enoch2090

7 ms·

31.
by enoch2090 3y ago
That's my point, maybe VectorDBs in production should have a fallback mechanism, for the documents inserted, 1. Generate embeddings using services such as OpenAI, which is usually more powerful; 2. Generate backup embeddings using loca
32.
by enoch2090 3y ago
Been playing with Phind for a while and my conclusion is: the Phind model works well on those long existing stuff like C++ libraries, but works generally bad on newer stuff, such as composing LCEL chains.
33.
by enoch2090 3y ago
This reminds us that, what if our databases are maintained using OpenAI's embeddings, and the API suddenly goes down? How do we find alternatives to match the already generated database?
34.
by enoch2090 3y ago
Ooh, this is what I've been looking for! But I agree with pbazarnik that their benchmark are still tackling with toy-level problems. Bits operations, Karnaugh map and other stuff are just one-level of transformation logic, really want
35.
by enoch2090 3y ago
That's called DSL(domain-specific language). Domains that have their DSLs will largely benefit as DSLs can work as an LLM interface. Even LangChain is inventing their own DSL (LCEL).
36.
by enoch2090 3y ago
The work is solid, the approach of domain-specific tokenizer is definitely effective. But the topic seems to be somehow misleading, as I expect the researchers to somehow find a way to connect LLMs for direct design of chips, while this pap
37.
by enoch2090 3y ago
This is cool! One thing that can be easily improved is that the narrator is still trying to describe everything, while people using Be My Eyes they usually got specific purposes (or "attention"). Would be nice to also implement wh
38.
by enoch2090 3y ago
This is sick, the file explorer kept crashed out until I rebooted, and then I have to scan my disk for risks for hours. No risks for now, don't try to open that file guys :(
39.
by enoch2090 3y ago
I tried to open this file while it says unsupported encoding, and crashed out my file explorer
40.
by enoch2090 3y ago
This is cool! Reminds me of that old iOS game Hacker Run :)
41.
by enoch2090 3y ago
When I use ChatGPT browse it's mainly two cases: 1. I don't want to read through all materials in that topic but want an overview. 2. I can't describe the topic with the precise language for it, hoping ChatGPT can translate m
42.
by enoch2090 3y ago
I have a feeling that in most use cases this is replacing grep and awk in a familiar way to Python programmers, especially the latter with its own grammar. Fun stuff!
43.
by enoch2090 3y ago
This is just so Edge. Unrelated with this post - I used to hate Edge so much for poping out and ask to replace Chrome. Therefore I insisted on using Chrome for years. One day I misclicked, and found out that Edge had exactly cloned my Chro
44.
by enoch2090 3y ago
Back in highschool they also had those kiosks for adding funds to our student cards, and it would speak out like "Success". I saw someone crashed out the app (probably with a race condition, too) and used the ethernet to download
45.
by enoch2090 3y ago
I think features like Google search and web scraping are really important, and you should keep adding them (like ChatGPT's official plugins). I would use Copilot for these features, but otherwise using Actions' built in ChatGPT ac
46.
by enoch2090 3y ago
The points that you don't care about is not essentially the points that other people don't care about. If you are strong the weight may not matter, but otherwise there is a significant difference imposed on fingers & wrists. I
47.
by enoch2090 3y ago
And I wish I could just trade the Type-C case with my lightning case (with at most a tiny refurbishing fee), there are a lot of people who just bought the Airpods pro 2 last or this year and not likely to purchase again in 2 yrs
48.
by enoch2090 3y ago
USB-C is great, but honestly I'm not transmitting data via the port, nor do I care about charging speed. Then it becomes weird that I still have to carry a lightning cable for my Airpods.Would be a huge plus if I can just carry one cab
49.
by enoch2090 3y ago
Wrote a poem for this a few years ago: My code was sent to the north pole. All the segment faults. Invalid access. And stack overflows. And I was told that one day the vault would be moved to Pluto. An archive for humanity, carved on giant
50.
by enoch2090 3y ago
Looking at the code and instantly I knew it's Huggingface's Transformers. Then I scrolled through the list for eons while the page kept hinting me things I already knew. That's just so frustrating.
51.
MetaGPT – A LLM powered simulation of software dev company(github.com)
2 points by enoch2090 3y ago|0 comments
52.
by enoch2090 3y ago
Kagi summarizer is good indeed! But after trying out Kagi search for a while I feel like it just can't replace Google :(
53.
by enoch2090 3y ago
It's probably fine for me if they just want to nerf the ability to bypass paywalls. But it's very common now for me that I found something lengthy and informative, pull the URL and ask ChatGPT to summarize. If that is also nerfed
54.
by enoch2090 3y ago
Now whenever I need to use a fancy new package that I never used of, I use LangChain to collect all documents from the package document site, load them into a vectorDB and start asking GPT questions. This method works in 80% of the time. On
55.
by enoch2090 4y ago
One fun thing about this. There's this teacher that I met in middle school who also knows this trick. I'm Chinese, that guy too. He was old (in his 70s) and he definitely didn't know how MIT does it so this is a common trick
56.
by enoch2090 4y ago
These are just frameworks that allows you to make an easy interface for your mostly-Python-based code. The problem is never that “I only want Python”, but “I want a web app but I don’t know much JS”.
57.
by enoch2090 4y ago
That sounds really promising though, you convinced me. Will have a try.
58.
by enoch2090 4y ago
Upvote for this request! It's really not some pleasant experience to implement all this in Streamlit. Should also support some third-party authentications.
59.
by enoch2090 4y ago
This is fun, but reading the commands[1] reminds me of learning a new language. There are just too of them, and they are all pipelined in a functional manner so I need to understand the I/O of them. In that case, why would one bother r
60.
by enoch2090 4y ago
Native Chinese speaker here. All transformations seem legit to me, the 兔/rabbit one was especially good since it is beautifully transformed, and I would say it still readable for a native speaker. For the 猫/cat and 冲浪/surfing

More ›