Welcome to Software Development on Codidact!
Will you help us build our independent community of developers helping developers? We're small and trying to grow. We welcome questions about all aspects of software development, from design to code to QA and more. Got questions? Got answers? Got code you'd like someone to review? Please join us.
Activity for mr Tsjolder
| Type | On... | Excerpt | Status | Date |
|---|---|---|---|---|
| Edit | Post #295887 | Initial revision | — | 5 months ago |
| Answer | — |
A: How to compute the peak performance of GPU tensor cores? It turns out there is a distinction between the API (i.e. the CUDA docs) and the actual implementation. In the Hopper whitepaper that is linked in the question (h100 SXM5), the implementation details are communicated by means of figures 8-11. Concretely, each figure shows 4 clusters for each ... (more) |
— | 5 months ago |
| Edit | Post #294939 | Initial revision | — | 11 months ago |
| Question | — |
How to compute the peak performance of GPU tensor cores? Not sure if this is the right place to ask, but I am trying to understand the reported theoretical performance of tensor cores on NVIDIA GPUs. Taking the H100 SXM5 as example, the reported peak performance for non-tensor cores is quite straightforward: the number of CUDA cores multiplied with ... (more) |
— | 11 months ago |
| Comment | Post #294753 |
Could it be related to this [huggingface issue](https://github.com/huggingface/transformers/issues/25179)? (more) |
— | 11 months ago |
| Edit | Post #294935 | Initial revision | — | 11 months ago |
| Answer | — |
A: How to add a percent_jitter strategy to backoff Python package and pass extra args in the decorator? One option is to write a (parameterised) decorator that transforms your jitter function with a given `percentage` value: ```python from typing import Callable def percentjitterdec(percentage: float = 0.2) -> Callable[[float], float]: def jitter(value: float) -> float: maxjitt... (more) |
— | 11 months ago |
| Comment | Post #294874 |
Not sure if it would work for this use-case, but have you considered [git submodules](https://git-scm.com/book/en/v2/Git-Tools-Submodules)? They seem to be designed to slightly modify projects your code depends on (e.g. to make `repository_dispatch` work). (more) |
— | 11 months ago |
| Edit | Post #294118 | Initial revision | — | over 1 year ago |
| Answer | — |
A: What is the reverse of an `ndarray.tolist()`? This should be possible using `np.array` as indicated in the Notes of the `tolist` docs (more) |
— | over 1 year ago |
| Comment | Post #294109 |
You're right. I forgot to check the tags. I'll leave it here in case anyone with interest in c++ stumbles across this question. (more) |
— | over 1 year ago |
| Edit | Post #294110 | Initial revision | — | over 1 year ago |
| Answer | — |
A: What binds together the loss function, the optimizer and the model in PyTorch? Your suspicion is right that PyTorch relies heavily on (hidden) global states. In your specific example, `loss.backward()` computes (all) gradients and accumulates them directly in the `grad` attribute of each parameter. You can verify this by printing the grad attributes of the parameters bef... (more) |
— | over 1 year ago |
| Edit | Post #294109 | Initial revision | — | over 1 year ago |
| Answer | — |
A: Macro to count the number of arguments If you are on C++11 (or newer), it should be possible to use variadic templates to write a simple (compile-time) function to count the number of arguments: ```cpp template constexpr int countArgs(const Ts&... args) { return sizeof...(args); } ``` With this function, we can go ahead a... (more) |
— | over 1 year ago |
| Edit | Post #293846 | Initial revision | — | over 1 year ago |
| Answer | — |
A: Why would the tokenizer for encoder-decoder model for machine translation use bos_token_id == eos_token_id? How does it know when a sequence ends? Not sure if this is really a coding question (maybe the ML or AI Tech proposals might be better suited). Not having distinct begin and end tokens is not that extraordinary. E.g. the GPT-2 tokenizer also uses only a single special token (rather than multiple distinct tokens). It is simply a... (more) |
— | over 1 year ago |
| Comment | Post #293717 |
does that mean that you are trying to get some instance of jupyterhub running on a bare metal server (and you don't have contact details of the admin)? If yes, what things did you try and what (didn't) work? These "details" should be added to the question in order for someone to understand what i... (more) |
— | over 1 year ago |
| Comment | Post #293717 |
Also, what do you mean by deploying? Are you able to deploy a container? Are you able to start a notebook server? Do you have open ports? In other words, where exactly does the error come from? (more) |
— | over 1 year ago |
| Comment | Post #292774 |
Why the negative undertone? There is no need to project your knowledge or your googlefu on OP. Makefiles can be confusing, especially if you're getting started. (more) |
— | almost 2 years ago |
| Edit | Post #292777 | Initial revision | — | almost 2 years ago |
| Answer | — |
A: make: How to compile all files in a directory. Each rule in the Makefile represents a single target. This means that your first makefile has a single target: `obj/library.o`. I guess that if there is only one target, `make` assumes that this is the target you want to build and does not require targets to be provided by the command line. As a ... (more) |
— | almost 2 years ago |
| Edit | Post #292770 |
Post edited: |
— | almost 2 years ago |
| Edit | Post #292770 | Initial revision | — | almost 2 years ago |
| Answer | — |
A: How to mock methods like `pathlib.Path.is_dir`? It turns out that this is exactly what autospeccing is for. Using ```python from pathlib import Path from unittest import mock somepath = Path("/some/path/") with mock.patch("pathlib.Path.isdir", autospec=True) as misdir: misdir.sideeffect = lambda p: p.name == "" print(Path.isdi... (more) |
— | almost 2 years ago |
| Edit | Post #292760 | Initial revision | — | almost 2 years ago |
| Question | — |
How to mock methods like `pathlib.Path.is_dir`? I used to have some testing code for mocking a simple directory structure when working with `pathlib`. I recently tried to run these tests to learn that some of the internals in `pathlib` have changed in recent Python versions, breaking my mocking setup. I managed to resolve some of the issues al... (more) |
