jofi1

Are H1Bs hired at big tech paid that low anywhere? From what I saw, they may be given the bottom of the job level band, but it's still usually quite high.

At the same time, it was obvious that having h1b engineers gave the managers a lot more power to overwork them and prevented them from speaking out.

Yeah it's absolutely demoralizing and made me leave some teams. Have very little faith in performance reviews and role guidelines

Being the odd one out also makes you a target for things like PIP quotas. It's very disturbing to watch, especially as managers try to justify it with phony evidence to save their own.

I don't have any affiliation with them, but Together is hardly random and very important to the open source community, putting out some of the most useful work around things like new architectures, distributed training, datasets.

The career path document is to try and make the process look objective and save the company from lawsuits. In reality it encourages bsing overly complicated solutions, and promotions are limited by things like budget rather than some objective criteria. The whole notion of this career ladder is pure nonsense, but like a religion to many people.

> It is something I have been wondering about: why did Meta not keep the training process going on while the loss curves seemed to go down? Could they conceivably release a Llama 2.1 being checkpoints taken a month after 2.0 was 'cut'? Maybe the expected gain is too small compared to what can be gained with fine/instruct tuning afterward anyway?

Because choosing the LR decay requires knowing the # of steps in advance. LR is too small after the 2T tokens, and changing it afterwards doesn't tend to help.

https://twitter.com/sherjilozair/status/1687837844729966592

It's often worse than random because it favors the political in-group mafia focused on climbing the ladder through any means necessary, instead of rewarding people doing good work or with actual leadership skills.