One of the big questions about Hugging Face accelerate during distributed PyTorch training is how do you optimize your DataLoaders to make use of your multiple GPUs.

Happy to share this all with you via another wonderful animated tutorial! youtube.com/watch?v=9Vfauv…

Next great microsoft release Microsoft Research ? 🤗

Twitter needs custom emojis like discord lol

Can't stop me Sparkling Ice!

My 4x NVME RAID 0 has been restored 🕺

Let's give er' a go!

Synthetic Data Synthetic Data Synthetic Data

Synthetic Data Synthetic Data Synthetic Data

I still can't believe microsoft put this guy in charge of anything lol

Confirmed all my nvme drives are safe and working.. Now just have to see if replacing mobo with exact copy will bring the RAID 0 config back to life

Great insights

.AMD Lisa Su what's the real problem here afraid of opensource or something? Come on we are all dying for you to become a competitive force...

Big data

I’m in pretty much total agreement

I think I destroyed my computer can anybody help lmao. Liquid spilled right onto an nvme shield on the motherboard, screen went black, left it off for 8 hours it started boot cycling, then completely stopped accepting power. The drives were 4x in raid 0. Is this recoverable

Help

Interesting. I would like to see GPT-5 before concurring

I think Meta and Llama-3 is the final nail in the coffin to several misconceptions I've been fighting against for the last year.

Llama-3 Chat was trained on over 10M Instruction/Chat samples, and is one of the only finetunes that shows significant improvements to MMLU.