enoch2090

You can achieve this with features that Tmux already have. Glued a tool that works:

https://gist.github.com/Enoch2090/5026c417f86ff6ff4fbe30c22b...

You better delete all third-party applications for they are having full disk access.

Looking forward to your progress! Just checked the paper and it says the underlying backbone is still DETR. My guess would be that SAM3 uses more video frames during the training process and caused the dilution of sparse engineering-paper-like data.

GW! How does Onyx differ from Open WebUI and its alike?

Surprisingly, SAM3 works bad on engineering drawings while SAM2 kinda works, and VLMs like Qwen3-VL works as well

Although a bit off the GPU topic, I think Apple's Rosetta is the smoothest binary transition I've ever used.

That's the essence of these services, they never explicitly mention the quota, or secretly lowers it at some point.

Played with the demo a bit and I got confused.

1. The chat context is always provided, and that introduces a bit of uncertainty - when the chat history mentioned something the model is always inclined to connect with it.

2. When I tried to set each context to an empty string, the model doesn't show any evidence of remembering concepts. I told it 5 times that I love cats, and when asked about its favorite animal, its output remains "honeybee" and "octopus".

Can't wait to replace my explorer with this, hope the support will come soon

Looks pretty good but all CJK characters are displayed as questionmarks (???) and switching to a CJK native font does no help. So my user folder is now crowded with ???????? which makes it hard to navigate :(

For llama3 just ask him to install ollama and serve the model. Ollama has auto memory management and will free the model when not used, and whenever you make a call to the API (do let your friend know before you do this) ollama will reload the model back to memory again.

Not sure whether there are anything similar for SD though.

I wonder how would they calculate the metrics if the result is generated instead of retrieved? Is it likely that the LLM can generate exactly the same output as the desired result?

I think Qt does have WASM compilation options, but I mean generating interfaces given models like Pydantic/Dataclasses/etc.

Streamlit's state management is so painful, feels like constantly writing hacks.

Actually, does Qt have something like FastUI? Been looking for one.

The TOC is a total mess

Totally makes sense - I'm already getting used to "ChatGPTing" instead of Googling these questions. I guess the future of these articles is that there are always fields that LLMs hallucinate at, fields that are not very common (explain the products of a small brand, pros &cons of their different models).

BM25 is just too famous in traditional searching and not worth explaining it all over again.

This is really cool! We are recently developing a project with heavy C++ and maybe a little Python scripts & wrappers and we are planning for a major refactor. Is it possible to adopt this with a C++ codebase?

Normally having a lot of choices is a good thing, but here we are facing a dozen of vector dbs with very similar features - to the root it's just some version of ANN implemented in C++/Rust/whatever, the "peak" means there's nothing new. People are flooding into this field not because there's something worth inventing, but more of fear to lag behind and miss the quick money. That's what I feel about vector DBs in Jan, 2024.

Once I accidentally lost 3 poems on my blog as well as the linked Notion database, and weirdly there's nothing in the version history. Thanks to the internet archive that I found those poems back.

Sometimes it's not the time saved, but a change to your mindset, that you don't need to even think of walking across the room to reach the switch before sleep. Such change accumulates and at a point I felt the difference, hopefully that works for you as well.

I've said this a few times previously yet I certainly want to say it again - "All You Need" titles are definetely not what all we need.

There's a browser embedded, so you can open the same website in the browser and recurse

This is great! Can you also add a preference menu to let users set custom fonts?

Nice concept and demonstration! Would argue though that it is the states in Jupyter notebooks that make it work so well for prototyping, and that should be its main purpose always. Using dependencies might break that.

As for the documentation purposes it should be the programmers/authors' duty to ensure that the code works out by executing all cells in order once.

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 local, more stable models, such as Llama2 embeddings or simply some BERT-family-model (which is more affordable).

When outages comes up you simply switch from one vector space to another. Though possible, model alignments are much harder and more expensive to achieve.

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.

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?

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 to see more in-depth examples such as high-speed circuit optimization (say a clk over 600MHz), complex filters and PID controls, etc. Wouldn't hurt even if they only achieve 20% correctness on these problems, as long as the approach can work once every few times.