It's very sad how illiberal hackernews gets when it comes to H-1B visas and immigration in general. People will say that they are concerned about the welfare of H-1B holders because they can't leave their jobs but I've literally never heard anyone suggest an improvement (better portability, etc). Instead all anyone talks about is how bad it is that they have to compete against people from around the world. I think people are just afraid of a true meritocracy and maybe that's a bearish signal for the industry overall.
Are there types of disability insurance you can buy that would cover you as a software engineer in case something like this happens and you're not able to go back to work or not work as effectively in this field? Does anyone have experiences with this?
Maybe I ate too much marketing but it does feel like having the PS5 support SSDs raised the bar for how fast games are expected to load, even across platforms.
What are some good ways to participate if you don't have an HF radio? (Alternatively, what's a good way to get into HF if you live in a small apartment)?
I'm a Canadian who moved to the SF bay area after graduating. A lot of my smartest friends who came with me at the same time are actively taking steps to move back due to the political environment.
I would love to build something like this, but it looks like the project is not quite complete yet? And also I don't see anywhere on the page where I can sign up for updates.
I was hit by this while working on a project for class and it was the most frustrating thing ever. The bot would completely hallucinate functions and docs and it confused everyone. I found one post where someone did the simple prompt injection of "ignore previous instructions and x" and it worked but I think it's delted now. Swore off ultralytics after that.
We definitely tried extending the lookahead, but I don't think we tried having a curriculum-style thing where we gradually increased the speed of the song. Great idea though!
We were able to make some improvements by tuning how the reward is distributed and also by first pretraining the agent on scales before fine-tuning them on the final pieces.
Thanks to Kevin Zakka for helping us get started with the RL environment!
Static linker supporter here, not to start a flame war but I like static linking because it means I don't have to worry about version mismatches of shared libraries between what I ship and what the user gets.
Without the actual training corpus, you can't know how much it takes to train. They could have trained on twice as many tokens for example (not saying they did!)
It's been a while since I took electrostatics, but I don't understand the theory behind this. If the rolls become charged and you are presumably neutral, wouldn't they attract you rather than repel you? That's what makes me think this story is apocryphal.