News for hackers. Stuff that matters.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by spudlyo on Sunday April 30, 2023 @12:14PM (#35763963)
I'm running this on my iPhone 13 Pro Max as part of the Test Flight beta, and it's interesting. I don't believe I've ever run anything that's ever pushed my phone this hard, and you can feel the heat. The text output performance is pretty inconsistent, it was very fast at first but slowed down considerably after a few answers. In terms of quality, it's prone to hallucinations, which is not unexpected based on the size and highly compressed nature of the model.

"The marvel is not that the bear dances well, but that the bear dances at all."

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by junrushao1994 on Sunday April 30, 2023 @12:23PM (#35764060)
Thanks for sharing! It's definitely a bit painstaking to get a real-world LLM running at all on an iPhone due to memory constraint. It's also quite compute-intense as it has 7B parameters, but we are glad that it's generating texts at reasonable speed!

The model we are using is a quantized Vicuna-7b, which I believe is one of the best open-sourced models. Hallucination is a problem to all LLMs, but I believe research on model side would gradually alleviate this problem :-)

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by Fr0styMatt88 on Sunday April 30, 2023 @05:01PM (#35766594)
WizardLM-7b would be a fantastic model to try out as well. Though that might be out of date tomorrow (or already!) given how many models are being released at the moment :)
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by david-gpu on Sunday April 30, 2023 @12:58PM (#35764429)
> I don't believe I've ever run anything that's ever pushed my phone this hard, and you can feel the heat. The text output performance is pretty inconsistent, it was very fast at first but slowed down considerably after a few answers.

Those two events are causally related. The OS has to throttle down the CPU or else it will overheat and malfunction.

It is one of the reasons why heavy number crunching is often performed on the cloud instead.

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by capableweb on Sunday April 30, 2023 @01:02PM (#35764468)
In my experience, heavy number crunching is more suitable to run on dedicated machines, rather than virtualized "cloud" cores. More consistent performance, no noisy neighbors and cheaper in the long term.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by halflings on Sunday April 30, 2023 @01:19PM (#35764606)
> cheaper in the long term

Citation needed :)) There are economies of scale and various optimizations that are just not possible with dedicated machines.

Anecdotal evidence: a company I worked in had a dedicated DC with hundreds/thousands of machines that mostly ran SQL queries on petabytes of data (any query would take ~5-30 minutes). Eye-watering budget and whole teams to maintain the cluster... They switched to GCP/BigQuery, got queries that ran in seconds at a fraction of the budget.

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by slimsag on Sunday April 30, 2023 @03:13PM (#35765680)
The inverse is also true; I don't have to pay Google or AWS to deploy my software to dedicated consumer devices. The user already paid for it.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by Negitivefrags on Sunday April 30, 2023 @03:36PM (#35765889)
This anecdote is the opposite of the ones I usually hear.

Economies of scale go in the other direction usually.

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by taneq on Sunday April 30, 2023 @08:12PM (#35767755)
Those economies are for the cloud provider, not you. The cost to you is set at just less than their best guess at what you think self-hosting will cost you.

Cloud hosting only makes sense in a case that your capacity needs are changing rapidly and unpredictably, or the case where you’re so big that the cloud hosting company is effectively a department. Any other time it will almost certainly be cheaper to self host.

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by rvnx on Sunday April 30, 2023 @07:33PM (#35767540)
Sounds like they could have switched to Clickhouse if they already had the hardware.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by david-gpu on Sunday April 30, 2023 @02:12PM (#35765093)
"The cloud" simply means remote servers. Those servers may or may not contain application-specific hardware acceleration.

Nowadays many cell phone application processors also have dedicated hardware to accelerate neural nets, but they will always be limited by thermal constraints.

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by smoldesu on Sunday April 30, 2023 @02:13PM (#35765115)
If you can saturate it, sure. The average person is not going to be running 24/7 AI compute tasks though, so paying for a month of A100 dedi when you only need 10-20 minutes of compute/week is a complete waste.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by trifurcate on Sunday April 30, 2023 @01:45PM (#35764838)
The slowdown here is going to be more about the increased context length than throttling.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by david-gpu on Sunday April 30, 2023 @02:12PM (#35765104)
Good point! Didn't think of that.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by kiratp on Sunday April 30, 2023 @03:08PM (#35765633)
You can do Stable Diffusion on your phone too, fully local.

https://apps.apple.com/app/id6444050820

Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by andrewaylett on Sunday April 30, 2023 @04:24PM (#35766279)
I'd suggest that everything output by an LLM is a hallucination, but it often enough hallucinates something that resembles reality.
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by strudey on Sunday April 30, 2023 @07:43PM (#35767583)
True for our minds too I think ;)
Re:MLC-LLM: GPT/Llama on consumer-class GPUs and phones by BoorishBears on Sunday April 30, 2023 @10:05PM (#35768400)
It honestly worries me when I see people paint the whole "everything is a hallucination" angle as uniquely applying LLMs...

If you're going to be that lose with your definition of hallucination, I'd really hope you apply it to yourself. That's a level of introspection that you need to avoid letting biases that all of us have deep down end up affecting your higher order function.

We humans rely on a ball of biases and interpolations and extrapolations to function, LLMs don't have a monopoly on that.