CultriX commited on
Commit
50c6ad3
1 Parent(s): 4b5b1c1

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +151 -0
README.md ADDED
@@ -0,0 +1,151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - moe
5
+ - frankenmoe
6
+ - merge
7
+ - mergekit
8
+ - lazymergekit
9
+ - CultriX/MonaTrix-v4
10
+ - mlabonne/OmniTruthyBeagle-7B-v0
11
+ - CultriX/MoNeuTrix-7B-v1
12
+ - paulml/OmniBeagleSquaredMBX-v3-7B
13
+ base_model:
14
+ - CultriX/MonaTrix-v4
15
+ - mlabonne/OmniTruthyBeagle-7B-v0
16
+ - CultriX/MoNeuTrix-7B-v1
17
+ - paulml/OmniBeagleSquaredMBX-v3-7B
18
+ ---
19
+
20
+ # MoNeuTrix-MoE-4x7B
21
+
22
+ MoNeuTrix-MoE-4x7B is a Mixture of Experts (MoE) made with the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
23
+ * [CultriX/MonaTrix-v4](https://huggingface.co/CultriX/MonaTrix-v4)
24
+ * [mlabonne/OmniTruthyBeagle-7B-v0](https://huggingface.co/mlabonne/OmniTruthyBeagle-7B-v0)
25
+ * [CultriX/MoNeuTrix-7B-v1](https://huggingface.co/CultriX/MoNeuTrix-7B-v1)
26
+ * [paulml/OmniBeagleSquaredMBX-v3-7B](https://huggingface.co/paulml/OmniBeagleSquaredMBX-v3-7B)
27
+
28
+ ## 🧩 Configuration
29
+
30
+ ```yaml
31
+ base_model: "CultriX/MonaTrix-v4"
32
+ dtype: bfloat16
33
+ gate:
34
+ type: "learned"
35
+ temperature: 0.1
36
+ scaling_factor: 10
37
+ experts:
38
+ - source_model: "CultriX/MonaTrix-v4" # Historical Analysis, Geopolitics, and Economic Evaluation
39
+ tasks:
40
+ - "Historical Analysis"
41
+ - "Geopolitical Evaluation"
42
+ - "Economic Insights"
43
+ - "Policy Analysis"
44
+ - "Socio-Economic Impacts"
45
+ - "Geopolitical Analysis"
46
+ - "Cultural Commentary"
47
+ - "Analyze geopolitical"
48
+ - "Analyze historic"
49
+ - "Analyze historical"
50
+ - "Assess the political dynamics of the Cold War and its global impact."
51
+ - "Evaluate the historical significance of the Silk Road in ancient trade."
52
+ exclude_tasks:
53
+ - "Technical Writing"
54
+ - "Mathematical Problem Solving"
55
+ - "Software Development"
56
+ - "Artistic Creation"
57
+ - "Machine Learning Development"
58
+ - "Storywriting"
59
+ - "Character Development"
60
+ - "Roleplaying"
61
+ - "Narrative Creation"
62
+
63
+ - source_model: "mlabonne/OmniTruthyBeagle-7B-v0" # Multilingual Communication and Cultural Insights
64
+ tasks:
65
+ - "Multilingual Communication"
66
+ - "Cultural Insights"
67
+ - "Translation and Interpretation"
68
+ - "Cultural Norms Exploration"
69
+ - "Intercultural Communication Practices"
70
+ - "Describe cultural significance"
71
+ - "Narrate cultural"
72
+ - "Discuss cultural impact"
73
+ exclude_tasks:
74
+ - "Scientific Analysis"
75
+ - "Creative Writing"
76
+ - "Technical Documentation"
77
+ - "Economic Modeling"
78
+ - "Historical Documentation"
79
+ - "Programming"
80
+ - "Algorithm Development"
81
+
82
+ - source_model: "CultriX/MoNeuTrix-7B-v1" # Creative Problem Solving and Innovation
83
+ tasks:
84
+ - "Innovation and Design"
85
+ - "Problem Solving"
86
+ - "Creative Thinking"
87
+ - "Strategic Planning"
88
+ - "Conceptual Design"
89
+ - "Innovation and Design"
90
+ - "Problem Solving"
91
+ - "Compose narrative content or poetry."
92
+ - "Create complex puzzles and games."
93
+ - "Devise strategy"
94
+ exclude_tasks:
95
+ - "Historical Analysis"
96
+ - "Linguistic Translation"
97
+ - "Economic Forecasting"
98
+ - "Geopolitical Analysis"
99
+ - "Cultural Commentary"
100
+ - "Historical Documentation"
101
+ - "Scientific Explanation"
102
+ - "Data Analysis Techniques"
103
+
104
+ - source_model: "paulml/OmniBeagleSquaredMBX-v3-7B" # Scientific and Technical Expertise
105
+ tasks:
106
+ - "Scientific Explanation"
107
+ - "Technical Analysis"
108
+ - "Experimental Design"
109
+ - "Data Analysis Techniques"
110
+ - "Scientific Innovation"
111
+ - "Mathematical Problem Solving"
112
+ - "Algorithm Development"
113
+ - "Programming"
114
+ - "Analyze data"
115
+ - "Analyze statistical data on climate change trends."
116
+ - "Conduct basic data analysis or statistical evaluations."
117
+ exclude_tasks:
118
+ - "Cultural Analysis"
119
+ - "Creative Arts"
120
+ - "Linguistic Challenges"
121
+ - "Political Debating"
122
+ - "Marketing Strategies"
123
+ - "Storywriting"
124
+ - "Character Development"
125
+ - "Roleplaying"
126
+ - "Narrative Creation"
127
+ ```
128
+
129
+ ## 💻 Usage
130
+
131
+ ```python
132
+ !pip install -qU transformers bitsandbytes accelerate
133
+
134
+ from transformers import AutoTokenizer
135
+ import transformers
136
+ import torch
137
+
138
+ model = "CultriX/MoNeuTrix-MoE-4x7B"
139
+
140
+ tokenizer = AutoTokenizer.from_pretrained(model)
141
+ pipeline = transformers.pipeline(
142
+ "text-generation",
143
+ model=model,
144
+ model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
145
+ )
146
+
147
+ messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
148
+ prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
149
+ outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
150
+ print(outputs[0]["generated_text"])
151
+ ```