Commit
•
75cf776
1
Parent(s):
8701926
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,223 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
KangalKhan-RawRuby-7B - bnb 8bits
|
11 |
+
- Model creator: https://huggingface.co/Yuma42/
|
12 |
+
- Original model: https://huggingface.co/Yuma42/KangalKhan-RawRuby-7B/
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
|
17 |
+
Original model description:
|
18 |
+
---
|
19 |
+
language:
|
20 |
+
- en
|
21 |
+
license: apache-2.0
|
22 |
+
tags:
|
23 |
+
- merge
|
24 |
+
- mergekit
|
25 |
+
- lazymergekit
|
26 |
+
- Yuma42/KangalKhan-Ruby-7B-Fixed
|
27 |
+
- Yuma42/KangalKhan-RawEmerald-7B
|
28 |
+
base_model:
|
29 |
+
- Yuma42/KangalKhan-Ruby-7B-Fixed
|
30 |
+
- Yuma42/KangalKhan-RawEmerald-7B
|
31 |
+
model-index:
|
32 |
+
- name: KangalKhan-RawRuby-7B
|
33 |
+
results:
|
34 |
+
- task:
|
35 |
+
type: text-generation
|
36 |
+
name: Text Generation
|
37 |
+
dataset:
|
38 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
39 |
+
type: ai2_arc
|
40 |
+
config: ARC-Challenge
|
41 |
+
split: test
|
42 |
+
args:
|
43 |
+
num_few_shot: 25
|
44 |
+
metrics:
|
45 |
+
- type: acc_norm
|
46 |
+
value: 66.89
|
47 |
+
name: normalized accuracy
|
48 |
+
source:
|
49 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Yuma42/KangalKhan-RawRuby-7B
|
50 |
+
name: Open LLM Leaderboard
|
51 |
+
- task:
|
52 |
+
type: text-generation
|
53 |
+
name: Text Generation
|
54 |
+
dataset:
|
55 |
+
name: HellaSwag (10-Shot)
|
56 |
+
type: hellaswag
|
57 |
+
split: validation
|
58 |
+
args:
|
59 |
+
num_few_shot: 10
|
60 |
+
metrics:
|
61 |
+
- type: acc_norm
|
62 |
+
value: 85.53
|
63 |
+
name: normalized accuracy
|
64 |
+
source:
|
65 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Yuma42/KangalKhan-RawRuby-7B
|
66 |
+
name: Open LLM Leaderboard
|
67 |
+
- task:
|
68 |
+
type: text-generation
|
69 |
+
name: Text Generation
|
70 |
+
dataset:
|
71 |
+
name: MMLU (5-Shot)
|
72 |
+
type: cais/mmlu
|
73 |
+
config: all
|
74 |
+
split: test
|
75 |
+
args:
|
76 |
+
num_few_shot: 5
|
77 |
+
metrics:
|
78 |
+
- type: acc
|
79 |
+
value: 63.46
|
80 |
+
name: accuracy
|
81 |
+
source:
|
82 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Yuma42/KangalKhan-RawRuby-7B
|
83 |
+
name: Open LLM Leaderboard
|
84 |
+
- task:
|
85 |
+
type: text-generation
|
86 |
+
name: Text Generation
|
87 |
+
dataset:
|
88 |
+
name: TruthfulQA (0-shot)
|
89 |
+
type: truthful_qa
|
90 |
+
config: multiple_choice
|
91 |
+
split: validation
|
92 |
+
args:
|
93 |
+
num_few_shot: 0
|
94 |
+
metrics:
|
95 |
+
- type: mc2
|
96 |
+
value: 57.09
|
97 |
+
source:
|
98 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Yuma42/KangalKhan-RawRuby-7B
|
99 |
+
name: Open LLM Leaderboard
|
100 |
+
- task:
|
101 |
+
type: text-generation
|
102 |
+
name: Text Generation
|
103 |
+
dataset:
|
104 |
+
name: Winogrande (5-shot)
|
105 |
+
type: winogrande
|
106 |
+
config: winogrande_xl
|
107 |
+
split: validation
|
108 |
+
args:
|
109 |
+
num_few_shot: 5
|
110 |
+
metrics:
|
111 |
+
- type: acc
|
112 |
+
value: 78.69
|
113 |
+
name: accuracy
|
114 |
+
source:
|
115 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Yuma42/KangalKhan-RawRuby-7B
|
116 |
+
name: Open LLM Leaderboard
|
117 |
+
- task:
|
118 |
+
type: text-generation
|
119 |
+
name: Text Generation
|
120 |
+
dataset:
|
121 |
+
name: GSM8k (5-shot)
|
122 |
+
type: gsm8k
|
123 |
+
config: main
|
124 |
+
split: test
|
125 |
+
args:
|
126 |
+
num_few_shot: 5
|
127 |
+
metrics:
|
128 |
+
- type: acc
|
129 |
+
value: 62.02
|
130 |
+
name: accuracy
|
131 |
+
source:
|
132 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Yuma42/KangalKhan-RawRuby-7B
|
133 |
+
name: Open LLM Leaderboard
|
134 |
+
---
|
135 |
+
|
136 |
+
# KangalKhan-RawRuby-7B
|
137 |
+
|
138 |
+
I suggest using ChatML (Use whatever system prompt you like, this is just an example!):
|
139 |
+
```
|
140 |
+
<|im_start|>system
|
141 |
+
You are a friendly assistant.<|im_end|>
|
142 |
+
<|im_start|>user
|
143 |
+
Hello, what are you?<|im_end|>
|
144 |
+
<|im_start|>assistant
|
145 |
+
I am an AI language model designed to assist users with information and answer their questions. How can I help you today?<|im_end|>
|
146 |
+
```
|
147 |
+
|
148 |
+
|
149 |
+
Q4_K_S GGUF:
|
150 |
+
https://huggingface.co/Yuma42/KangalKhan-RawRuby-7B-GGUF
|
151 |
+
|
152 |
+
More GGUF variants by [mradermacher](https://huggingface.co/mradermacher):
|
153 |
+
WARNING: I have observed that these versions output typos in rare cases. If you have the same problem, use my Q4_K_S GGUF above.
