vykthur

7 ms·

1.
Show HN: I wrote a book on How to build your own agent framework(amazon.com)
2 points by vykthur 10mo ago|0 comments
2.
by vykthur 1y ago
This is pretty cool. I can see it saving users a bit of time. Feedback - it might be worth adding an info button on what DSCR is and how your tool is calculating it (seems like a critical bit to understand to really get the value of your t
3.
by vykthur 1y ago
I tend to agree here. Alot of the use cases are around STDIO and using MCPs within existing apps like Cursor/Windsurf etc. Most developers want to build and integrate their own tools and the complexity required here (build out the ser
4.
by vykthur 1y ago
the blog is hosted on substack which supports feeds. https://newsletter.victordibia.com/feed
5.
by vykthur 1y ago
I tend to agree with this. No, MCP's have NOT Won (Yet) https://newsletter.victordibia.com/p/no-mcps-have-not-won-ye...
6.
by vykthur 1y ago
This a good characterisation of functionality MCP might enable. Thanks. In your opinion, what percentage of apps might benefit from this model where end users bring their own MCP tools to extend the capabilities of your app. What are some g
7.
by vykthur 2y ago
Fully multimodal on the input side: The same model can take in text, images, audio tokens, reason about them, and generate text as outputs. Compact yet powerful: At just 5.8 billion parameters, this model can be optimized for deployment wor
8.
by vykthur 2y ago
True. And that is a step forward. I notice that they make the plan, and THEN write the code in the same forward pass/generation sequence. The challenge here is that all of the incorrect assumptions get "lumped" into this pass
9.
by vykthur 2y ago
I find that it is IMPORTANT to never start these coding sessions with "write X code". Instead, begin with a "open plan" - something the author does allude to (he calls it prompt engineering, I find it also works as the s
10.
by vykthur 3y ago
True. This is an open area of research. Tools like guidance (or other implementations of constrained decoding with llms [1,2]) will likely help improve this problem. [1] A guidance language for controlling large language models. https:&#x2
11.
LIDA – A library (and UI) for automatic generation of visualizations using LLMs(github.com)
5 points by vykthur 3y ago|1 comments
12.
by vykthur 3y ago
LIDA is a library for generating data visualizations and data-faithful infographics. LIDA is grammar agnostic (will work with any programming language and visualization libraries e.g. matplotlib, seaborn, altair, d3 etc) and works with mult
13.
by vykthur 4y ago
https://github.com/victordibia/peacasso Peacasso is a (yet another but really nice) user interface for interacting with stable diffusion models. The current implementation of Peacasso provides a GUI for basic operation
14.
by vykthur 4y ago
Yes. Captum is a great library and a few of my colleagues have used it with good results in the past. Like you mention, most of the few examples that demonstrate gradient based explanation methods for Huggingface models typically focus on P
15.
Implementing gradient explanations for a HuggingFace text classification model(victordibia.com)
49 points by vykthur 4y ago|5 comments
16.
Short review of the AlphaCode Feb 2022 paper from DeepMind(victordibia.com)
1 points by vykthur 5y ago|0 comments
17.
Top Machine Learning and Design Insights from Google IO 2021(victordibia.com)
3 points by vykthur 5y ago|0 comments
18.
State of Deep Learning for Object Detection – You Should Consider CenterNets(victordibia.com)
1 points by vykthur 5y ago|0 comments
19.
Recent Breakthroughs in AI (Feb 11 Talk by Andrej Karpathy, Socher, Johnson)(victordibia.com)
3 points by vykthur 6y ago|1 comments
20.
by vykthur 6y ago
TLDR; The discussion mostly looked at the novelty of transformer based multimodal models such as CLIP and DALL·E which have both shown interesting results. TLDR - Data is king! Getting better data might be the single biggest bang for buck i
21.
by vykthur 6y ago
Current approach being explored to expand the training dataset to include faces/heads labels such that the model better discriminates this from hands.
22.
by vykthur 6y ago
Hi, handtack.js author here, I am exploring some size and speed optimizations which might help with this (smaller models such as efficientdet, quantization). Will share once something is ready.
23.
by vykthur 6y ago
Author here, this is a known issue and being worked on for handtrack.js 2.0. The face/head errors is likely an artifact of transfer learning (handtrackjs is finetuned on the object detection model trained using the coco object detectio
24.
by vykthur 6y ago
Author here, this is a known issue and being worked on for handtrack.js 2.0. The face/head errors is likely an artifact of transfer learning (handtrackjs is finetuned on the an object detection model trained using the coco object detec