So lately the idea has been to develop a sleep inducing podcast, called “The Daily Yawn”. Where a narrator reads rather dry material to induce that same coma we all experienced in high-school math classes. Here’s a sample episode (beta version): Video set to premiere on YouTube 9-3-2024. Link will be embedded shortly.
So the black box problem of NLP models is beginning to be probed here, with some interesting results. This post is simply to store Notes on this paper and a video by bycloud interpreting the interpretability paper. Paper: https://transformer-circuits.pub/2024/scaling-monosemanticity/index.html Video:
Some notes and AI texts generated for organizing some training for an IT role. llama-3-1028k-gradient LLM: (RESPONSE) Hi! I’d be happy to assist you in your career development journey. It’s great that you’re taking proactive steps towards upskilling yourself for the role you want. Here’s a suggested path forward: Step 1: Get familiar with the
Just a reference note for this blender project.
The concept of embeddings (See [1]) utilizes vectors, hyperdimensional space, matrix algebra and ultimately hardware architecture to embody the fully realized mathematical object nature of a vocabulary of words and their semantic meaning. The use of these mathematical objects, the vector class, allows for similarity evaluation at a higher level of information abstraction than simple
So I decided I wanted to build a chat bot, specifically for Oxford House; which will allow natural language queries concerning the rules and data associated with the particular house and organization in my locale. After looking into this a bit, and experimenting with some more naive implementations using selenium and other automation solutions, I
This is a post to collect notes and musings on Jupter Notebook and Lab. The issue with aligning the ipyKernel and the correct python binary of the virtual environment is discussed in the following post:
So, after building an intersting scraping algorithm, I need to do things like remove extremely similar or identical sentences repeated in close proximity, remove captured code elements, and other non-alphanumeric noise. To that end here are some notes on research: Okay, so after experimenting with some AI assisted implementations of BERT embeddings and K-Means clustering,
Proceeding to implement and test python based AI projects from github has necessitated some seriously finicky python version dependency management. Initially I was using venv however, this became unwieldy in some cases and am now experimenting with miniconda3. A shell script and symbolic link were created to facilitate running the conda.sh script with the command
So, I decided I wanted to have a local installation of TriplePlaneGaussian from github for research and experimentation. Ran into version conflicts, as usual; so I created a venv (python virtual environment). No big deal so far. Well, when going to install the appropriate versions of pyTorch and associated packages, (1.13.1+cu113); I couldn’t find them