Update README.md
Browse files
README.md
CHANGED
@@ -1,11 +1,82 @@
|
|
1 |
---
|
2 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
datasets:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
- xz56/react-llama
|
5 |
- BeIR/hotpotqa
|
6 |
-
-
|
7 |
-
|
8 |
-
-
|
|
|
|
|
|
|
9 |
---
|
10 |
# SpydazWeb AI React Project
|
11 |
Quote for Motivation:
|
@@ -21,6 +92,54 @@ The SpydazWeb AI React Project was initiated to build advanced AI agents capable
|
|
21 |
|
22 |
## Training Methodology:
|
23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
24 |
### Foundation Building:
|
25 |
|
26 |
The initial phase involved training the model on binary yes/no questions without any explicit methodology. This was crucial in establishing a baseline for the model’s decision-making capabilities.
|
|
|
1 |
---
|
2 |
+
base_model:
|
3 |
+
- LeroyDyer/SpydazWeb_AI_CyberTron_Ultra_7b
|
4 |
+
- LeroyDyer/LCARS_AI_StarTrek_Computer
|
5 |
+
- LeroyDyer/_Spydaz_Web_AI_ActionQA_Project
|
6 |
+
- LeroyDyer/_Spydaz_Web_AI_ChatML_512K_Project
|
7 |
+
- LeroyDyer/SpyazWeb_AI_DeepMind_Project
|
8 |
+
- LeroyDyer/SpydazWeb_AI_Swahili_Project
|
9 |
+
- LeroyDyer/_Spydaz_Web_AI_08
|
10 |
+
- LeroyDyer/_Spydaz_Web_AI_ChatQA_001
|
11 |
+
- LeroyDyer/_Spydaz_Web_AI_ChatQA_001_SFT
|
12 |
+
- LeroyDyer/_Spydaz_Web_AI_ChatQA_003
|
13 |
+
- LeroyDyer/_Spydaz_Web_AI_ChatQA_004
|
14 |
+
- LeroyDyer/_Spydaz_Web_AI_ChatQA_ReAct_Project_UltraFineTuned
|
15 |
+
library_name: transformers
|
16 |
+
language:
|
17 |
+
- en
|
18 |
+
- sw
|
19 |
+
- ig
|
20 |
+
- so
|
21 |
+
- es
|
22 |
+
- ca
|
23 |
+
- xh
|
24 |
+
- zu
|
25 |
+
- ha
|
26 |
+
- tw
|
27 |
+
- af
|
28 |
+
- hi
|
29 |
+
- bm
|
30 |
+
- su
|
31 |
datasets:
|
32 |
+
- gretelai/synthetic_text_to_sql
|
33 |
+
- HuggingFaceTB/cosmopedia
|
34 |
+
- teknium/OpenHermes-2.5
|
35 |
+
- Open-Orca/SlimOrca
|
36 |
+
- Open-Orca/OpenOrca
|
37 |
+
- cognitivecomputations/dolphin-coder
|
38 |
+
- databricks/databricks-dolly-15k
|
39 |
+
- yahma/alpaca-cleaned
|
40 |
+
- uonlp/CulturaX
|
41 |
+
- mwitiderrick/SwahiliPlatypus
|
42 |
+
- swahili
|
43 |
+
- Rogendo/English-Swahili-Sentence-Pairs
|
44 |
+
- ise-uiuc/Magicoder-Evol-Instruct-110K
|
45 |
+
- meta-math/MetaMathQA
|
46 |
+
- abacusai/ARC_DPO_FewShot
|
47 |
+
- abacusai/MetaMath_DPO_FewShot
|
48 |
+
- abacusai/HellaSwag_DPO_FewShot
|
49 |
+
- HaltiaAI/Her-The-Movie-Samantha-and-Theodore-Dataset
|
50 |
+
- HuggingFaceFW/fineweb
|
51 |
+
- occiglot/occiglot-fineweb-v0.5
|
52 |
+
- omi-health/medical-dialogue-to-soap-summary
|
53 |
+
- keivalya/MedQuad-MedicalQnADataset
|
54 |
+
- ruslanmv/ai-medical-dataset
|
55 |
+
- Shekswess/medical_llama3_instruct_dataset_short
|
56 |
+
- ShenRuililin/MedicalQnA
|
57 |
+
- virattt/financial-qa-10K
|
58 |
+
- PatronusAI/financebench
|
59 |
+
- takala/financial_phrasebank
|
60 |
+
- Replete-AI/code_bagel
|
61 |
+
- athirdpath/DPO_Pairs-Roleplay-Alpaca-NSFW
|
62 |
+
- IlyaGusev/gpt_roleplay_realm
|
63 |
+
- rickRossie/bluemoon_roleplay_chat_data_300k_messages
|
64 |
+
- jtatman/hypnosis_dataset
|
65 |
+
- Hypersniper/philosophy_dialogue
|
66 |
+
- Locutusque/function-calling-chatml
|
67 |
+
- bible-nlp/biblenlp-corpus
|
68 |
+
- DatadudeDev/Bible
|
69 |
+
- Helsinki-NLP/bible_para
|
70 |
+
- HausaNLP/AfriSenti-Twitter
|
71 |
+
- aixsatoshi/Chat-with-cosmopedia
|
72 |
- xz56/react-llama
|
73 |
- BeIR/hotpotqa
|
74 |
+
- YBXL/medical_book_train_filtered
|
75 |
+
- SkunkworksAI/reasoning-0.01
|
76 |
+
- THUDM/LongWriter-6k
|
77 |
+
- WhiteRabbitNeo/WRN-Chapter-1
|
78 |
+
- WhiteRabbitNeo/Code-Functions-Level-Cyber
