GB |
|
33 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q5_K_S.gguf) | Q5_K_S | 6.89GB |
|
34 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q5_K.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q5_K.gguf) | Q5_K | 7.08GB |
|
35 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q5_K_M.gguf) | Q5_K_M | 7.08GB |
|
36 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q5_1.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q5_1.gguf) | Q5_1 | 7.51GB |
|
37 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q6_K.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q6_K.gguf) | Q6_K | 8.2GB |
|
38 |
+
| [SOLAR-tail-10.7B-Merge-v1.0.Q8_0.gguf](https://huggingface.co/RichardErkhov/PracticeLLM_-_SOLAR-tail-10.7B-Merge-v1.0-gguf/blob/main/SOLAR-tail-10.7B-Merge-v1.0.Q8_0.gguf) | Q8_0 | 10.62GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
language:
|
46 |
+
- en
|
47 |
+
- ko
|
48 |
+
license: cc-by-nc-sa-4.0
|
49 |
+
pipeline_tag: text-generation
|
50 |
+
model-index:
|
51 |
+
- name: SOLAR-tail-10.7B-Merge-v1.0
|
52 |
+
results:
|
53 |
+
- task:
|
54 |
+
type: text-generation
|
55 |
+
name: Text Generation
|
56 |
+
dataset:
|
57 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
58 |
+
type: ai2_arc
|
59 |
+
config: ARC-Challenge
|
60 |
+
split: test
|
61 |
+
args:
|
62 |
+
num_few_shot: 25
|
63 |
+
metrics:
|
64 |
+
- type: acc_norm
|
65 |
+
value: 66.13
|
66 |
+
name: normalized accuracy
|
67 |
+
source:
|
68 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0
|
69 |
+
name: Open LLM Leaderboard
|
70 |
+
- task:
|
71 |
+
type: text-generation
|
72 |
+
name: Text Generation
|
73 |
+
dataset:
|
74 |
+
name: HellaSwag (10-Shot)
|
75 |
+
type: hellaswag
|
76 |
+
split: validation
|
77 |
+
args:
|
78 |
+
num_few_shot: 10
|
79 |
+
metrics:
|
80 |
+
- type: acc_norm
|
81 |
+
value: 86.54
|
82 |
+
name: normalized accuracy
|
83 |
+
source:
|
84 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0
|
85 |
+
name: Open LLM Leaderboard
|
86 |
+
- task:
|
87 |
+
type: text-generation
|
88 |
+
name: Text Generation
|
89 |
+
dataset:
|
90 |
+
name: MMLU (5-Shot)
|
91 |
+
type: cais/mmlu
|
92 |
+
config: all
|
93 |
+
split: test
|
94 |
+
args:
|
95 |
+
num_few_shot: 5
|
96 |
+
metrics:
|
97 |
+
- type: acc
|
98 |
+
value: 66.52
|
99 |
+
name: accuracy
|
100 |
+
source:
|
101 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0
|
102 |
+
name: Open LLM Leaderboard
|
103 |
+
- task:
|
104 |
+
type: text-generation
|
105 |
+
name: Text Generation
|
106 |
+
dataset:
|
107 |
+
name: TruthfulQA (0-shot)
|
108 |
+
type: truthful_qa
|
109 |
+
config: multiple_choice
|
110 |
+
split: validation
|
111 |
+
args:
|
112 |
+
num_few_shot: 0
|
113 |
+
metrics:
|
114 |
+
- type: mc2
|
115 |
+
value: 60.57
|
116 |
+
source:
|
117 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0
|
118 |
+
name: Open LLM Leaderboard
|
119 |
+
- task:
|
120 |
+
type: text-generation
|
121 |
+
name: Text Generation
|
122 |
+
dataset:
|
123 |
+
name: Winogrande (5-shot)
|
124 |
+
type: winogrande
|
125 |
+
config: winogrande_xl
|
126 |
+
split: validation
|
127 |
+
args:
|
128 |
+
num_few_shot: 5
|
129 |
+
metrics:
|
130 |
+
- type: acc
|
131 |
+
value: 84.77
|
132 |
+
name: accuracy
|
133 |
+
source:
|
134 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0
|
135 |
+
name: Open LLM Leaderboard
|
136 |
+
- task:
|
137 |
+
type: text-generation
|
138 |
+
name: Text Generation
|
139 |
+
dataset:
|
140 |
+
name: GSM8k (5-shot)
|
141 |
+
type: gsm8k
|
142 |
+
config: main
|
143 |
+
split: test
|
144 |
+
args:
|
145 |
+
num_few_shot: 5
|
146 |
+
metrics:
|
147 |
+
- type: acc
|
148 |
+
value: 65.58
|
149 |
+
name: accuracy
|
150 |
+
source:
|
151 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0
|
152 |
+
name: Open LLM Leaderboard
|
153 |
+
---
|
154 |
+
|
155 |
+
# **SOLAR-tail-10.7B-Merge-v1.0**
|
156 |
+
|
157 |
+
## Model Details
|
158 |
+
|
159 |
+
**Model Developers** Kyujin Han (kyujinpy)
|
160 |
+
|
161 |
+
**Method**
|
162 |
+
Using [Mergekit](https://github.com/cg123/mergekit).
