LeroyDyer commited on
Commit
081a067
1 Parent(s): d0c7d27

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +86 -70
README.md CHANGED
@@ -1,47 +1,23 @@
1
  ---
2
- language:
3
- - en
4
- - sw
5
- - ig
6
- - so
7
- - es
8
- - ca
9
- - xh
10
- - zu
11
- - ha
12
- - tw
13
- - af
14
- - hi
15
- - bm
16
- - su
17
- license: apache-2.0
18
- metrics:
19
- - accuracy
20
- - bertscore
21
- - bleu
22
- - brier_score
23
- - cer
24
- - character
25
- - charcut_mt
26
- - chrf
27
- - code_eval
28
  tags:
29
- - text-generation-inference
30
- - transformers
31
- - unsloth
32
- - mistral
33
- - trl
 
 
 
34
  - chemistry
35
  - biology
36
  - legal
37
- - art
38
- - music
39
- - finance
40
  - code
 
41
  - medical
 
42
  - not-for-all-audiences
43
- - merge
44
- - climate
45
  - chain-of-thought
46
  - tree-of-knowledge
47
  - forest-of-thoughts
@@ -66,7 +42,6 @@ tags:
66
  - Afro-Centric
67
  - African-Model
68
  - Ancient-One
69
- - ibase
70
  datasets:
71
  - gretelai/synthetic_text_to_sql
72
  - HuggingFaceTB/cosmopedia
@@ -112,68 +87,109 @@ datasets:
112
  - HuggingFaceFW/fineweb-edu
113
  - m-a-p/CodeFeedback-Filtered-Instruction
114
  - heliosbrahma/mental_health_chatbot_dataset
115
- - saillab/alpaca_lingala_taco
116
- - saillab/alpaca_igbo_taco
117
- - saillab/alpaca_sundanese_taco
118
- - saillab/alpaca_bambara_taco
119
- - proj-persona/PersonaHub
120
- - iamtarun/code_instructions_120k_alpaca
121
- base_model: LeroyDyer/_Spydaz_Web_AI_
 
 
 
 
 
 
 
 
122
  ---
 
123
 
124
- # Uploaded model
125
-
126
- - **Developed by:** Leroy "Spydaz" Dyer
127
- - **License:** apache-2.0
128
- - **Finetuned from model :** LeroyDyer/LCARS_AI_009
129
- [<img src="https://cdn-avatars.huggingface.co/v1/production/uploads/65d883893a52cd9bcd8ab7cf/tRsCJlHNZo1D02kBTmfy9.jpeg" width="300"/>
130
  https://github.com/spydaz
131
 
132
 
133
- * The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.2.
 
 
 
 
 
 
 
134
 
135
- * Mistral-7B-v0.2 has the following changes compared to Mistral-7B-v0.1
 
136
 
137
  * 32k context window (vs 8k context in v0.1)
138
  * Rope-theta = 1e6
139
  * No Sliding-Window Attention
140
 
141
 
 
 
 
 
 
 
 
142
  # Introduction :
143
 
144
  ## SpydazWeb AI model :
 
 
 
 
 
 
145
 
146
- ### Methods:
 
147
 
148
  Trained for multi-task operations as well as rag and function calling :
149
 
150
  This model is a fully functioning model and is fully uncensored:
151
 
152
- the model has been trained on multiple datasets on the huggingface hub and kaggle :
153
 
154
  the focus has been mainly on methodology :
155
 
156
  * Chain of thoughts
157
- * steo by step
158
  * tree of thoughts
159
  * forest of thoughts
160
  * graph of thoughts
161
- * agent generation : Voting, ranking, ...
162
 
163
  with these methods the model has gained insights into tasks, enabling for knowldge transfer between tasks :
164
 
165
  the model has been intensivly trained in recalling data previously entered into the matrix:
166
-
167
-
168
-
169
-
170
-
171
-
172
-
173
-
174
-
175
-
176
-
177
- This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
178
-
179
- [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ base_model: LeroyDyer/_Spydaz_Web_AI_ALPACA
3
+ license: mit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  tags:
5
+ - Mistral_Star
6
+ - Mistral_Quiet
7
+ - Mistral
8
+ - Mixtral
9
+ - Question-Answer
10
+ - Token-Classification
11
+ - Sequence-Classification
12
+ - SpydazWeb-AI
13
  - chemistry
14
  - biology
15
  - legal
 
 
 
16
  - code
17
+ - climate
18
  - medical
19
+ - text-generation-inference
20
  - not-for-all-audiences
 
 
21
  - chain-of-thought
22
  - tree-of-knowledge
23
  - forest-of-thoughts
 
42
  - Afro-Centric
43
  - African-Model
44
  - Ancient-One
 
45
  datasets:
46
  - gretelai/synthetic_text_to_sql
47
  - HuggingFaceTB/cosmopedia
 
