danieldk

12 ms·

181.
by danieldk 6y ago
It's not like Intel is stellar in terms of privacy. I think the grandparent means that using an M1 implies using macOS, which had its share of potential privacy issues recently (regularly phoning home on launching apps, having Apple
182.
by danieldk 6y ago
I surely hope that this is one of the outcomes of Apple and Amazon's ARM push. Perhaps Intel will realize that's it is better to have competition on its own architecture than complete ecosystems moving to foreign architectures. On
183.
by danieldk 6y ago
I don't know, perhaps vote for politicians that do have your rights and not corporation's rights at heart? Convince people that protecting their rights and livelihoods is not socialism or unamerican? I know that this is difficult
184.
by danieldk 6y ago
Interesting! Do you have a source?
185.
by danieldk 6y ago
Well unless they also support acceleration on AMD GPUs, this is not so interesting. Training on x86_64 CPU cores or integrated Intel GPUs is really slow compared to training on modern NVIDIA GPUs with Tensor Cores (or AMD GPUs with ROCm,
186.
by danieldk 6y ago
Interesting, because the README in the repository states that it will only use CPU cores in eager mode: Please note that in eager mode, ML Compute will use the CPU.
187.
by danieldk 6y ago
So far, of the models that run on GPUs with 8-16GiB VRAM XLM-RoBERTa has been the best for these specific tasks. It worked better than the multi-lingual BERT model and language-specific BERT models by quite a wide margin.
188.
by danieldk 6y ago
Yes https://pytorch.org/tutorials/beginner/hybrid_frontend/learn... https://pytorch.org/docs/stable/jit.html But people normally use PyTorch in eager mode.
189.
by danieldk 6y ago
As far as I understand, Apple's ML Compute framework cannot use the neural cores in eager mode (only in graph mode). So, the neural cores would probably only be used with TorchScript. (The TensorFlow implementation has the same limitat
190.
by danieldk 6y ago
That's just a 8.4KiB ZIP file of the repository, which obviously does not contain anything interesting.
191.
by danieldk 6y ago
I wonder how much it is even possible to train cutting-edge models models. I am sure that there are still tasks where a simple feed forward network or RNN. However, you can just barely finetune any for the base pretrained transformer models
192.
by danieldk 6y ago
There's nothing preventing you from sharing weights across layers, and would be interesting to see some research about that. E.g. the ALBERT model does that: https://arxiv.org/abs/1909.11942 I have done model di
193.
by danieldk 6y ago
That's only for pretraining a model. Very few groups pretrain models, since it is so expensive in terms of GPU time. Finetuning a model for a specific task typically only takes a few hours. E.g. I regularly train multitask syntax model
194.
by danieldk 6y ago
These issues (allow updates within limits) are what I understand as the point of dependency resolution, I'm trying to understand how you do without it. In such a case e.g. nixpkgs makes two attributes: a_1 and a_2 (and alias a to
195.
by danieldk 6y ago
I thought that meant that package X might specify a dependency on A In Nix and Guix, you would indeed not specify that X needs package A version 2.4.2. You could see it as X using the 'build recipe' of A as its dependency. So, i
196.
by danieldk 6y ago
file system behaviour is also a Linux thing It's required by POSIX, so all modern UNIX-like systems have this behavior: When the file's link count becomes 0 and no process has the file open, the space occupied by the file shall
197.
by danieldk 6y ago
Unless you use anything that does JIT compilation (e.g. JVM), e.g. JetBrains IDEs are really slow onder Rosetta. Another pain point is code that relies heavily on SIMD. AVX/AVX2/AVX-512 instructions are not translated, so in tho
198.
by danieldk 6y ago
Not intended in a nasty way, just factual: The only reason I drive it is because it's so comfortable to sit higher up, and to have a better view of the road that way. And thereby blocking the view of 'regular' cars.
199.
by danieldk 6y ago
macOS grep is from FreeBSD, at least as of macOS Mojave (10.14). Yep, it still is. On Big Sur on M1: $ grep --version grep (BSD grep) 2.5.1-FreeBSD
200.
by danieldk 6y ago
Linux is the only unix OS that gets traction Darwin (macOS, iOS, iPadOS, tvOS, watchOS)? And I guess that technically Minix is one of the most successful Unix-like systems. [1] [1] https://www.cs.vu.nl/~ast/intel&#x2F
201.
by danieldk 6y ago
That would serve to reduce effective memory usage or need to fall back to swap. I would not expect something like that to be responsible for more than, say, a 5-10% RAM usage decrease though. I think you can reach a lot more than that. Pr
202.
by danieldk 6y ago
I also have an Intel 660P, works fine with NixOS. It shows up as /dev/nvme0n1 .
203.
by danieldk 6y ago
I don't have enough background, but I recently read that some distributions could be a fire hazard due to the lack of proper temperature regulation: https://xnux.eu/log/#017 https://xnux.eu/log&#x2
204.
by danieldk 6y ago
followed by no standard trusted file access UI. If an application uses Gtk+3 or Qt5 then portals will be used automatically for Open/Save dialogs: https://docs.flatpak.org/en/latest/sandbox-permissions.html#
205.
by danieldk 6y ago
Do you? I think it often feels old, because we associate it with a certain time period. But is what we currently have for desktops so much better than e.g. Windows 2000? At least it was easy to recognize interactive UI elements and there wa
206.
by danieldk 6y ago
You don't know what they did or didn't do to adapt the M1 for computers. There has to be some degree of PCI-E support since they are using thunderbolt ports. There is ethernet on the mac mini with the M1 chip. Also, from the Doc
207.
by danieldk 6y ago
Desktop performance is too poor to run at 4K, even though it works fine in Windows. I have used 4k with Intel, AMD, and NVIDIA GPUs on Linux (both Wayland and X11) and it is a smooth as it is on macOS. My only data point is GNOME though,
208.
by danieldk 6y ago
Eventually, it hit end of life and became AOCL [2]. I've not tried it, but I'm sure it's fine. It's ok. I did some experiments with transformer networks using libtorch. The numbers on a Ryzen 3700X were (sentences per
209.
by danieldk 6y ago
and if you think you can just build the package and get the correct hash from the error message, you're mistaken Actually, you can. In any place a SHA256 hash is required, you can put *lib.fakeSha256" (which is an all-zeros hash
210.
by danieldk 6y ago
Besides NixOS, which is mentioned by jdally987, Fedora Silverblue also fits that bill. A basic immutable desktop OS with atomic updates/rollbacks. Applications are installed through Flatpaks or containers (per Fedora Toolbox or podman)

More ›