Dusko

ukraine conflict biggest impact is that people are now star struck about comedians (not since Charlie Chaplin). ukraine is seeking a serious amount of stealth bombers to get a chance in the conflict- but trumps only answer is strengthening own forces with the new 6th gen fighter jet plans.

since i've started chatting with chatGPT this was a goal- a AI-agent which can generate and publish content generated by a ML powered (positive and negative keywords) web-spider, input with urls are from Travily, crawl of the filtered urls with nutch, index store in solr, content is created by ollamas llama3.1 and qwen2.5-coder, this is the...

making agent which can manage contacts, calendar (multiple entries at once), write emails and blog posts (https://thestammtischcom.wordpress.com/) is getting simpler- only the imaginations are the limits... i'm also testing some visual content generating LLMs- rave JLO is banging on the dance floor...   #dichipcoin

-bought two p40s -testing apps flowise, n8n, aythingllm, open webui, langflow -the interesting aspect is that i've tested pythonapi (fastapi) calling from inside a n8n node...that inspires me a lot- i've searched for this type of functionality, cause it make coding simpler. -flowise can use faiss out of the box

https://www.tiktok.com/@huggyattack/video/7367979440548187435 jlos shooting star was an offspring of a kpop group. since the script is doing well and the both tesla m40s are heating up too much, i've purchaised two p40s. to get executable scripts isn't that easy and there are no free llms which can make code refraction, see https://aider....

some content about plans about family...and having a daisy duck... and being confused sitting on too hard balls...getting softer balls is more comfortable

the measurement avoid to be on a loose the classifier managment of solid liquid neural network is giving a notebook to a llm to make it to a semi-liquid neural network. making notes make sense for sam since "early 2024".

i've achieved the next milestone with the autocoder-script when creating a tic-tac-toe game. this are the LLMs: "OLLAMA_API_MODEL = 'deepseek-coder:33b-instruct-fp16'OLLAMA_HELPER_API_MODEL = 'phi3:14b-medium-128k-instruct-q8_0'OLLAMA_API_URL = 'http://localhost:11434/api'coder_LLM = 'dolphincoder:15b-starcoder2-fp16'coder_LLM_2 =...

how to save the world from pollution and not affordable energy-consumption. mr. gates gives a bit of a push to the thin bone african soup... and is the gainer of the commercialsation of openai.

#dichipcoin the swarm-coder is tested and working. i'm working on this for months and have tried a lot of improvements while learning by doing. implementing memories and web-search for a LLM is crucial. beyond this i'm experimenting with ML BoW models creating a pool of classifiers (vectorizers) and managing these. the magic code results...

i've started a journey into the world of AI and coding (with chatGPT), with the goal to make some scirpts on github and to learn about ML and AI. now i have a Swarm-Coder, RAG and a Crawler... who could expect that- i got even a call from an agency which is lunching start-ups into trading markets...a call just for info ;-) the swarm-coder has...

the swarm coder is doing well- sometimes almost perfect. it seems there is a huge variation in the created outputs depending on LLMs point of view. if the point of view which is enabling the advanced AI can be somehow defined- advanced AI-workloads can be achieved. i've recognized memories (with chromadb) give a huge advantage for the LLM.

my project Auto-coder with chatGPT (i'm new with python-coding and AI, so chatGPT is doing the coding for me) has succeeded. basically all the scripts i was interested in and have discuss with chatGPT (because from the orig. source not working) have been successfully rearranged together with chatGPT. what's next?

i've just released a multi-agents AutoCoder app on github the goal is to have multiple AIs working in team together to solve problems (f.e. coding, maintenance,...)

i did a nutch web-crawl and had a chat with a local LLM (AI) about it. here are some used scripts (developed witch chatGPT): https://github.com/DuskoPre Please enter your question (or type 'exit' to end): what is Nutch? Apache Nutch is an open-source web crawling and indexing tool. It is designed to automatically download, parse, and index...

March 20, 2024 05:50 PMLocA01: To create a new workstation for 70B LLVM inference, we'll need to consider the following components: GPU: Given the size of the model (70B), a high-end NVIDIA Ampere architecture GPU like the NVIDIA A100 or NVIDIA GeForce RTX 3090 is recommended for fast inference performance. You can find these GPUs on websites...