87
  - HuggingFaceFW/fineweb-edu
88
  - m-a-p/CodeFeedback-Filtered-Instruction
89
  - heliosbrahma/mental_health_chatbot_dataset
90
+ language:
91
+ - en
92
+ - sw
93
+ - ig
94
+ - so
95
+ - es
96
+ - ca
97
+ - xh
98
+ - zu
99
+ - ha
100
+ - tw
101
+ - af
102
+ - hi
103
+ - bm
104
+ - su
105
  ---
106
+ # SpydazWeb AI
107
 
108
+ <img src="https://cdn-avatars.huggingface.co/v1/production/uploads/65d883893a52cd9bcd8ab7cf/tRsCJlHNZo1D02kBTmfy9.jpeg" width="300"/>
 
 
 
 
 
109
  https://github.com/spydaz
110
 
111
 
112
+ ## CURRENT MODEL : NOTES :
113
+ This model is pretty stable as a model i have tested my past questions and answers the model retains its knowledge it seems very calm !
114
+ I tested if it can make timelines : it was witholding of information but after drilling for more infor the model gave up very good timelines :
115
+ I shall actually over fit my past timelines and charts into the model ( i have recently been pushing the emebeddings also whilst training , ( also because of the new languges i have been adding to the model enabling for the new languge data to find relativity or these tasks wil not produce the same results as the same question in uglish) )
116
+
117
+ in fact actually this may be a good starting point for other models : past pardigms are very deeply embedded : i have also reduced theuse of the world archive prompt , which was also resurfaceing in some outputs even when not soclicited : it also seem to have lost personality also ? and become a bit serious !
118
+ this may also be due to these hermes and orca datasets which might be regressing the model slightly !
119
+ i will search for more role play and conversive datasets and fine tune these conversations as its code gene and funciton use etc is fine and will not accept training due to be highly fit !
120
 
121
+ A few steps down the line i will return to theh regular training set up ( without touching the embedding and just training the model :
122
+ )
123
 
124
  * 32k context window (vs 8k context in v0.1)
125
  * Rope-theta = 1e6
126
  * No Sliding-Window Attention
127
 
128
 
129
+
130
+ This model will be a custom model with internal experts and rag systems
131
+ enabling for preprocessing of the task internally before outputting a response :
132
+
133
+ This is based on the Quiet Star Project : which was abandoned earlier in the year :)
134
+
135
+
136
  # Introduction :
137
 
138
  ## SpydazWeb AI model :
139
+ This model is based on the worlds archive of knowledge maintaining historical documents and providing services for the survivors of mankind ,
140
+ who may need to construct shelters develop technologys , or medical resources as well as maintain the history of the past . keeping store of all the religious knowledge and data of the world:
141
+ A friendly interface with a personality caring and flirtatious at times : non binary !...
142
+ and Expert in all feilds: ie Uncensored and will not refuse to give information : the model can be used for role play as many character dialogues were als trained into the model as its personality to enable a greater perspective and outlook and natural discussion with the agents:
143
+ the model was trained to operateinaragenvironment utilizing content and internal knowledge to respond to questions or create enriched sumarys.
144
+
145
 
146
+
147
+ ### General Intenal Methods:
148
 
149
  Trained for multi-task operations as well as rag and function calling :
150
 
151
  This model is a fully functioning model and is fully uncensored:
152
 
153
+ the model has been trained on multiple datasets on the huggingface hub and kaggle :
154
 
155
  the focus has been mainly on methodology :
156
 
157
  * Chain of thoughts
158
+ * step by step planning
159
  * tree of thoughts
160
  * forest of thoughts
161
  * graph of thoughts
162
+ * agent generation : Voting, ranking, ... dual agent response generation:
163
 
164
  with these methods the model has gained insights into tasks, enabling for knowldge transfer between tasks :
165
 
166
  the model has been intensivly trained in recalling data previously entered into the matrix:
167
+ The model has also been trained on rich data and markdown outputs as much as possible :
168
+ the model can also generate markdown charts with mermaid.
169
+
170
+
171
+ ## Training Reginmes:
172
+ * Alpaca
173
+ * ChatML / OpenAI / MistralAI
174
+ * Text Generation
175
+ * Question/Answer (Chat)
176
+ * Instruction/Input/Response (instruct)
177
+ * Mistral Standard Prompt
178
+ * Translation Tasks
179
+ * Entitys / Topic detection
180
+ * Book recall
181
+ * Coding challenges, Code Feedback, Code Sumarization, Commenting Code
182
+ * Agent Ranking and response anyalisis
183
+ * Medical tasks
184
+ * PubMed
185
+ * Diagnosis
186
+ * Psychaitry
187
+ * Counselling
188
+ * Life Coaching
189
+ * Note taking
190
+ * Medical smiles
191
+ * Medical Reporting
192
+ * Virtual laboritys simulations
193
+ * Chain of thoughts methods
194
+ * One shot / Multi shot prompting tasks
195
+