— | almost 2 years ago |
| Comment | Post #291931 |
This answer is missing one of the main reasons in the context of [OOP](https://en.wikipedia.org/wiki/Object-oriented_programming): indicating that a method (and therefore its class) is abstract. (more) |
— | about 2 years ago |
| Comment | Post #291611 |
I would argue that the ambiguity of the term "package" is a problem with Python rather than with this question. I could not imagine someone asking this kind of question to actually know about the ambiguity of "package". (more) |
— | over 2 years ago |
| Comment | Post #291611 |
I must admit that I shortly forgot about that, but it could be that the OP is actually using a Python version < 3.3 and is therefore having problems... (more) |
— | over 2 years ago |
| Comment | Post #291606 |
Although Python is not necessarily an efficiency-oriented language, it might be useful to point out that this approach is probably not as efficient as a bisection algorithm. (more) |
— | over 2 years ago |
| Comment | Post #291609 |
I believe the answer would depend on the kind of setup that you are using for managing your environment. Are you using a system-level installation of python, virtual environments, conda/mamba or something else entirely? (more) |
— | over 2 years ago |
| Comment | Post #291611 |
Is adding a(n empty) `__init__.py` file in each subdirectory considered to be too much overhead? After all, this is the only thing necessary to make Python interpret a directory as a [Python package](https://docs.python.org/3/tutorial/modules.html#packages). (more) |
— | over 2 years ago |
| Comment | Post #291310 |
Are you saying that using pointers is just fine and custom iterators are only useful if iteration becomes more complex? (more) |
— | over 2 years ago |
| Edit | Post #291310 | Initial revision | — | over 2 years ago |
| Question | — |
When to use custom iterators versus pointers I am working on a toy project where I have a container for which I would like to write an iterator that iterates over the values in the container. Because the values are stored in a (c-style) array, this would look something like this: ```c++ class MyClass { int data[10] {}; public: cla... (more) |
— | over 2 years ago |
| Edit | Post #289919 | Initial revision | — | almost 3 years ago |
| Answer | — |
A: How to overwrite lines of STDOUT in Python? You can use the carriage return control character `"\r"` to return to the beginning of the line and overwrite text. When using the `print` function you additionally have to make sure not to write the line feed control character (`"\n"`) is not printed. The easiest way to achieve this is probabl... (more) |
— | almost 3 years ago |
| Edit | Post #289892 | Initial revision | — | almost 3 years ago |
| Answer | — |
A: Regex for simplifying Amazon product URLs The following expression should capture the `{dp|gp}/$ID` part: ```plaintext https://www\.amazon\.com/([gd]p/[A-z0-9]{10}) ``` A quick explanation: - the `\.` are there to match periods only (otherwise it would match any symbol), - `[gd]p` matches either `gp` or `dp`, - `[A-z0-9]{1... (more) |
— | almost 3 years ago |
| Comment | Post #289700 |
The problem is that it does not really make sense to write an answer about `BitSet` in this general context, since e.g. Python does not really have a data structure like that. This is probably also the reason why I assumed this was a Python and not a Java question.
My point is that by abstract... (more) |
— | about 3 years ago |
| Comment | Post #289625 |
Roughly speaking, tensorflow-graph is tensorflow-v1 and tensorflow-eager is tensorflow-v2. I don't think these tags would be that useful. Especially since there are generally not that much tensorflow questions thus far. (more) |
— | about 3 years ago |
| Comment | Post #289251 |
I definitely agree that showing non-working attempts is generally a good thing. I do feel that the non-working attempts should be conceivable from the standpoint of someone who could have this question, however. I just can't imagine that someone would think about tuples in this kind of scenario. (more) |
— | about 3 years ago |
| Comment | Post #289700 |
First of all, I hope it is clear that I did not intend to "attack" your question. I just found the topicality documentation a little ambiguous about this kind of question.
To me, software design/architecture/modelling is one level above the actual algorithm/data structure considerations. On a ... (more) |
— | about 3 years ago |
| Comment | Post #289694 |
I have no problem with the language-agnostic tag. I think my main issue is that this question feels like it originated from Python and was then artificially generalised to cover all languages (or all languages with a data structure for sets that is implemented by a hashtable). In the end, this qu... (more) |
— | about 3 years ago |
| Edit | Post #289694 | Initial revision | — | about 3 years ago |
| Question | — |
Are questions about (abstract) algorithms and datastructures considered on-topic? I just noticed this question about data structures in the Q&A. Although algorithms and data structures are often (part of) a solution to a problem when writing software, this question is constructed in a way that aims to leave out the software part entirely. Especially, because I have a strong ... (more) |
— | about 3 years ago |
| Edit | Post #289693 | Initial revision | — | about 3 years ago |
| Answer | — |
A: Should self-answered Q&A use separate answers for different techniques/approaches (even if there's a caveat that applies overall)? I believe one of the main strengths of Q&A websites is that they provide a more digestible alternative to dense documentation. Especially for programmers who do not yet know how to find/use the documentation for the language that they are working with. The answer (and even the question [^1]) i... (more) |
— | about 3 years ago |
| Comment | Post #289251 |
I agree with @#53398 that this question comes over as too artificial. In the list you linked in the answer to the meta question, you list these top questions. One common aspect of these questions seems to be that they have very short and simple question statements. These questions are popular bec... (more) |
— | about 3 years ago |