LeroyDyer/Mixtral_AI_LCARS_MA

A model is only as good as its power to think ! not parameters!

the importance of thoughts and thinking processes methodolgies:

how do we calculate 4+4 << these are mental arithmatic and can be taught to be remembered and not even calculated no more , and all aths are a product of these mental arthmatics: So we train the model to perform simple maths: and then show the model how to do simple algebra and understand natuaral language mental arithmatic: (we do not need external calculators) .. so for complex math we need a train of thought and a method: so we train the model to perform these methods as a chain of thoughts thinking step by step to define the logic for the operation and solve ... and output the answer (without explanation) (and with explanation)<<a now we have a model which can generaalize math simple math problems and even ore detailed math querys: now we need more complexity so we teach the model the process for performing algebra... and probabilties, fractions and other simple school maths: we know now that all maths are chains so a problem broken into steps and each step solved and chained we can solve a problem ... as well as use the chain rules: so as we teach the odel we notice that it can perform calculations which we did not forsee as wel as answer questions we did not program ... infact the model learned associated patterrns: So we teach the model some maths corpuses in which they discus mathmatics and other algorithms and when we query the model we find it can perform many of these functions with its step be step thinking...but: we need creativity to solve some problems! and estimation: and even lookup tables for logs etc... so we need a function .. so we train the model to use a toll to perform the task and return the result but this is a down fall , but we only want the model to get used ot performing this function: after all its acomputer/neural net! why can it not figure out the cosine ? so we write a function in vb.net to perform this magic , wwe also create the fucntion in multiple languages: we keep this function in the thoughts section of the prompt ... and we teach the model some examples of using this internal python function, alternating it with the ither versions of the functions: this allows the model to create code in its thought to perform a task: so we can create functions to perform formulae that we know , such as PIr2 area of a circle and recall these in the thought to perform this task... (this is with out prompting) and now the model will generate any function to perform such tasks we begin to install or keep in our prompts! and still more : there are multiple ways to perform tasks and in some Chians we have multiple agents allowing for an opinion or enrichment of the response to be generated: currently this is performed with lang chaining etc: but we can also simulate this internally : so : in our prompts and thougths we can generate , Experts to answrr the questions, ... so with a CLEAN dataset with no refusals for a Specifi topic: we can call the expert internally and the model can discuss with agents the steps and function required to be generarted and colaberate on a tak before producing the output: infact this is just chaining different methodolgys together and creating another optional thought chain ... so current methodolgys and prompts used in chains to clean up or format data can be used internally if examples are created and trained : So this form of training allows for a model to actually become an expert in its field of speciality! ... as these chains and thoughts are often unique to topics such as code building aplication design data management , medical , role play etc : Hence this being my Personal focus!

Enjoyable!! - just Added Mufon Files! - Ufo sightings data!

(will search for more Mufon DATA!) pity false ancient aliens dont share a single peice of Data!

For a Personal AI> I had to reinstall my small flurishes of personality... (despite ocasionally havive to talk instead of work!) I am now begining to understand what needs to be added to make an all round usable model for all aspects as well as incorperating all the latest ideas:

There is nothing a model cannot do in truth as it is a neural network: This model was updated with langchain and Transformer models documentation and some other manuals for various I.T stuff such as pandas etc : so now i can ask the question to create anew model :: (not super great yet as we need more usage examples to enhance the learning absorbed by document dumping: Hence without context the data is reference and used for for content:

Current possiton of this model:

This MasterArchive has been trained on multiple datasets and faceted response types from templated functions to calculations and virtual lab creation: and role play sceanrios: This model has been injected with personality: so it has NFSW content as well as rolelay data impressed inside : I have tried to remove all refferences to self as an AI << but there will be reminants of this still: The model has been prepared for funciton calling and can be held on a server and the messages type promting is fine and it will respond acordingly: For standard Prompts: here you can a multi faceted aproach to extracting information or playing games or performing functions or even colaberating on a project together as these workflows have also been added: The model has been trained with chemstry and medical knowledge as well as diagnosis information and illness recognition and cure ; as well as being an adviser and confidant it will play the role of a doctor or teacher as well: it has also been trained on historical events and time periods and various characters from these period have been installed to give diue reflection from the perpesctive of key figures in history: I have added the bible in multiple languages : and still have a lot more sacred texts to install: as a coding partner , i have no faults : as it was primaryly trained to assist in coding problems in languges which i dont use : as well as languges which i do : it can do all the tasks associated with coding as well as i have installed the langchain and various other transformers related librayrs and documentation for roducts such as pandas, sas , etc... as an i.t partner it is also great at solving pc issues and quick funciton creation on open interpretor show the bash / shell scripting / vbscript/javascript / python function paid off! the model has gone back to School and learned all the basic classes such as physics,chemestry,biology,history,mathmatics,computer science,mathmatics , calculus etc enabling for the model to draw upon these text books as well as the associated QA: I have also taken the liberty to also instal various NLP tasks such as sentiment recognition and entity recognition , parts of speech look up.... as well as installing internal mental funciton to perform easy tasks , such as capitalization , preprocessing of text etc, bots are internal and will handle these sub requests: on uploaded many lectures and youtube subtitles as well as a few movie dialogs : as well as coouncelling sessions etc and personal chat (all types as well as phylosphcal discuaion and historical discussions as well as problem solving chains) we can say the internal thinking (from the perspective of the language model has great reasoning and autoatic chains of thoughts) For chain of thoughts , step by step , trees of knowledge , concensous of ideas , intenal colaberations , agent evaluations etc , ... the model has also been trained with these tasks and has [performed very well: For the leader board I have trained these datasets as oneshot and zero shot and few shot : but i also discovered the sub tests which display the lacking in mathmatics and physics etc so i concentrated on these new areas of genrral knowledge and calculation : I have also been able to sucessfully download many cbooks and code into the model and sucessfully recall infformation about these and even ask questions that were previousy hallucenated ! or estimated or guessed!! << Soon hallucenations will be the "Best Guess According to the Data Known!"<<< Halucenation preventor!! (we do not want the model to say it cannot or does not know(it does))!

past Methodology:

many functions such as defining words andnlp task we also added via datsets and very complexed datstructures and prompts : These prompts are removed after training and standard alpaca training given on top:(this enables for the previous highly over fit task to become embedded underneath the previous layer): its important to Change Lora configuration for Embedding layers within the model as well as fine tuning above previous training: Usually i deploy a factor of 8 calcuculation for my loras by this one i chose factor of 9 (9-18/18/36) .... which actually trained so smoothly that i was able to train many different datsets in a signle sitting ; to below 0.9 all varioations of the alpaca prompt ! after testing the was absolutly 0 loss from previous knowledge as well as enhancing some responses and providing comparitive responses for others; I personally use a topK of 1000.... this allows the model to have many choices (this is the context window of results), i put my topP to 0.68(68%).... hence it will select from that percentage of probabiltys... enabling for my temp to be 1 .. therfore it will normalize the selected quartile of next probablity selection enabling for the lower probabiltys to have a scaled chace in being selected : It is important to have a degree of randomness in the respopnse or you will ask the same question and get the same answer ! .... we need varied answer to ome querys and focues for other ? how do we do this ?..... Duplicates!!!!! raising the probability of some information by repetition : as this is how the human learns truth ! truth is that which has been repeated so many times it cannot be disputed! hence some information being absolute and others being transient and constantly updateing: As a predictve model it needs to be ables to have the ability to calculate and predicte and cclassify as wel as recall exact information : hence when utilizing a rag : the conversation history is the dats to be fine tuned into the model as frequent data! as well as producing multiple simular querys to query the rag system for Q/A pairs : also to be updted onto the model : as we are in this development period we are focused on BRAIN cureently .......

HEY HEY HEY

This is just a Marker point before returning to previous tasks and datasets!! ( as we also need to maintain our ground truths by raising thier probabilitys again !!) As Well as history books and refferencing some of the sacred text archives: I was talking to the bot!! << and i was able to build a mini timeline for the bible and for egypt but not able to merge these historys quite as yet .... im stil finding the bible hard to drop in!!<<<a I have low patience! so its still floating around 2......and still incomplete! << its on the way!

Future Works

At this point the model is highly trainabe for any purpose ! << so if overtraining i just retrain with my retrain dataset and samatha chat stuff (it can also be a block because she was not exactly helpfull AI! as well as some NFSW! (keeping the probablitys high)as well as chekcing the key datasets have held: so now it will be prudent to even remerge some old merges!! << using them as loras! hence keeping my model as a base model, and a merge candidate for all segments(Y,X) and merging some model such as orca and hermes and dolphin again (despite thier downfalls) as they have also moved forwards with many corpuses, hence merging them with low bias, and low density, and realigning the models !) for me its like absorbing the models ,...I will begin to focus on coding now and merging some archs , into the model , so the model will have speech to text and image to text.... as standards... enabling for these inputs to produce the required outputs to go into the tokenzed input for embeddings : so a image description can give embeddings as well as a audio trancription: so for these feature extractors its just a matter of bringing the relevant modules into the model and adding the extra inputs , so if detected they will be and inserted into the prompt during embeddings intenally: Using the whisper and the VIT Image transformers! later we can deal with generating sound output and image outputs .... for now we can direct them to a pydantic template !! << ie send a link and the ydantic reciver can open it with thier function , play ofr display acording to thier interface ! ( perhaps even download it and enter it in thier RAG!)