Three passions, simple but overwhelmingly strong, have governed my life: the longing for love, the search for knowledge, and unbearable pity for the suffering of mankind. These passions, like great winds, have blown me hither and thither, in a wayward course, over a great ocean of anguish, reaching to the very verge of despair.
I have sought love, first, because it brings ecstasy - ecstasy so great that I would often have sacrificed all the rest of life for a few hours of this joy. I have sought it, next, because it relieves loneliness--that terrible loneliness in which one shivering consciousness looks over the rim of the world into the cold unfathomable lifeless abyss. I have sought it finally, because in the union of love I have seen, in a mystic miniature, the prefiguring vision of the heaven that saints and poets have imagined. This is what I sought, and though it might seem too good for human life, this is what--at last--I have found.
With equal passion I have sought knowledge. I have wished to understand the hearts of men. I have wished to know why the stars shine. And I have tried to apprehend the Pythagorean power by which number holds sway above the flux. A little of this, but not much, I have achieved.
Love and knowledge, so far as they were possible, led upward toward the heavens. But always pity brought me back to earth. Echoes of cries of pain reverberate in my heart. Children in famine, victims tortured by oppressors, helpless old people a burden to their sons, and the whole world of loneliness, poverty, and pain make a mockery of what human life should be. I long to alleviate this evil, but I cannot, and I too suffer.
This has been my life. I have found it worth living, and would gladly live it again if the chance were offered me.
This is asking the wrong question: for example, there are antibiotics you are prescribed for IBS because generally they tend to have a balancing effect on the gut microbiome. I've done a course every year or two when things get unmanageable, since it's a good reset.
In this context, a treatment which reduced risk would likely be a combination of antibiotics and probiotics to try and shift the biome to a different one.
This is antibiotic marketing bs. There are antibiotic that harm pathogenic or prone-to-overgrowth bacteria harder than others, but they ALL lead to a state of dysbiosis and susceptibility to subsequent colonization/overgrowth/dysregulation
And this is the exact reason whymost of the time, GI problems come back or are replaced with different symptoms after antibiotics, and why we need more Fecal Microbiota Transplant research.
No, it isn't. The whole point of antibiotics for IBS treatment is that you already have an overgrowth, and you want to disrupt that and level the playing field.
This is literally the reason I've been on short course antibiotics (<1 week or a single 4 tab dose) at different points, because it eliminates the constant bloating and gives probiotics a chance to work.
The exact reason why overgrowth happens is because an opportunistic strain takes advantage of a dysbiotic state - even if you suppress it with antibiotics, your gut still lacks the robustness of composition necessary to be resilient. In fact, no antibiotic is selective enough to not impoverish your microbiome.
Probiotics might help, but a few or a few dozen non-gut-sourced strains simply cannot compare to the hundreds present in healthy stool. The effects on microbiota-dependent conditions are often weak to none.
tl;dr: Long-term harm that might happen to kill off whatever's harming you in the process, but don't count on it.
Long version: It depends on specific antibiotic- generally, they wipe a portion of the strains completely, suppress most of the rest, and give anything resistant the opportunity to grow out of control, a state which persists for weeks to months during which you're extra susceptible to external pathogens, food poisoning, etc. A healthy microbiome is resilient, and sadly if you don't have one, FMT is the only way to do it.
Probably they would be close - M1 still needs to use memory for the OS and other stuff, while 3090 can use fp16/mixed precision, which in many cases almost doubles effective memory. Also if we talk about training, then a more mature CUDA implementation of things like batch normalization and optimizers can also result in lower memory usage compared to a likely less mature TF Metal support.
The MacBook Pro M1 Max can be ordered with either 32 GB or 64 GB of unified memory. The geekbench report shows 64 GB of memory (maxed), not sure why only 42.7 GB is usable by OpenCL--so I guess we have to assume that's the max, unless there's some software fix to get it up to 64 GB.
> unless there's some software fix to get it up to 64 GB.
I very much doubt that this would even be possible. The OS and other on-chip devices require dedicated memory as well and need to reserve address space.
You simply cannot max out physical RAM with a single device in a UMA configuration.
It’s all about the RAM. 64GB would allow input of larger image sizes and/or nets with more parameters. Right now, the consumer card with the most RAM is the rtx 3090 which is only 24GB, and in my opinion overpriced and inefficient in terms of wattage (~350W). Even the ~$6000 RTX A6000 cards are only 48GB.
I don't think replacing a workstation with a Macbook because of RAM makes too much sense: If running one minibatch of your model already takes up all the memory you have, where would the rest of your training data sit? In the M1, you don't have a separate main memory.
Also, software support for accelerated training on Apple hardware is extremely limited: Out of the main frameworks, only tensorflow seems to target it, and even there, the issues you'll face won't be high on the priority list.
I know that nvidia GPUs are very expensive, but if you're really serious about training a large model, the only alternative would be paying rent to Google.
Good points. I look forward to some benchmarks. Just hoping for an alternative to Nvidia sooner than later. Dreaming apple will solve it and offer a 1.5 TB mac pro.
Their comparison charts showed the performance of mobile GPUs, not the desktop ones. So, I wouldn’t call this “practical”. Most likely depends on what kind of models you are building and what software you use and how optimized it is for M1.
It will be definitely handy for finetuning large models due to huge ram, but in training from scratch 3090 is certainly better. They are seemingly cooking 128 core GPU, if they release this kraken it will beat 3090 in pretty much everything.
