zetazzed

16 ms·

121.
by zetazzed 6y ago
A100s support MIG (multi instance GPUs), so you can carve each A100 up into 7 separate GPUs for inference of small models. If your inference workload is too small to feed this beast all at once, it can be pretty handy: https://do
122.
by zetazzed 6y ago
My wild, unfounded conjecture: the attacker discovered this recently and had only a short, fixed time window in which to run a scam. Maybe the time before some maintenance update? So none of the more sophisticated approaches (like selling t
123.
by zetazzed 6y ago
From the actual opinion: "Oklahoma replies that its situation is different because the affected population here is large and many of its residents will be surprised to find out they have been living in Indian country this whole time. B
124.
by zetazzed 6y ago
NVIDIA also gave out dividends (I believe Tesla does not?). Looks like $500 in NVDA in 2009 becomes almost $20k of total return by today. ( https://dqydj.com/stock-return-calculator/ )
125.
by zetazzed 6y ago
Both PyTorch and Tensorflow teams have announced they'll support TF32. By design, it interoperates well with existing code, since calling code can just treat it as a regular FP32 value.
126.
by zetazzed 6y ago
I think this is the most in-depth article on Ampere: https://developer.nvidia.com/blog/nvidia-ampere-architecture... Lots of architectural changes like MIG, new floating point formats, etc. Great to see GCP getting VMs
127.
Trump will sign an order suspending work visas (including H-1B)(nytimes.com)
15 points by zetazzed 6y ago|3 comments
128.
by zetazzed 6y ago
Agreed that Great Courses has a sweet spot... they bring in great professors who are particularly engaging and good at teaching. So, much more engaging than Coursera on average but still a high level of expertise. For a tough technical subj
129.
by zetazzed 6y ago
If this article develops your interest in eels, I highly recommend following "Surprised Eel Historian" ( https://twitter.com/greenleejw ) on Twitter for approximately daily facts about the role of eels in medieval E
130.
by zetazzed 6y ago
This goes beyond the existing Spark+XGBoost GPU acceleration to include ETL, Spark SQL, etc. Coming for Spark 3. Full details here: https://www.nvidia.com/en-us/deep-learning-ai/solutions/data...
131.
by zetazzed 6y ago
Interestingly, a top Kaggler (Ahmet) just posted a nice contest solution with SVMs for the TReNDS Neuroimaging contest: https://www.kaggle.com/aerdem4/rapids-svm-on-trends-neuroima...
132.
by zetazzed 7y ago
Hydro disasters can be dams bursts and are no joke! Thousands dead, towns wiped away, stuff like that. https://en.wikipedia.org/wiki/1979_Machchhu_dam_failure Luckily not that common especially with well-maintained dam
133.
by zetazzed 7y ago
Ars review: https://arstechnica.com/gaming/2020/02/nvidia-has-shaken-up-... (Disclaimer: I'm an NVIDIA employee who works on unrelated projects, but I really enjoyed using the beta of GeForce Now.)
134.
by zetazzed 7y ago
People will disbelieve that, but it's absolutely true... I'll never forget the interview I did with an engineer who described an elaborate Hadoop-based solution to some past problem. When I asked him what type of data he was worki
135.
by zetazzed 7y ago
Could you expand on that a little? Are you mostly fighting with high latency for deep learning models for imaging/audio or for traditional ML models on tabular data too? What sorts of latency SLAs do you aim for? (I do some work in thi
136.
by zetazzed 7y ago
UMAP is awesome, and we usually recommend it to people who are struggling with tSNE performance issues. If you have access to a GPU, it's worth checking out cuML's UMAP implementation ( https://rapidsai.github.io/pr
137.
by zetazzed 7y ago
(NVIDIAn here but speaking for myself) - We still have a ways to go, but we're working on it! Really trying to integrate with open data/ML standards, though sometimes an existing codebase doesn't exactly match the GPU approac
138.
by zetazzed 7y ago
I'm curious - do any HN readers invest heavily in actively-managed mutual funds or hedge funds where you aren't personally involved? I feel like it has become a widely-held belief among young, somewhat-affluent techies that index
139.
by zetazzed 9y ago
The most important part of this article is the way that Gelman derives the value. He writes a simple R script to simulate the outcomes. Many, many introductory stats classes will teach you a variety of formulas to estimate the "statist
140.
by zetazzed 9y ago
Prescription drugs represent about 10% of healthcare spending in the US, so while their prices are very high in the US, it's hard for them to explain very much of our high cost system. See https://www.cms.gov/Research-S
141.
by zetazzed 9y ago
Or, the lowest 20% of earners according to CES under-report their income, which may include tips and cash payments, so CES thinks they are overspending.
142.
by zetazzed 10y ago
Under $35k per year AFTER you've paid the $98k up-front cost for a reserved instance. The blog post shows pricing for reserved instances. Hourly price is $13.338 without up-front payment. Expensive, but still nice to have this availabl
143.
by zetazzed 11y ago
The authors may want to consider how much of this work could be done easily and effectively with deep learning. For content-based search and image similarity, even simple, pre-trained convnets will likely crush the histogram-based approache
144.
by zetazzed 12y ago
I found myself getting mad at Math.random() pretty quickly. "Argh, you just got to promote two pawns, and you made them both black bishops!?!? I oughta fire you and hire an LFSR!"
145.
by zetazzed 12y ago
Xen is an example, but, more generally, virtualization has been the big story in systems software for the period mentioned in this article (post-2000). VMware came out of systems research at Stanford and Xen out of Cambridge. The cloud worl
146.
by zetazzed 12y ago
(1) Be very happy that this happened before you actually uprooted your life and moved to a new country, only to find yourself jobless there! (2) Reach out again to the managers behind some of your past job offers. We have hired people who p
147.
by zetazzed 12y ago
Does this mean that Pearson books will start disappearing from Safari?
148.
by zetazzed 13y ago
29% of large, randomized trials did not have published results. This is actually a much better stat than I was expecting, given the significant incentives for industry to hide bad outcomes. There's still an unquantified pile of hidden
149.
by zetazzed 13y ago
The line actually changes - it was recorded twice, once with each emphasis. Coppola expressed regret about the decision, since he also considered just using one take of the line and allowing the audience to imagine the different inflections
150.
by zetazzed 14y ago
This is clearly illuminating the Ponzi-like elements of Bitcoin. As in a Ponzi scheme, early arrivals to the system earn easy money, funded by later arrivals. Tales of bitcoin fortunes draw more and more entry to the system, leading to larg

More ›