DeepSpeed VS fickling

DeepSpeed can handle parallelism concerns, and even offload data/model to RAM, or even NVMe (!?) . I'm surprised I don't see this project used more.

While on the surface, a 192GB Mac Studio seems like a great deal (it's not much more than a 48GB A6000!), there are several reasons why this might not be a good idea:

* I assume most people have never used llama.cpp Metal w/ large models. It will drop to CPU speeds whenever the context window is full: https://github.com/ggerganov/llama.cpp/issues/1730#issuecomm... - while sure this might be fixed in the future, it's been an issue since Metal support was added, and is a significant problem if you are actually trying to actually use it for inferencing. With 192GB of memory, you could probably run larger models w/o quantization, but I've never seen anyone post benchmarks of their experiences. Note that at that point, the limited memory bandwidth will be a big factor.

* If you are planning on using Apple Silicon for ML/training, I'd also be wary. There are multi-year long open bugs in PyTorch[1], and most major LLM libs like deepspeed, bitsandbytes, etc don't have Apple Silicon support[2][3].

You can see similar patterns w/ Stable Diffusion support [4][5] - support lagging by months, lots of problems and poor performance with inference, much less fine tuning. You can apply this to basically any ML application you want (srt, tts, video, etc)

Macs are fine to poke around with, but if you actually plan to do more than run a small LLM and say "neat", especially for a business, recommending a Mac for anyone getting started w/ ML workloads is a bad take. (In general, for anyone getting started, unless you're just burning budget, renting cloud GPU is going to be the best cost/perf, although on-prem/local obviously has other advantages.)

[1] https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3A...

[2] https://github.com/microsoft/DeepSpeed/issues/1580

[3] https://github.com/TimDettmers/bitsandbytes/issues/485

[4] https://github.com/AUTOMATIC1111/stable-diffusion-webui/disc...

[5] https://forums.macrumors.com/threads/ai-generated-art-stable...

And https://github.com/microsoft/deepspeed

DeepSpeed Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales (https://github.com/microsoft/DeepSpeed/tree/master/blogs/deepspeed-chat)

You're right! You should probably use Trail of Bits Fickling tool to investigate. https://github.com/trailofbits/fickling

I tested the Anything V3 pruned from Hugging Face, and indeed nothing funny in its pickle. I used the Fickling library to decompile it. I do not use Windows so my interests in .ckpt security are largely related to Pickle exploits— which could extract malicious code from a data file and then do something with it, but the data files themselves are not executed. I will edit this comment with lines referencing that data file.

I've been using Diffusion Bee on my Mac, and it's just gained the ability to import models (which it converts), but it is unpickling to do so— but barely. It unpickles, figures out what sort of data is in every data file and then computes what it wants from them on its own. I would love it to not use unpickling at all, so my intention is if I can figure it out, to write a script to decode the pickle file (with Fickling or otherwise) and then just do the weight calculation/assignment.

Something you won't gather from skim-reading the headline is that this is that the author has also created a tool, Fickling: https://github.com/trailofbits/fickling - to aid in playing around with pickle files.

From the article: [Fickling] can help you reverse engineer, test, and even create malicious pickle files.