gpugreg

7 ms·

1.
by gpugreg 4d ago
I currently use DeepSeek-V4.1-Flash and the V4-Pro and -Flash versions before that, because they are extremely cheap. I have spent less than $25 for over a billion token so far (1B cached, 15M out, 19M in). I even prefer the DeepSeek models
2.
by gpugreg 5d ago
I've had some success with tokens as a measure of complexity instead of number of lines, but should be combined with additional rules, e.g. disallowing lambdas, exec, eval, compile, __import__ and complex list comprehensions for Python
3.
by gpugreg 5d ago
Both DeepSeek-V4.1-Flash and GLM-5.3-Flash failed to decode your embedded example text. I failed, too, but I only spent a minute trying to figure out your repo before giving up and telling AI to do it. Anyway, maybe you want to improve your
4.
by gpugreg 6d ago
For me, "Improve the model for everyone" was "On", although I disabled a similar-sounding checkbox in the past (Germany).
5.
by gpugreg 7d ago
They did, but it was not well-received. Perhaps they want to try something different.
6.
by gpugreg 7d ago
> those things are not deterministic Determinism was an explicit goal of DeepSeek-V4. From their paper: https://arxiv.org/html/2606.19348v1#S3.SS3 > we implement end-to-end, bitwise batch-invarian
7.
by gpugreg 8d ago
It is easier to trust what you can understand.
8.
by gpugreg 8d ago
Is there any cryptocurrency that uses AES?
9.
by gpugreg 9d ago
Not necessarily. These days, SSDs can go up to multiple millions of random reads per second. TLS termination (or self-inflicted software bottlenecks) will become an issue much earlier.
10.
by gpugreg 10d ago
I also thought about building one of those AI honeypots, but I stopped when I realized that it would quickly be turned into a command and control server by botnet operators, followed by mail from a three letter agency. Is there any way to a
11.
by gpugreg 12d ago
I scrolled through https://simonwillison.net/tags/pelican-riding-a-bicycle/ but did not see any image where the spokes were correct. For a moment, I thought that the text-to-image model might have gotten it right,
12.
by gpugreg 13d ago
MiMo-V2.5-Pro-UltraSpeed gets pretty close with over 1000 TPS on 8x B200. It has 1.02T total parameters and 42B active, compared to 27B total/active for Qwen3.8-27B. Also, B300 are out now. I think 1500 TPS for Qwen3.8-27B should be do
13.
by gpugreg 13d ago
Cached tokens count towards the limit as well. For example, if your context window is 50,000 tokens, it takes 9 requests to reach that limit without generating a single token.
14.
by gpugreg 13d ago
I was wondering whether this was any good for programming, but it is too fast for its own good. There is a limit of 450,000 tokens per minute. I hit this limit in about 90 seconds and burned through $1.10 while doing so. This is because cac
15.
by gpugreg 13d ago
Thanks, that is very kind of you. I have submitted bugs in the past and was enthusiastic for GPU support for over 15 years now, but I lost my faith.
16.
by gpugreg 13d ago
You can query polars data frames with SQL: https://docs.pola.rs/api/python/stable/reference/expressions... Unfortunately, polars does not support parameterized queries, so the risk of SQL injection is ex
17.
by gpugreg 14d ago
I tried it and the experience was not great. I first enabled chrome://flags/#force-enable-webgpu-interop which did nothing. Next, I enabled chrome://flags/#enable-unsafe-webgpu which made some WebGPU demos work
18.
by gpugreg 17d ago
Not all their research, but certainly a lot: https://github.com/orgs/deepseek-ai/repositories?q=sort%3Ast...
19.
by gpugreg 17d ago
The kettlebell in this [1] image looks a bit like the one [2] that was recalled due to radioactivity. It's probably a different one, but I thought I should mention it anyway. Better safe than sorry. [1] https://greenlightnin
20.
by gpugreg 18d ago
I could have used more precise terminology. rfind is average case O(n + m), worst case O(n * m). Imho the worst case performance is more important than the average case performance, since it tells us whether there is any risk for attack
21.
by gpugreg 18d ago
Because the use case is very niche and nobody optimized it yet. https://github.com/python/cpython/issues/135824#issuecomment... `x in range(n)` is already optimized, but that was easier since the `__contains_
22.
by gpugreg 18d ago
Notable pitfalls: - s[i:j] is O(j - i) because it creates a copy instead of a view - max(range(n)) is O(n) - substring search is O(n), which is good, but rfind is O(n m) - iterative string concatenation (for c in ...: s += c) can be O(n^2)
23.
by gpugreg 1mo ago
Sure! But where?
24.
by gpugreg 1mo ago
To learn about sentiment analysis, I'd look for related datasets and then look at recent code, e.g. here: https://www.kaggle.com/datasets?search=sentiment+analysis For more LLM-specific stuff, you can pick some agent t
25.
by gpugreg 1mo ago
Agents usually start with ingesting the existing code base, and DeepSeek can use those code bases for pretraining. And they will have filters on top of that to throw out garbage. I am not sure how they are using the data for post-training,
26.
by gpugreg 1mo ago
Oh, I messed up. Half-way through, I thought it would be a good idea to double the numbers so I don't have to deal with half millions, but forgot to also double the 98.5. Unfortunately, I can not edit it anymore. I think the margins of
27.
by gpugreg 1mo ago
Agentic workloads are somewhere around 1%/0.5%/98.5% input/output/cached tokens. Cached tokens are pretty much free for inference providers (if they implement sparse and compressed attention properly) and throughput for
28.
by gpugreg 1mo ago
Personally, I prefer QDirStat. I just tried to use FileLight to compare, but the package seems to be broken on Lubuntu.
29.
by gpugreg 1mo ago
> you're having issues handling files properly? I guess they were using ollama, which does not tell you where it puts the models it downloads.
30.
by gpugreg 2mo ago
I get the following error: Traceback (most recent call last): File "/app.py", line 1, in <module> import spider ModuleNotFoundError: No module named 'spider' Steps to reproduce: 1. V

More ›