RichardErkhov
commited on
Commit
•
952494e
1
Parent(s):
ae4ce28
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,243 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
RolePlayLake-7B - GGUF
|
11 |
+
- Model creator: https://huggingface.co/fhai50032/
|
12 |
+
- Original model: https://huggingface.co/fhai50032/RolePlayLake-7B/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [RolePlayLake-7B.Q2_K.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q2_K.gguf) | Q2_K | 2.53GB |
|
18 |
+
| [RolePlayLake-7B.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.IQ3_XS.gguf) | IQ3_XS | 2.81GB |
|
19 |
+
| [RolePlayLake-7B.IQ3_S.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.IQ3_S.gguf) | IQ3_S | 2.96GB |
|
20 |
+
| [RolePlayLake-7B.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q3_K_S.gguf) | Q3_K_S | 2.95GB |
|
21 |
+
| [RolePlayLake-7B.IQ3_M.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.IQ3_M.gguf) | IQ3_M | 3.06GB |
|
22 |
+
| [RolePlayLake-7B.Q3_K.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q3_K.gguf) | Q3_K | 3.28GB |
|
23 |
+
| [RolePlayLake-7B.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q3_K_M.gguf) | Q3_K_M | 3.28GB |
|
24 |
+
| [RolePlayLake-7B.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q3_K_L.gguf) | Q3_K_L | 3.56GB |
|
25 |
+
| [RolePlayLake-7B.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.IQ4_XS.gguf) | IQ4_XS | 3.67GB |
|
26 |
+
| [RolePlayLake-7B.Q4_0.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q4_0.gguf) | Q4_0 | 3.83GB |
|
27 |
+
| [RolePlayLake-7B.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.IQ4_NL.gguf) | IQ4_NL | 3.87GB |
|
28 |
+
| [RolePlayLake-7B.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q4_K_S.gguf) | Q4_K_S | 3.86GB |
|
29 |
+
| [RolePlayLake-7B.Q4_K.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q4_K.gguf) | Q4_K | 4.07GB |
|
30 |
+
| [RolePlayLake-7B.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q4_K_M.gguf) | Q4_K_M | 4.07GB |
|
31 |
+
| [RolePlayLake-7B.Q4_1.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q4_1.gguf) | Q4_1 | 4.24GB |
|
32 |
+
| [RolePlayLake-7B.Q5_0.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q5_0.gguf) | Q5_0 | 4.65GB |
|
33 |
+
| [RolePlayLake-7B.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q5_K_S.gguf) | Q5_K_S | 4.65GB |
|
34 |
+
| [RolePlayLake-7B.Q5_K.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q5_K.gguf) | Q5_K | 4.78GB |
|
35 |
+
| [RolePlayLake-7B.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q5_K_M.gguf) | Q5_K_M | 4.78GB |
|
36 |
+
| [RolePlayLake-7B.Q5_1.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q5_1.gguf) | Q5_1 | 5.07GB |
|
37 |
+
| [RolePlayLake-7B.Q6_K.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q6_K.gguf) | Q6_K | 5.53GB |
|
38 |
+
| [RolePlayLake-7B.Q8_0.gguf](https://huggingface.co/RichardErkhov/fhai50032_-_RolePlayLake-7B-gguf/blob/main/RolePlayLake-7B.Q8_0.gguf) | Q8_0 | 7.17GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
license: apache-2.0
|
46 |
+
tags:
|
47 |
+
- merge
|
48 |
+
- mergekit
|
49 |
+
