Nostr Web Client

Replying to

Obscura VPN

Today is the day: Obscura VPN is NOW AVAILABLE!

Obscura is the first VPN that:

- CAN'T log your activity by design

- Outsmarts network filters

We believe Obscura sets the standard for a new generation of VPNs, and hope you’ll check it out!

🔗 https://obscura.net/

----

Contrary to popular belief, traditional VPNs (even “no-log” ones) can track you – they see both who you are and what you do, just like your ISP.

Your ISP is no better. Since 2017, they've been able to legally sell your sensitive data.

Obscura is different – we never see your decrypted internet packets in the first place.

It’s simply impossible for us to log your internet activity, even if we were compelled to, or if our servers were compromised.

We achieve this by using a fully-independent exit hop run by @mullvadvpn. Ensuring that our servers never see your traffic, and the exit hop never sees your identity.

In fact, you can check your connected server’s public key against those listed on Mullvad’s server page!

But that’s not all…

You may have had the frustrating experience of trying to use your VPN on a restrictive WiFi: in an airport, a hotel, or certain jurisdictions.

Other VPNs will often fail to connect, as their off-the-shelf protocol is easily detected and blocked.

With Obscura, we built our own custom stealth protocol that is much harder to block.

Our protocol blends in with regular internet traffic using the same technology that powers HTTP/3 – QUIC – making it much harder for censors or network filters to detect or block.

To celebrate our launch, Obscura is just $6/month.

Our team has put in the hours to make this all a seamless experience, and I hope you’ll take Obscura for a spin.

🔗 https://obscura.net/

Thanks for reading, and I’ll see you on the free and open internet. 🏄

P.S. For those looking for an exploration of our technical choices, here's our blog post! https://obscura.net/blog/bootstrapping-trust/

brittenedborȝ 10mo ago

Does anything make it impossible for obscura and mullvad to collude and correlate identity with traffic?

Replying to

brittenedborȝ

Imparting "Knowledge" with LoRa / QLoRa has been challenging IME, unless you have *highly* structured data like Q&A with all of the right prompt template tokens for the given model (e.g. https://www.llama.com/docs/model-cards-and-prompt-formats/llama3_1/)

The effect that LoRa training has is kind of noisy, which is why people say that they like them for image models where the results they're looking for are 'thematic' instead of structural or 'domain knowledge', but they are just as effective at imparting thematic/stylistic 'color' on LLM's in my (limited) experience.

brittenedborȝ 1y ago

My suggestion is to go for ~low epoch training and iterate your dataset / parameters frequently instead of wasting compute on big training sessions, you will learn so much more and get much better results doing 10x small training sessions than 3x medium for example

Replying to

jb55

#LoRA makes me bullish on #ai hacking on consumer hardware while still leveraging large pretrained models as a base. It works by putting lower rank matrices at each layer of the transformer stack in a larger base model like llama and training those.

The base model’s weights are frozen, but you can train these low-rank “adapters” which are much smaller and require less memory/compute.

Nice thing about fine tuning is that you are basically teaching the ai new things that it won’t forget all the time. So we can give it lots of domain knowledge about nostr, nips, etc. hardest part is setting up a good training dataset.

brittenedborȝ 1y ago

Imparting "Knowledge" with LoRa / QLoRa has been challenging IME, unless you have *highly* structured data like Q&A with all of the right prompt template tokens for the given model (e.g. https://www.llama.com/docs/model-cards-and-prompt-formats/llama3_1/)

The effect that LoRa training has is kind of noisy, which is why people say that they like them for image models where the results they're looking for are 'thematic' instead of structural or 'domain knowledge', but they are just as effective at imparting thematic/stylistic 'color' on LLM's in my (limited) experience.

brittenedborȝ 1y ago