This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
This is not sufficient. The rig as described by neonbjb is only 192GB of vram; fine-tuning an LM with 130B params (in the best possible case of GLM-130B; the less said about the shoddy performance of OPT/BLOOM, the better) requires somewhere in the ballpark of ~1.7TB of vram (this is at least 20+ A100s), and that's on batch size 1 with gradient checkpointing and mixed precision and 8bit adam and fused kernels without kv cache and etc. If you don't have an optimised trainer ready to go (or god forbid, you're trying distributed training), you should expect double the requirements.
The cost of that isn't too bad, of course. Maybe $25 bucks an hour on LL, any machine learning engineer can surely afford that. The larger doubt I have is that any of this will take place.
Respectfully, I think GLM-130B is not the right scale for the present-day present-time personal assistant. Ideally, someone (Carper?) would release a 30B or 70B Chinchilla-scaled LM for us to use as a base, but barring that lucky outcome (not sure if carper will be allowed to) I'd go with UL20B or a smaller Flan-T5, or one of several available 10-20B decoder-only models.
In this setting I have in mind, GLM-130B zero-shot prompted with what amounts to our values could be used either as a source of custom base CoT-dialogue finetune dataset or as a critique-generator and ranker in the Anthropic's constitutional AI setting. So, their inference-only config which supports servers as small as 4x RTX3090 could be used. Granted, the performance of GLM-130B in its current raw shape is somewhere between "GPT-3.5" and older Instruct-GPT-3, but it should suffice for the purpose described here.
Wearing my ML engineer hat I could say that while this is a conventional requirement, if we were determined to tune this LLM on a few batches on a given single server, we could use DeepSpeed's Zero-3 offload mode and maybe a bit of custom code to swap most of the parameters to the CPU RAM, which is much cheaper and is surprisingly efficient given large enough batches. One transformer layer worth of VRAM would be enough. One server likely wouldn't be enough for the complete operation, but used infiniband cards and cables are surprisingly cheap.
Regarding the kv cache, I expect the next generation of the transformer-like models to use advanced optimizations which lower kv cache pressure, specifically memorizing transformer. There are other competitive inventions, and discussion of the highest performing stack of tricks to get to the most efficient LM would be interesting, if exhausting.
More options
Context Copy link
More options
Context Copy link