- mistral
|
50 |
+
- SanjiWatsuki/Silicon-Maid-7B
|
51 |
+
- senseable/WestLake-7B-v2
|
52 |
+
base_model:
|
53 |
+
- SanjiWatsuki/Silicon-Maid-7B
|
54 |
+
- senseable/WestLake-7B-v2
|
55 |
+
model-index:
|
56 |
+
- name: RolePlayLake-7B
|
57 |
+
results:
|
58 |
+
- task:
|
59 |
+
type: text-generation
|
60 |
+
name: Text Generation
|
61 |
+
dataset:
|
62 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
63 |
+
type: ai2_arc
|
64 |
+
config: ARC-Challenge
|
65 |
+
split: test
|
66 |
+
args:
|
67 |
+
num_few_shot: 25
|
68 |
+
metrics:
|
69 |
+
- type: acc_norm
|
70 |
+
value: 70.56
|
71 |
+
name: normalized accuracy
|
72 |
+
source:
|
73 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=fhai50032/RolePlayLake-7B
|
74 |
+
name: Open LLM Leaderboard
|
75 |
+
- task:
|
76 |
+
type: text-generation
|
77 |
+
name: Text Generation
|
78 |
+
dataset:
|
79 |
+
name: HellaSwag (10-Shot)
|
80 |
+
type: hellaswag
|
81 |
+
split: validation
|
82 |
+
args:
|
83 |
+
num_few_shot: 10
|
84 |
+
metrics:
|
85 |
+
- type: acc_norm
|
86 |
+
value: 87.42
|
87 |
+
name: normalized accuracy
|
88 |
+
source:
|
89 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=fhai50032/RolePlayLake-7B
|
90 |
+
name: Open LLM Leaderboard
|
91 |
+
- task:
|
92 |
+
type: text-generation
|
93 |
+
name: Text Generation
|
94 |
+
dataset:
|
95 |
+
name: MMLU (5-Shot)
|
96 |
+
type: cais/mmlu
|
97 |
+
config: all
|
98 |
+
split: test
|
99 |
+
args:
|
100 |
+
num_few_shot: 5
|
101 |
+
metrics:
|
102 |
+
- type: acc
|
103 |
+
value: 64.55
|
104 |
+
name: accuracy
|
105 |
+
source:
|
106 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=fhai50032/RolePlayLake-7B
|
107 |
+
name: Open LLM Leaderboard
|
108 |
+
- task:
|
109 |
+
type: text-generation
|
110 |
+
name: Text Generation
|
111 |
+
dataset:
|
112 |
+
name: TruthfulQA (0-shot)
|
113 |
+
type: truthful_qa
|
114 |
+
config: multiple_choice
|
115 |
+
split: validation
|
116 |
+
args:
|
117 |
+
num_few_shot: 0
|
118 |
+
metrics:
|
119 |
+
- type: mc2
|
120 |
+
value: 64.38
|
121 |
+
source:
|
122 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=fhai50032/RolePlayLake-7B
|
123 |
+
name: Open LLM Leaderboard
|
124 |
+
- task:
|
125 |
+
type: text-generation
|
126 |
+
name: Text Generation
|
127 |
+
dataset:
|
128 |
+
name: Winogrande (5-shot)
|
129 |
+
type: winogrande
|
130 |
+
config: winogrande_xl
|
131 |
+
split: validation
|
132 |
+
args:
|
133 |
+
num_few_shot: 5
|
134 |
+
metrics:
|
135 |
+
- type: acc
|
136 |
+
value: 83.27
|
137 |
+
name: accuracy
|
138 |
+
source:
|
139 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=fhai50032/RolePlayLake-7B
|
140 |
+
name: Open LLM Leaderboard
|
141 |
+
- task:
|
142 |
+
type: text-generation
|
143 |
+
name: Text Generation
|
144 |
+
dataset:
|
145 |
+
name: GSM8k (5-shot)
|
146 |
+
type: gsm8k
|
147 |
+
config: main
|
148 |
+
split: test
|
149 |
+
args:
|
150 |
+
num_few_shot: 5
|
151 |
+
metrics:
|
152 |
+
- type: acc
|
153 |
+
value: 65.05
|
154 |
+
name: accuracy
|
155 |
+
source:
|
156 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=fhai50032/RolePlayLake-7B
