sroecker

Yes, but ramalama defaults to podman. Podman Desktop is very easy to install and use.

He also built a nice "little" cluster with Nat for their startups: https://andromeda.ai/

Yes, used by Grok: https://twitter.com/qdrant_engine/status/1721097971830260030 Oh wow, completely missed that they're German. Should have noticed their "Impressum"..

Care to enlighten us why?

You probably don't need fine-tuning, at least if it's just new content (and no new instructions). It may even be detrimental, since LLMs are als good at forgetting: https://twitter.com/abacaj/status/1739015011748499772

Good question, as you can see I haven't touched it for a month. I wanted to show what's possible then with open source and (open) local models and there's already so much new stuff out there.

I'll probably fix some things this week and then either update it or start from scratch. Guided generation, structured extraction, function calling and multi-modal are things I wanted to add and chainlit looks interesting.

We just held a workshop about this a few weeks ago: https://red.ht/llmappdev We created a simple chatbot using local models with Ollama (llamacpp), LlamaIndex and streamlit. Have a look at the streamlit folder, it's super easy.

I used this simple example to teach about RAG, the importance of the system prompt and prompt injection. The notebook folder has a few more examples, local models can even do natural language SQL querying now.

I can't even see the pay what you like feature on the app.

That version of libnc_cuda.so doesn't seem to be compatible with ts_server though.

$25B and they can't even afford an API..

If you want analytics you have to rely on a 3rd party provider who somehow gets the information by logging into your account. I guess if you're a key account they will send you some data and maybe even some nice visualizations.

I did some research and people really track their audience with spread sheets. Influencers send screenshots of their accounts to advertisment companies to prove how many followers they have. Seriously.. If they want to monetarize their platform they should open at least part of their API.

Please do not use Telegram. It' closed source and uses some half-baked crypto. Signal is open source and is actually end-to-end encrypted.

There is a NeuroBayes plugin for TMVA. I've put it on Github with some additional patches: https://github.com/sroecker/tmva-neurobayes

I did some research:

Around 0.6% of all Bitcoin are known to be lost, so with unknown losses it could be 1-2%(https://bitcointalk.org/index.php?topic=7253.120)

Excluding the Pirate scam, 2.9% have been stolen. These are still viable, some are tainted since the receiver address is known. (https://bitcointalk.org/index.php?topic=83794.0)

It has been estimated on bitcointalk that around 1-2% have been lost. That's not really a large amount. You can still divide bitcoins into satoshis (1E-8) or even further with a protocol change in the future.

.onion sites don't need SSL as Tor traffic itself is already encrypted.

Who decides what is legal and what not? Is it even legal for an ISP in the US to monitor your traffic or to act like this without a court order on behalf of a shady company? I don't get it that nobody is protesting about this...

You can even donate some Bitcoin if you don't have a Paypal or Amazon account.

Yes I know, this is a bug in Google Translate. However, the bug was found by a user using Google Now and was reported in the Android issue tracker. The issue was described very well and is reproducible. I am criticizing that nobody seems to check the reported issues and the time it takes for Google to fix the reported problems.

Wow, this speaks volumes about their QA process, especially for Android. This bug is known since October '12, how hard can it be for a company with 50k employees to fix a bug like this?

I would guess this won't be easy in rural India.

It would be awesome if they load it up with lots of free ebooks and a decent offline copy of wikipedia (wikidroyd) from the start.

Yep. MongoDB for example is used for their Dataset Bookkeeping Service. JSON is indeed used but only for 'good run' lists.