ibuildthings

5 ms·

1.
Bringing Mobius Labs' Aana to Dropbox: multimodal understanding at scale(dropbox.tech)
15 points by ibuildthings 11mo ago|1 comments
2.
Low Bit Triton Kernels [video](youtube.com)
1 points by ibuildthings 2y ago|1 comments
3.
by ibuildthings 2y ago
Talk on optimizing matrix multiplication with Triton kernels, focusing on low-bit processing and efficient quantization for high-performance AI models.
4.
Aana SDK: open-source library for building multimodal AI applications(mobiusml.github.io)
2 points by ibuildthings 2y ago|1 comments
5.
by ibuildthings 2y ago
Aana SDK is an open-source toolkit for building cutting-edge multimodal AI applications: https://github.com/mobiusml/aana_sdk It addresses key challenges in multimodal AI development: - Managing diverse inputs - Scalin
6.
Aanaphi-2 3B(huggingface.co)
2 points by ibuildthings 3y ago|1 comments
7.
by ibuildthings 3y ago
Currently leading in LLM benchmarks among the 3B categories of models
8.
New Mixtral HQQ Quantzied 4-bit/2-bit configuration(huggingface.co)
5 points by ibuildthings 3y ago|1 comments
9.
by ibuildthings 3y ago
We are releasing new 2-bit Mixtral models. These ones use a mixed HQQ 4-bit/2-bit configuration, resulting in a significantly improved model (ppl 4.69 vs. 5.90) with a negligible 0.20 GB VRAM increase. Base: https://huggingf
10.
2-bit and 4-bit versions of Mixtral(huggingface.co)
4 points by ibuildthings 3y ago|1 comments
11.
by ibuildthings 3y ago
We are releasing 2-bit and 4-bit quantized versions of Mixtral utilizing the HQQ method that we just published https://mobiusml.github.io/hqq_blog/ and https://github.com/mobiusml/hqq . The 2-bit v
12.
Half-Quadratic Quantization of Large Machine Learning Models(mobiusml.github.io)
8 points by ibuildthings 3y ago|1 comments
13.
by ibuildthings 3y ago
Sharing our work on model quantization. - Blog: https://mobiusml.github.io/hqq_blog/ - Code: https://github.com/mobiusml/hqq - Models: https://huggingface.co/mobiuslabsgmbh/
14.
by ibuildthings 3y ago
Github repo should be visible now. It is not distilling the model, it is reducing the model weights on the fly and uses LoRA for training/fine-tuning. After the training phase, we explain how to merge the LoRA weights with the pruned w
15.
Low-Rank Pruning of Llama2(mobiusml.github.io)
2 points by ibuildthings 3y ago|3 comments
16.
by ibuildthings 3y ago
I'm sharing a blog post https://mobiusml.github.io/low-rank-llama2/ on our approach to pruning the Llama2 model by leveraging low-rank structures. In a nutshell, we've managed to reduce the model's param
17.
Fine grained human expression analysis using deep learning(blog.mobiuslabs.com)
1 points by ibuildthings 6y ago|0 comments
18.
by ibuildthings 6y ago
In this particular case, why is testing/demonstrating on diverse set of individuals a bad thing ? A personal anecdote is that a few years back the automatic door sensors in my university did not work on my skin tone.
19.
by ibuildthings 8y ago
While agreeing to the general principle, incentive structures are wired quiet differently in academia vs end consumer oriented gig/service industries. Publications ( number, when, where, citations) is the primary currency/value in
20.
by ibuildthings 8y ago
First principles of doing a PhD and taking up an industrial jobs are quite different, which this article sidesteps. I am talking from the perspective of someone who did a PhD, postdoc and migrated to be a founder/CEO. A PhD system trai
21.
by ibuildthings 10y ago
Author here. Just to clarify, motive of this work is to ease curation, with the massive amount of content being created; but by no means an attempt at creativity or originality. The work is not at all contradictory to Adorno, especially in
22.
by ibuildthings 10y ago
The author here. I used the term "understanding", not as in machines understanding the images, but more as scientific attempt in understanding aesthetics. ( <snippet from the text>"empowering me to develop systems for
23.
by ibuildthings 11y ago
I agree with you completely about "bad industry-focused research". It serves no end. My question is that, is it just a reflection of mediocracy and gaming/dishonesty being everywhere, including academica ? In this case, pleas
24.
by ibuildthings 11y ago
Pushing away from color lines is very easy. For negative values of lambda in Eq.1 of the paper, (i.e. reverse the cost for color lines ), the optimization tries to push the puzzle shapes away from the color lines. We had tried a few in this
25.
by ibuildthings 11y ago
One of the authors here. Since it is a optimization, the difficulty can be controlled as a parameter ( the lambda parameter in Eq.1 in the paper ). But you are right, some of the puzzles can be super-hard ( for example, the Seurat puzzle )
26.
by ibuildthings 11y ago
The site has an interesting history. The former Stadtschloss suffered serious destruction during WWII ( https://en.wikipedia.org/wiki/City_Palace,_Berlin ) and the Communist East tore it completely down to build the Pa
27.
by ibuildthings 12y ago
I this is a bit of legacy ( like the Stanford Bunny ). One of the seminal work of lighting was Photon mapping by Henrik Jensen , and he used Ludwig Mies van der Rohe's structure as an example ( http://graphics.ucsd.edu&#x2F
28.
by ibuildthings 12y ago
Throwing in my personal goto reference on gradient descent : http://www.cs.cmu.edu/~quake-papers/painless-conjugate-gradi... "An Introduction to the Conjugate Gradient Method Without the Agonizing Pain" - Jo
29.
EyeEm Acquires Computer Vision Startup Sight.io(techcrunch.com)
7 points by ibuildthings 12y ago|0 comments
30.
by ibuildthings 13y ago
I do empathize with the original article a lot. I used to have/still have a strong fear of failing, especially in intellectual tasks. According to my own introspection this is primary angst that caused/causes me to procrastinate.

More ›