|
154 |
+
https://huggingface.co/mradermacher/KangalKhan-RawRuby-7B-GGUF
|
155 |
+
weighted/imatrix GGUF by [mradermacher](https://huggingface.co/mradermacher):
|
156 |
+
https://huggingface.co/mradermacher/KangalKhan-RawRuby-7B-i1-GGUF
|
157 |
+
|
158 |
+
|
159 |
+
|
160 |
+
KangalKhan-RawRuby-7B is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
|
161 |
+
* [Yuma42/KangalKhan-Ruby-7B-Fixed](https://huggingface.co/Yuma42/KangalKhan-Ruby-7B-Fixed)
|
162 |
+
* [Yuma42/KangalKhan-RawEmerald-7B](https://huggingface.co/Yuma42/KangalKhan-RawRuby-7B)
|
163 |
+
|
164 |
+
## 🧩 Configuration
|
165 |
+
|
166 |
+
```yaml
|
167 |
+
slices:
|
168 |
+
- sources:
|
169 |
+
- model: Yuma42/KangalKhan-Ruby-7B-Fixed
|
170 |
+
layer_range: [0, 32]
|
171 |
+
- model: Yuma42/KangalKhan-RawEmerald-7B
|
172 |
+
layer_range: [0, 32]
|
173 |
+
merge_method: slerp
|
174 |
+
base_model: Yuma42/KangalKhan-Ruby-7B-Fixed
|
175 |
+
parameters:
|
176 |
+
t:
|
177 |
+
- filter: self_attn
|
178 |
+
value: [0.1, 0.55, 0.35, 0.75, 0.97]
|
179 |
+
- filter: mlp
|
180 |
+
value: [0.9, 0.45, 0.65, 0.25, 0.03]
|
181 |
+
- value: 0.5
|
182 |
+
dtype: bfloat16
|
183 |
+
```
|
184 |
+
|
185 |
+
## 💻 Usage
|
186 |
+
|
187 |
+
```python
|
188 |
+
!pip install -qU transformers accelerate
|
189 |
+
|
190 |
+
from transformers import AutoTokenizer
|
191 |
+
import transformers
|
192 |
+
import torch
|
193 |
+
|
194 |
+
model = "Yuma42/KangalKhan-RawRuby-7B"
|
195 |
+
messages = [{"role": "user", "content": "What is a large language model?"}]
|
196 |
+
|
197 |
+
tokenizer = AutoTokenizer.from_pretrained(model)
|
198 |
+
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
199 |
+
pipeline = transformers.pipeline(
|
200 |
+
"text-generation",
|
201 |
+
model=model,
|
202 |
+
torch_dtype=torch.float16,
|
203 |
+
device_map="auto",
|
204 |
+
)
|
205 |
+
|
206 |
+
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
207 |
+
print(outputs[0]["generated_text"])
|
208 |
+
```
|
209 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
210 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Yuma42__KangalKhan-RawRuby-7B)
|
211 |
+
|
212 |
+
| Metric |Value|
|
213 |
+
|---------------------------------|----:|
|
214 |
+
|Avg. |68.95|
|
215 |
+
|AI2 Reasoning Challenge (25-Shot)|66.89|
|
216 |
+
|HellaSwag (10-Shot) |85.53|
|
217 |
+
|MMLU (5-Shot) |63.46|
|
218 |
+
|TruthfulQA (0-shot) |57.09|
|
219 |
+
|Winogrande (5-shot) |78.69|
|
220 |
+
|GSM8k (5-shot) |62.02|
|
221 |
+
|
222 |
+
|
223 |
+
|