|
79 |
+
- WhiteRabbitNeo/Code-Functions-Level-General
|
80 |
---
|
81 |
# SpydazWeb AI React Project
|
82 |
Quote for Motivation:
|
|
|
92 |
|
93 |
## Training Methodology:
|
94 |
|
95 |
+
|
96 |
+
# NEW PROMPTING METHOD DEPLOYED :
|
97 |
+
I have discovered that its possiblle to generatre graphs on the fly internallly within a model with a simple Prompt :slight_smile:
|
98 |
+
|
99 |
+
here is an example in which i invoke the ReaCt Prompt Loop !
|
100 |
+
```yal
|
101 |
+
1. **Question**: {Insert user question here}
|
102 |
+
2. **Thought**: Think step by step about how to approach this question.
|
103 |
+
3. **Action**: Determine what action to take next:
|
104 |
+
- [Search]: Look for relevant information online.
|
105 |
+
- [Analyze]: Break down the problem into smaller parts.
|
106 |
+
- [Summarize]: Provide a summary of known facts related to the question.
|
107 |
+
4. **Action Input**: Specify any details needed for the action.
|
108 |
+
5. **Observation**: Describe what was found or learned from the action taken.
|
109 |
+
|
110 |
+
Repeat steps 2-5 as necessary to refine your answer.
|
111 |
+
|
112 |
+
6. **Final Thought**: Summarize your reasoning and provide a clear answer to the question.
|
113 |
+
```
|
114 |
+
|
115 |
+
In this prompt you will note an inner prompt !
|
116 |
+
this is the prompt within the action !
|
117 |
+
here we can state a methodology ad even a loop , so we can deploy a refiner in the loop or even a tester component : like so !
|
118 |
+
```yaml
|
119 |
+
1. **Question**: {Insert user question here}
|
120 |
+
2. **Thought**: Think step by step about how to approach this question.
|
121 |
+
3. **Action**: Determine what action to take next:
|
122 |
+
- [Plan]: Create a plan or methodolgy for the task , select from known methods if avaliable first.
|
123 |
+
- [Test]: Break down the problem into smaller parts testing each step befor moveing to the next:
|
124 |
+
- [Act]: Provide a summary of known facts related to the question. generate full answere from sucessfull steps :
|
125 |
+
4. **Action Input**: Specify any details needed for the action.
|
126 |
+
5. **Observation**: Describe what was found or learned from the action taken.
|
127 |
+
|
128 |
+
Repeat steps 2-5 as necessary to refine your answer.
|
129 |
+
|
130 |
+
6. **Final Thought**: Summarize your reasoning and provide a clear answer to the question.
|
131 |
+
```
|
132 |
+
# TRSAINING METHODS:
|
133 |
+
### STEP 1
|
134 |
+
I often train the model with 500 samples MyTrainSet ! First !::: Then i overfit these samples ! to 0.001 :Here i train all heads ! and Gates! Sometimes if the task is not accepting is switch to lm-head
|
135 |
+
### STEP 2
|
136 |
+
Then i use the next bulk 5000 samples ! just to add Mass examples until they diverges to 0.3/4 : here i only trai the ATTENTION HEADS ONLY:
|
137 |
+
this give the modle the attention required to solve all future tasks
|
138 |
+
i dont not add more samples than 10k of a single task !
|
139 |
+
### ONGOING !
|
140 |
+
so i keep taining these samples while i train other methods : to keep these training sets in place ! ie the model stays aligned to the old sets while training the new: so you will see i used over 48 datasets they dont seem to be changing but they are :
|
141 |
+
|
142 |
+
# Past Method :
|
143 |
### Foundation Building:
|
144 |
|
145 |
The initial phase involved training the model on binary yes/no questions without any explicit methodology. This was crucial in establishing a baseline for the model’s decision-making capabilities.
|