lysecret

30 ms·

421.
by lysecret 5y ago
Well the way I mostly solve it is by using the cache from docker. As well as versioning important packages. As well as multiple requirements files. However, i feel like if with every small change to your libraries your code breaks there wou
422.
by lysecret 5y ago
Ah that I totally understand.
423.
by lysecret 5y ago
Pip and Docker and i never had any issues with python packaging.
424.
by lysecret 5y ago
Not sure why this gets so much hate. I think he's a bit too optimistic about AGI prospects. But in his position these are all interesting and reasonable options. I'm a bit sceptical on this 10,5,1 whatever year ahead metric he pul
425.
by lysecret 5y ago
Hmm I don't know Nix very well. But the features described in the blog post. Aren't they very similar to the features of docker ? And docker doesn't hat all of these "paper cuts".
426.
by lysecret 5y ago
Tangentially related. But i really love the podcast from the janestreet people: signals and threads.
427.
by lysecret 5y ago
I have a theory about current state of the are of recommendations, automation or ML in general. I think a lot of the valid criticisms we see with recommendation systems (building of bubbles, extremization etc.) is due to the simplicity of t
428.
by lysecret 5y ago
About this section: Sadly, dataclasses ignore type hints as well this is excatly what pydantic was made for. As well as its integration with FastApi. FastApi, is really the most beautiful way, I have seen. To use the python type system to i
429.
by lysecret 5y ago
Love typing, if you are writing code with a lot of Data Frame Inputs and Outputs I encourage you to check out Pandera.
430.
by lysecret 5y ago
It's kind of funny hearing both potential employers and employees coming up with "concise" or good enough code example. I think this exactly the beauty of this approach. If you don't talk about some simple sorting Algo c
431.
by lysecret 5y ago
Main thing I am missing in SQLite is array or Json support.
432.
by lysecret 5y ago
I will never go away from a 3:2 Laptop ever again. I am using the Huawei 14, bought it for 700. The only issue is that i cant replace the RAM, and the camera is really bad, being behind the keyboard.
433.
by lysecret 5y ago
Yea same here. What Nosql DB did you use for these lookups? Im currently using postgres for it but seems a bit like a waste. Even though the array field is nice for feature vectors.
434.
by lysecret 5y ago
Interesting post. On thing to note, this seems to be about "on request" ranking. E.g. googleing something and in 500ms you need the recommended content. However, a lot of usecases are time insensitive rankings. Like recommending c
435.
by lysecret 5y ago
Yea, that is also my experience working mostly in startups. I wonder if they will keep up this credit strategy.
436.
by lysecret 5y ago
Could you elaborate a bit? I find this really interesting. My story is: I started with invoke (with boto3) then I wanted some more abstraction and use normal Requests. So i moved them to serverless framework with API Gateway. Then the 30 se
437.
by lysecret 5y ago
https://leloew.com/ I write about data engineering as well as poems :D
438.
by lysecret 5y ago
Really Cool addition. I just moved my lambdas from API Gateway to ALB (because of API G limit to 30s). I also use Serverless framework. It was a day of work, but developing with ALB is a bit more of pain. Maybe this would be better. Are the
439.
by lysecret 5y ago
Would help to know what exactly their product is. When you are building a crud type simple website / web-app this seems like a big overkill. I am doing a lot of data-engineering and serverless is an absolute god-send. I have set up all
440.
by lysecret 5y ago
have you heard of docker?
441.
by lysecret 5y ago
I used selenium. Really like it and very well maintained.
442.
by lysecret 5y ago
I am a bit confused
443.
by lysecret 5y ago
The pain for sure is real. The way we approach it: Whenever there is a data change we re-run our models and see if it improves performance or some other metrics (which are very specific to us). I wonder how valuable "generic" metr
444.
by lysecret 5y ago
Heyo, I looked a bit more at your service, since I am migrating one of our text classification models anyways right now. I decided against using it but maybe my reasoning could still be helpful. (and I see a lot of potential so I want to he
445.
by lysecret 5y ago
One other quick question. Have you built the ide yourself or are you using a provider? It looks pretty cool.
446.
by lysecret 5y ago
Makes a lot of sense! It really is true that ML models kind of live in its own little world with their training loop and should interact with the rest through a rest api. And now and then new data gets added for training, as well as the api
447.
by lysecret 5y ago
Congrats on the launch. I'm quite impressed. Here are my unordered thoughts. So it seems a lot like an improved colab with a deployment stage. Which sounds good to me, it will be much more expensive than colab though. I like the pitch
448.
by lysecret 5y ago
Operationalizing ML is hard. But it has nothing to do with the models at all. It is hard because the main use-case (besides Images and Text processing) is feature fusion, you generate a bunch of distinct features about say People and their
449.
by lysecret 5y ago
Wow. This sounds a lot like the Resnet moment for Transformers.
450.
by lysecret 5y ago
Hijacking this article. But can someone reference me a "how to make Markdown look nice tutorial" I just have zero aesthetic intuition and would love some simple guidelines to follow.

More ›