freeqaz

22 ms·

301.
by freeqaz 4y ago
I thought this was pretty cool and wanted to share it here. This makes me think about how the last time we actually took up-close photos of Uranus was with Voyager tech, and time has marched on since then! It makes me wonder what data can b
302.
by freeqaz 4y ago
They also said "the camera placement on the old HW3 cars was insufficient to see things very near the vehicle needed for parking, so it was never going to work well." So it's probably more of that!
303.
LLaMA-Adapter: Efficient Fine-Tuning of LLaMA(github.com)
5 points by freeqaz 4y ago|0 comments
304.
by freeqaz 4y ago
When I've played around with this on my box, I haven't been able to get the output to be very decent. Maybe I'm just used to ChatGPT though. Does anybody have some examples of output that it generates to explain what this is
305.
by freeqaz 4y ago
Yeah there are some models that I played with that can do this. They only work for 2 or 3 speakers currently though. They term for this is "diarization". https://huggingface.co/pyannote/speaker-diarization
306.
by freeqaz 4y ago
So ~18.6% trained relative to LLaMa. That's not _nothing_ but it's also not great. Thanks for digging into this!
307.
by freeqaz 4y ago
That makes sense especially since they're not intending to deploy this model to production. For models like GPT-3/4 it makes sense why they would train them more because the costs of running the inference "in production"
308.
Cerebras-GPT vs. LLaMA AI Model Performance Comparison(lunasec.io)
263 points by freeqaz 4y ago|70 comments
309.
by freeqaz 4y ago
If you have to ask, you can't afford it! Mostly teasing but my guess would be $500k+ since they'll likely price it so that it is the same $ as the equivalent NVIDIA cluster (or very close to it).
310.
by freeqaz 4y ago
I'm on mobile and struggled to compare these two tables properly. Would you mind posting a summary of your findings? Here are some values but I don't know what they mean. LLama 60B on the left, Cerebras 13B on the right. PiQA: 82.
311.
by freeqaz 4y ago
Is there an equivalent to Nvidia-Docker for ROCm to make it brain dead easy? Where does the complexity come in for AMD GPUs?
312.
by freeqaz 4y ago
Taking a paper and turning it into working production code is a non-trivial process, 100%. Training big models takes a lot of random reads/writes and those tend to be pretty latency sensitive. There _may_ be a way to train this BitTorr
313.
by freeqaz 4y ago
This video doesn't seem to have the extra wheels. Not sure why! https://www.youtube.com/watch?v=9lc72Uvq29U
314.
by freeqaz 4y ago
Where is the limit with copyright law then? Is ChatGPT allowed to be trained on books? If the "terms" of the book ban it, can I just read the book and then write notes which I upload to ChatGPT? (Would a Microsoft employee be paid
315.
Discord hops the generative AI train with ChatGPT-style tools(arstechnica.com)
1 points by freeqaz 4y ago|0 comments
316.
by freeqaz 4y ago
How does that performance scale though? If you double the inputs you train on, do you get twice the performance? Less? More? I've heard conflicting things and I'm curious to hear your reasoning!
317.
by freeqaz 4y ago
Are you planning to make this Open Source? We run an Open Source project[0] and we would love to work with y'all since this is related to where our heads are at, but we would want to have the core be open for that! 0: https:/&#x2
318.
by freeqaz 4y ago
I'd like to be a bit more optimistic that we could move quickly on something that would greatly help us curb greenhouse gas emmisions!
319.
by freeqaz 4y ago
It's definitely not happening at the high end of the market (NVIDIA A100s with 40GB or 80GB of RAM). The cards that were used for mining have since crashed in terms of prices, but those were always gamer cards and very rarely Datacente
320.
by freeqaz 4y ago
It's also worth mentioning that, because Microsoft is an investor, they're likely getting these at cost or subsidized. OpenAI doesn't have to make money right away. They can lose a small bit of money per API request in exch
321.
by freeqaz 4y ago
Per instance (worker serving an API request) it requires 8x GPUs. I believe they have thousands of these instances and they scale them up with load. Because the model isn't dynamic (it doesn't learn) it is stateless and can be sca
322.
by freeqaz 4y ago
There is a cost to fixing every CVE. Dependabot is designed to fix every CVE which is highly inefficient when only a fraction of those CVEs matter/are exploitable. Approaches like what EdgeBit is taking are very important to reducing f
323.
by freeqaz 4y ago
What's growth rate on that revenue though? If it's like 20-30% growth per year... you're talking a pretty short amount of time (3-5 years) until that revenue multiplier is at parity with many publicly traded tech companies.
324.
by freeqaz 4y ago
I've seen this effect get amplified also when somebody puts a "bad" answer in a public place like StackOverflow. It is possible to have quite a large blast radius from something like this!
325.
by freeqaz 4y ago
I agree that it is a threat. I don't think this instance is (it's too noisy). I wrote a comment on the NPM thread earlier ( https://news.ycombinator.com/threads?id=freeqaz ) that I'll quote here: > "Whi
326.
by freeqaz 4y ago
Much appreciated!
327.
by freeqaz 4y ago
It's hard for the enterprise world because they need to mirror everything internally. (That's a doable problem and I believe Google proxies all Golang installs already.) There are still issues with things like history being re-wri
328.
by freeqaz 4y ago
The lead is buried at the bottom of the article (imo). > "While being flooded with spam is never good, it gets immediately noticed and mitigated. It's harder for open source projects to spot and stop rare one-offs" This is
329.
by freeqaz 4y ago
How did Microsoft sour developers with Copilot? I know dozens of people that pay for it (including myself) and I feel like it is widely regarded as a "no brainer" for the price that it's offered at. Please help me understand!
330.
by freeqaz 4y ago
Starred! We've been looking to build something similar so I appreciate you sharing this here. The only other project that I've seen that's doing something close to this is this one: https://github.com/getbuff&

More ›