GaggiX

40 ms·

961.
by GaggiX 3y ago
Deleted by the author, no DMCA.
962.
by GaggiX 3y ago
The repo was deleted by the author, no DMCA. I imagine the author realized that the code uploaded do not prove any of the claims made on the README.
963.
by GaggiX 3y ago
It shows how incompetent the attacker was, I report below what Retr0id wrote in the issue: "tl;dr: The "leak" seems real, but doesn't prove any of the claims made in the readme. This statement from Peiyuan Liao, the rabb
964.
by GaggiX 3y ago
The repo falsely claimed to have released the language action model code, you can read Retr0id's analysis here: https://github.com/rabbitscam/rabbitr1/issues/11
965.
by GaggiX 3y ago
I don't want to ruin the party, but I find it hard to believe that this would have any tangible effect.
966.
by GaggiX 3y ago
I submitted this analysis from Retr0id because I first believed the claims in the repo were real and then felt like an idiot.
967.
Rabbit R1 source code analysis by Retr0id(github.com)
4 points by GaggiX 3y ago|2 comments
968.
by GaggiX 3y ago
We already know that the large action model would not be available at launch, but I wonder how well it does works as an AI assistant.
969.
by GaggiX 3y ago
What would the file extension be?
970.
by GaggiX 3y ago
Claude 3 Sonnet is free like ChatGPT-3.5 but much more powerful, for what I need to do these chatbots + GPT-4-turbo can do most of what I need, from scripts for managing files to templates, rewriting code, actually solving difficult tasks (
971.
by GaggiX 3y ago
Yeah, chatbots have completely changed the way I program, and I'm not a "self-mythologizing executive and founder in Silicon Valley", also it's not only very useful but also entertaining, quite a lot of fun creating stuf
972.
by GaggiX 3y ago
The article does mention LoRom and HiRom, I imagine you're talking about the same thing.
973.
by GaggiX 3y ago
Yeah that's why today Cyberpunk is a very different game.
974.
by GaggiX 3y ago
Are there any companies that would actually pay the ransom? Because I have never seen one.
975.
by GaggiX 3y ago
Nothing really, the source code is also quite old, Cyberpunk 2077 is very different now.
976.
MeshLRM: Large Reconstruction Model for High-Quality Meshes(sarahweiii.github.io)
2 points by GaggiX 3y ago|0 comments
977.
by GaggiX 3y ago
Reading the github repo, it doesn't seem that Llama 3 70B has won the game but "is the only open weight model, which can play through the first pass of the game".
978.
Dynamic Typography: Bringing Text to Life via Video Diffusion Prior(animate-your-word.github.io)
75 points by GaggiX 3y ago|16 comments
979.
by GaggiX 3y ago
Some models are better than others, for example Claude 3 Sonnet will refuse to answer because it doesn't know who Tim Bray is, and I imagine GPT-4 will do the same, unfortunately open models are not at that level yet.
980.
by GaggiX 3y ago
The dataset is 7 times bigger than the dataset used for Llama 2 as reported by Meta.
981.
by GaggiX 3y ago
>Llama 3 is pretrained on over 15T tokens Well, I remember when training GPT-3 on 300B was a lot.
982.
by GaggiX 3y ago
The real value of Stable Diffusion models are the finetuned models when the base model is released.
983.
by GaggiX 3y ago
That's 25% fewer errors.
984.
Bringing generative AI to video editing workflows in Adobe Premiere Pro(blog.adobe.com)
3 points by GaggiX 3y ago|0 comments
985.
by GaggiX 3y ago
Both Deepl and LLM use transformer decoders, a type of generative models.
986.
by GaggiX 3y ago
I believe that with AC is easier and cheaper to change the voltage using transformers compare to DC.
987.
by GaggiX 3y ago
I wonder if there was a conversion between HDR and standard image format to make the images look so dark, I'm not too familiar with this technology.
988.
by GaggiX 3y ago
It seems to respect it as the majority of the requests are for the robots.txt.
989.
by GaggiX 3y ago
The point is that it also beats the baseline on every other size (1B and 3B). So it wouldn't be surprising to see it beat a 7B transformer model like the 6B model. Note 2 on page 5 probably explains why the sizes are different.
990.
by GaggiX 3y ago
All the MQA transformers, Hawk and Griffin are trained on the same MassiveText dataset so no.

More ›