|
157 |
+
name: Open LLM Leaderboard
|
158 |
+
---
|
159 |
+
|
160 |
+
# RolePlayLake-7B
|
161 |
+
|
162 |
+
RolePlayLake-7B is a merge of the following models :
|
163 |
+
* [SanjiWatsuki/Silicon-Maid-7B](https://huggingface.co/SanjiWatsuki/Silicon-Maid-7B)
|
164 |
+
* [senseable/WestLake-7B-v2](https://huggingface.co/senseable/WestLake-7B-v2)
|
165 |
+
|
166 |
+
|
167 |
+
`In my current testing RolePlayLake is Better than Silicon_Maid in RP and More Uncensored Than WestLake`
|
168 |
+
|
169 |
+
|
170 |
+
|
171 |
+
`I would try to only merge Uncensored Models with Baising towards Chat rather than Instruct `
|
172 |
+
|
173 |
+
|
174 |
+
## 🧩 Configuration
|
175 |
+
|
176 |
+
```yaml
|
177 |
+
slices:
|
178 |
+
- sources:
|
179 |
+
- model: SanjiWatsuki/Silicon-Maid-7B
|
180 |
+
layer_range: [0, 32]
|
181 |
+
- model: senseable/WestLake-7B-v2
|
182 |
+
layer_range: [0, 32]
|
183 |
+
merge_method: slerp
|
184 |
+
base_model: senseable/WestLake-7B-v2
|
185 |
+
parameters:
|
186 |
+
t:
|
187 |
+
- filter: self_attn
|
188 |
+
value: [0, 0.5, 0.3, 0.7, 1]
|
189 |
+
- filter: mlp
|
190 |
+
value: [1, 0.5, 0.7, 0.3, 0]
|
191 |
+
- value: 0.5
|
192 |
+
dtype: bfloat16
|
193 |
+
```
|
194 |
+
|
195 |
+
## 💻 Usage
|
196 |
+
|
197 |
+
```python
|
198 |
+
!pip install -qU transformers accelerate
|
199 |
+
|
200 |
+
from transformers import AutoTokenizer
|
201 |
+
import transformers
|
202 |
+
import torch
|
203 |
+
|
204 |
+
model = "fhai50032/RolePlayLake-7B"
|
205 |
+
messages = [{"role": "user", "content": "What is a large language model?"}]
|
206 |
+
|
207 |
+
tokenizer = AutoTokenizer.from_pretrained(model)
|
208 |
+
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
209 |
+
pipeline = transformers.pipeline(
|
210 |
+
"text-generation",
|
211 |
+
model=model,
|
212 |
+
torch_dtype=torch.float16,
|
213 |
+
device_map="auto",
|
214 |
+
)
|
215 |
+
|
216 |
+
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
217 |
+
print(outputs[0]["generated_text"])
|
218 |
+
```
|
219 |
+
|
220 |
+
|
221 |
+
# Why I Merged WestLake and Silicon Maid
|
222 |
+
|
223 |
+
Merged WestLake and Silicon Maid for a unique blend:
|
224 |
+
1. **EQ-Bench Dominance:** WestLake's 79.75 EQ-Bench score. (Maybe Contaminated)
|
225 |
+
2. **Charm and Role-Play:** Silicon's explicit charm and WestLake's role-play prowess.
|
226 |
+
3. **Config Synergy:** Supports lots of prompt format out of the gate and has a very nice synergy
|
227 |
+
|
228 |
+
Result: RolePlayLake-7B, a linguistic fusion with EQ-Bench supremacy and captivating role-play potential.
|
229 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
230 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_fhai50032__RolePlayLake-7B)
|
231 |
+
|
232 |
+
| Metric |Value|
|
233 |
+
|---------------------------------|----:|
|
234 |
+
|Avg. |72.54|
|
235 |
+
|AI2 Reasoning Challenge (25-Shot)|70.56|
|
236 |
+
|HellaSwag (10-Shot) |87.42|
|
237 |
+
|MMLU (5-Shot) |64.55|
|
238 |
+
|TruthfulQA (0-shot) |64.38|
|
239 |
+
|Winogrande (5-shot) |83.27|
|
240 |
+
|GSM8k (5-shot) |65.05|
|
241 |
+
|
242 |
+
|
243 |
+
|