MacBook Pro M4 Max Chip with Python Install and LLMs

[@IoannisNousias]

Would be nice to see how it fares when training small models, compared your 4090s. Can it be used for small experiments before, say, posting it to a GPU server?

[@andresv8743]

it would be nice to have elastic models. A model that starts tiny and lives in your personal computer or cloud. It should have "reasoning" to realize when it doesn't know something. As you interact then it can go learn from bigger models and grow as needed. It just feels inefficient that millions of people are using the same model as me.

[@apm]

It is so cost-effective to run a small language model that this strategy makes a lot of sense. You could also run a couple small language models, and if they don’t agree, use a larger model to gain consensus or a better answer.

[@petermiller1946]

dude! thank you for making this video. very helpful

[@apm]

I am glad the video was helpful!

[@refixed]

Is there a reason you went with the 14" binned M4 Max? How's it working? I've been trying to decide on what to get -- I ordered a 14" M4 Max with 64gb this morning, but I considered the binned version long and hard...

[@DS-pk4eh]

Well with lower version of Max chip, you are limited with the max RAM you can have. So it is not really a free choice if you want high amount of the RAM. Probably something Apple artificially made so they would sell full Max chip. PS: By the way, full Max chip is just that. Not binned. The lower ones are probably some binned full Max chips with one or two cores damaged, and then deactivated.

[@FishStickington]

@@DS-pk4eh Not entirely correct, all chips, both the full Max and lower core versions, are binned, as in they are all subjected to the binning process. This is something that all CPU and GPU manufacturers have been doing for a long time. Apple fans just found out about it after Apple started making their own silicon and have hijacked to term to mean a less-than-full-spec chip, but this is nothing new to the industry, and it's contrary to how to the term has been used historically. An intel i5 is just a cut down i7 or i9, but that's implied because that's how CPUs are made, so if someone says "binned i5" implies that it was actually compared to other i5s and proven to perform slightly above average compared to the others, i.e. a BETTER than average i5, but realistically all chips are subject to a binning process at some point. Turning off faulty cores to sell a lower model CPU isn't anything new. A lower spec isn't any more or less "binned" by Apple because they have to put all of them through the binning process to figure which ones will be the 14c versions or 16c. So while calling a 14c Max the "binned" version isn't entirely false, because they're all binned, it also doesn't make much sense, because they're all binned. They don't sell a CPU that hasn't been subject to the binning process, its just a matter of whether or not they decide to disable any cores based on the results of the binning process, so people should really just be referring to THAT, the # OF CORES, to specify spec versions. Saying 14c Max is just as easy, if not easier, than saying binned, its much more clear, and it makes more sense. They are ALL binned just say the spec

[@nascentnaga]

shoulda maxed the max with 128 GB RAM :D

[@seattledude2022]

fed up with videos talk about nano texture and video editing on macbook. Get some real work done, people. The so called content creation are mostly Internet junk creation

[@apm]

Yes, need more reviews on ML / Data Science work. Hopefully this video contributes.