Senin, 04 September 2023

New top story on Hacker News: Ask HN: In person industry tours and site visits

Ask HN: In person industry tours and site visits
26 by helghardt | 15 comments on Hacker News.
The head master of my primary school took a small group of students every semester on industry site visits. I believe these site visits played a meaningful role in developing my engineering and entrepreneurial interests/thinking. Although I only vaguely remember the details, I do have a strong lasting impression of the locations/factories we visited and people we met. I got to see a plastic pipe molding facility, coke cola bottling factory, first wind power turbines in Cape Town area and assembled a door alarm prototype at the neighboring university. This was all before the age of 13. In university I had similar practical exposure doing an internship at a boiler manufacturing factory, chicken processing plant and finally a tech startup. I truly cherish these experiences and glimpses into the real world. Obviously I knew very little of what was really going on, but these experiences helped me build a sort-of mental map to unpack my options at the time. Do you think practical site visits as a teenager is a good idea? Have you had similar exposure and did it have a lasting impact on you too? Do you think we need to create more opportunities like this for students?

New top story on Hacker News: VSCodium – Libre Open Source Software Binaries of VS Code

VSCodium – Libre Open Source Software Binaries of VS Code
27 by Brajeshwar | 3 comments on Hacker News.


New top story on Hacker News: Show HN: finetune LLMs via the Finetuning Hub

Show HN: finetune LLMs via the Finetuning Hub
8 by rsaha7 | 0 comments on Hacker News.
Hi HN community, I have been working on benchmarking publicly available LLMs these past couple of weeks. More precisely, I am interested on the finetuning piece since a lot of businesses are starting to entertain the idea of self-hosting LLMs trained on their proprietary data rather than relying on third party APIs. To this point, I am tracking the following 4 pillars of evaluation that businesses are typically look into: - Performance - Time to train an LLM - Cost to train an LLM - Inference (throughput / latency / cost per token) For each LLM, my aim is to benchmark them for popular tasks, i.e., classification and summarization. Moreover, I would like to compare them against each other. So far, I have benchmarked Flan-T5-Large, Falcon-7B and RedPajama and have found them to be very efficient in low-data situations, i.e., when there are very few annotated samples. Llama2-7B/13B and Writer’s Palmyra are in the pipeline. But there’s so many LLMs out there! In case this work interests you, would be great to join forces. GitHub repo attached — feedback is always welcome :) Happy hacking!

New top story on Hacker News: TinyLlama project aims to pretrain a 1.1B Llama model on 3T tokens

TinyLlama project aims to pretrain a 1.1B Llama model on 3T tokens
27 by cmitsakis | 1 comments on Hacker News.


New top story on Hacker News: Black Art of 3D Game Programming (1995)

Black Art of 3D Game Programming (1995)
7 by stefankuehnel | 0 comments on Hacker News.