fgfm

8 ms·

1.
First verifiable quantum experiment by Google [video](youtube.com)
1 points by fgfm 11mo ago|0 comments
2.
Eureka Labs: Andrej Karpathy new venture in education and AI(eurekalabs.ai)
1 points by fgfm 2y ago|0 comments
3.
OpenAI ARR breakdown: API revenue only represents 15%(futuresearch.ai)
2 points by fgfm 2y ago|0 comments
4.
Runway Gen-3 Alpha: temporal consistency coming to video generation(youtube.com)
5 points by fgfm 2y ago|0 comments
5.
Mistral releases the v0.3 of its 7B LLM(huggingface.co)
6 points by fgfm 2y ago|0 comments
6.
Captable – open-source cap table management(github.com)
9 points by fgfm 2y ago|2 comments
7.
Sequoia: Speculative decoding boosting LLM inference by 8-10x(infini-ai-lab.github.io)
3 points by fgfm 3y ago|0 comments
8.
by fgfm 3y ago
It's funny how discoveries in NLP & computer vision complement each other. The replacement of multiplication by additions made me think about the AdderNet paper ( https://arxiv.org/abs/1912.13200 ), which conclu
9.
The Era of 1-bit LLMs: ternary parameters for cost-effective computing(arxiv.org)
1040 points by fgfm 3y ago|447 comments
10.
The Case for Open Source AI(press.airstreet.com)
1 points by fgfm 3y ago|0 comments
11.
AlphaCodium – SOTA approach for code generation(codium.ai)
7 points by fgfm 3y ago|2 comments
12.
Wavecoder – a CodeLLM with 6.7B params scoring just behind GPT4(twitter.com)
36 points by fgfm 3y ago|10 comments
13.
OpenChat-3.5 Update 0106: ChatGPT-level performances accessible locally(huggingface.co)
2 points by fgfm 3y ago|0 comments
14.
CRUXEval: A Benchmark for Code Reasoning, Understanding and Execution(arxiv.org)
1 points by fgfm 3y ago|0 comments
15.
by fgfm 3y ago
That last part feels very relatable to me: I've seen organizations who are mindful of the licenses of tools they use to avoid further problems, and others assuming that because it's closed source the problem won't ever arise.
16.
by fgfm 3y ago
Haha I don't know what your poison is, but the same goes for: - using the syntax of Python 3.11 for asynchronous tasks; - using Promises vs. Observables in Javascript Was the demo example confusing, or not challenging enough perhaps? I
17.
by fgfm 3y ago
Thanks for sharing, that's an interesting social component of the equation. From your comment, I assume you're referring to something I've also encountered as a maintainer: we filter out signals where no efforts were put in.
18.
by fgfm 3y ago
We'll do our best to consistently report it since this can indeed influence the financial decisions of developers, especially if they go through third-party paying LLM APIs. In our early experiments, we've seen about 200-250 token
19.
by fgfm 3y ago
I appreciate the feedback about clarity, thanks! We'll update the documentation and agree to reflect that more accurately. For now, we've started with VSCode as an IDE and used GitHub for authentication. But actually, we're a
20.
Show HN: Quack Companion – VSCode extension for OSS contribution assistance(github.com)
49 points by fgfm 3y ago|11 comments
21.
by fgfm 3y ago
This approach feels like pruning, but the speedup is considerably higher. Interestingly, I'm curious how this will play out on more recent transformer architectures though: I guess the speedup will be more important for the largest arc
22.
Exponentially Faster Language Modelling(arxiv.org)
2 points by fgfm 3y ago|1 comments
23.
by fgfm 3y ago
Orca 2-13B consistently beat Llama 2-70B on most benchmarks in 0-shot. Hopefully, research papers will start to include Mistral/Zephyr 7B & Openchat 3.5. Even though they're smaller, they're getting competitive against mu
24.
Orca 2: Teaching Small Language Models How to Reason(arxiv.org)
310 points by fgfm 3y ago|80 comments
25.
OpenChat 3.5: 7B model with comparable perf to ChatGPT(huggingface.co)
6 points by fgfm 3y ago|2 comments
26.
by fgfm 3y ago
The Alignement AI Lab just published OpenChat 3.5, which is outperforming ChatGPT (march version) on most benchmarks apart from MMLU (67.3% vs 64.3%) & BBH-CoT (70.1% vs 63.5%).
27.
Loading Llama-2 70B 20x faster with Anyscale Endpoints(anyscale.com)
1 points by fgfm 3y ago|1 comments
28.
by fgfm 3y ago
The Anyscale team shared how you can achieve considerable speedups for model loading in production with examples on the Llama 2 variants.
29.
by fgfm 3y ago
Well, that was a blog post, but they just released a research paper. And in comparison to the blogpost, they indeed added more information regarding the attention mechanism they used, details about the architecture, more evaluation results
30.
by fgfm 3y ago
The research paper by Mistral about their Mistral 7B v0.1

More ›