fgfm

7 ms·

31.
Mistral 7B(arxiv.org)
267 points by fgfm 3y ago|123 comments
32.
by fgfm 3y ago
The Mistral AI team added an article with additional context on their efforts beyond their new 7B model & their commitment to OSS!
33.
Mistral AI – Bringing open AI models to the frontier(mistral.ai)
3 points by fgfm 3y ago|1 comments
34.
OctoPack: Instruction Tuning Code Large Language Models(arxiv.org)
1 points by fgfm 3y ago|1 comments
35.
by fgfm 3y ago
New code LLM released by the BigCode team from HuggingFace achieving 46.2 on HumanEval by including commits in the training set.
36.
Pixi: A new package manager for Python written in Rust(twitter.com)
7 points by fgfm 3y ago|1 comments
37.
by fgfm 3y ago
A new open-source package manager created by the authors of Mamba (C++ re-implementation of Conda). Looks like poetry and conda in one blazing fast project, looking forward to try it out!
38.
LPython: Novel, Fast, Retargetable Python Compiler(lpython.org)
265 points by fgfm 3y ago|100 comments
39.
by fgfm 3y ago
A new Python high-performance compiler that could compete with Mojo before it even releases.
40.
Universal and Transferable Attacks on Aligned Language Models(llm-attacks.org)
3 points by fgfm 3y ago|1 comments
41.
by fgfm 3y ago
Study on adversarial attacks on LLMs to steer their objective into misalignment.
42.
PanGu-Coder2: SOTA for Code LLMs on 15B(arxiv.org)
8 points by fgfm 3y ago|2 comments
43.
by fgfm 3y ago
There is a new code LLM in town, and with only 15B params, it reaches 62.20% pass@1 on HumanEval!
44.
by fgfm 3y ago
Goat AI just released their new model Goat-7b which surpasses Vicuna, Alpaca & LLaMA 2 in their 7b versions.
45.
Goat AI releases GOAT-7B(twitter.com)
3 points by fgfm 3y ago|1 comments
46.
The economic impact of GitHub Copilot(github.blog)
1 points by fgfm 3y ago|1 comments
47.
by fgfm 3y ago
GitHub just released a report about their ongoing experimentation with Copilot over a million users. Interestingly enough, it highlights the increase in acceptance rate over time. But I'd be curious to distinguish pure performance incr
48.
by fgfm 3y ago
Really neat job guys! Some people see it as an inferior version of Salesforce, which goes beyond a CRM. But for a dev who loves open source & is looking for something that does only what a CRM does, it looks surprisingly good! I have to
49.
by fgfm 3y ago
The repo the paper is pointing to indicates that the code will be released within ~1 week! If the sheer difference in VRAM requirements and latency holds up, this will seriously be a major breakthrough for LLM architectures! Can't wait
50.
by fgfm 3y ago
The funnily interesting aspect here is that they actually managed to get consistent hand generation, which has been quite a problem for Midjourney, Dall-E & co until now! All details in the paper: https://shorturl.at/fnA
51.
by fgfm 3y ago
Replit just released a report on OpenAI dominance in AI-related projects & the growing importance of HuggingFace and LangChain. The competition between HuggingFace & Replicate on open source models hosting looks fierce. But I was mo
52.
AI Safety and the Age of Dislightenment(fast.ai)
15 points by fgfm 3y ago|4 comments