|
163 |
+
- [upstage/SOLAR-10.7B-v1.0](https://huggingface.co/upstage/SOLAR-10.7B-v1.0)
|
164 |
+
- [Yhyu13/LMCocktail-10.7B-v1](Yhyu13/LMCocktail-10.7B-v1)
|
165 |
+
|
166 |
+
**Merge config**
|
167 |
+
```
|
168 |
+
slices:
|
169 |
+
- sources:
|
170 |
+
- model: upstage/SOLAR-10.7B-v1.0
|
171 |
+
layer_range: [0, 48]
|
172 |
+
- model: Yhyu13/LMCocktail-10.7B-v1
|
173 |
+
layer_range: [0, 48]
|
174 |
+
|
175 |
+
merge_method: slerp
|
176 |
+
base_model: upstage/SOLAR-10.7B-v1.0
|
177 |
+
|
178 |
+
parameters:
|
179 |
+
t:
|
180 |
+
- filter: self_attn
|
181 |
+
value: [0, 0.5, 0.3, 0.7, 1]
|
182 |
+
- filter: mlp
|
183 |
+
value: [1, 0.5, 0.7, 0.3, 0]
|
184 |
+
- value: 0.5 # fallback for rest of tensors
|
185 |
+
tokenizer_source: union
|
186 |
+
|
187 |
+
dtype: float16
|
188 |
+
```
|
189 |
+
|
190 |
+
# **Model Benchmark**
|
191 |
+
|
192 |
+
## Open Ko leaderboard
|
193 |
+
- Follow up as [Ko-link](https://huggingface.co/spaces/upstage/open-ko-llm-leaderboard).
|
194 |
+
|
195 |
+
| Model | Average | ARC | HellaSwag | MMLU | TruthfulQA | Ko-CommonGenV2 |
|
196 |
+
| --- | --- | --- | --- | --- | --- | --- |
|
197 |
+
| PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0 | 48.32 | 45.73 | 56.97 | 38.77 | 38.75 | 61.16 |
|
198 |
+
| jjourney1125/M-SOLAR-10.7B-v1.0 | 55.15 | 49.57 | 60.12 | 54.60 | 49.23 | 62.22 |
|
199 |
+
|
200 |
+
- Follow up as [En-link](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).
|
201 |
+
| Model | Average | ARC | HellaSwag | MMLU | TruthfulQA | Winogrande | GSM8K |
|
202 |
+
| --- | --- | --- | --- | --- | --- | --- | --- |
|
203 |
+
| PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0 | 71.68 | 66.13 | 86.54 | **66.52** | 60.57 | **84.77** | **65.58** |
|
204 |
+
| kyujinpy/Sakura-SOLAR-Instruct | **74.40** | **70.99** | **88.42** | 66.33 | **71.79** | 83.66 | 65.20 |
|
205 |
+
|
206 |
+
|
207 |
+
## lm-evaluation-harness
|
208 |
+
```
|
209 |
+
gpt2 (pretrained=PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0), limit: None, provide_description: False, num_fewshot: 0, batch_size: None
|
210 |
+
| Task |Version| Metric |Value | |Stderr|
|
211 |
+
|----------------|------:|--------|-----:|---|-----:|
|
212 |
+
|kobest_boolq | 0|acc |0.5021|± |0.0133|
|
213 |
+
| | |macro_f1|0.3343|± |0.0059|
|
214 |
+
|kobest_copa | 0|acc |0.6220|± |0.0153|
|
215 |
+
| | |macro_f1|0.6217|± |0.0154|
|
216 |
+
|kobest_hellaswag| 0|acc |0.4380|± |0.0222|
|
217 |
+
| | |acc_norm|0.5380|± |0.0223|
|
218 |
+
| | |macro_f1|0.4366|± |0.0222|
|
219 |
+
|kobest_sentineg | 0|acc |0.4962|± |0.0251|
|
220 |
+
| | |macro_f1|0.3316|± |0.0113|
|
221 |
+
```
|
222 |
+
|
223 |
+
|
224 |
+
# Implementation Code
|
225 |
+
```python
|
226 |
+
### KO-Platypus
|
227 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
228 |
+
import torch
|
229 |
+
|
230 |
+
repo = "PracticeLLM/SOLAR-tail-10.7B-Merge-v1.0"
|
231 |
+
OpenOrca = AutoModelForCausalLM.from_pretrained(
|
232 |
+
repo,
|
233 |
+
return_dict=True,
|
234 |
+
torch_dtype=torch.float16,
|
235 |
+
device_map='auto'
|
236 |
+
)
|
237 |
+
OpenOrca_tokenizer = AutoTokenizer.from_pretrained(repo)
|
238 |
+
```
|
239 |
+
|
240 |
+
---
|
241 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
242 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_PracticeLLM__SOLAR-tail-10.7B-Merge-v1.0)
|
243 |
+
|
244 |
+
| Metric |Value|
|
245 |
+
|---------------------------------|----:|
|
246 |
+
|Avg. |71.68|
|
247 |
+
|AI2 Reasoning Challenge (25-Shot)|66.13|
|
248 |
+
|HellaSwag (10-Shot) |86.54|
|
249 |
+
|MMLU (5-Shot) |66.52|
|
250 |
+
|TruthfulQA (0-shot) |60.57|
|
251 |
+
|Winogrande (5-shot) |84.77|
|
252 |
+
|GSM8k (5-shot) |65.58|
|
253 |
+
|
254 |
+
|
255 |
+
|