REDUCE ONLINE DPO MEMORY CONSUMPTION WITH UNSLOTH QLORA (OPEN SOURCE) Find my open source contributions and details about how you can reduce VRAM usage for ONLINE DPO finetuning below in my substack blog! (All code in the blog!) I believe that more official Online RLHF support for Unsloth will be coming out sometime soon. Special thanks to the Unsloth AI team (Daniel Han and Michael Han (Unsloth)), Edward Kim and Costa Huang for helping me make this project come together! blog: https://lnkd.in/eCej2J5q
Such an imbalanced y_train 😂
hell yeah!
🤯
👏
Love this thanks Keith Truongcao! :)
Very informative keith!
Very interesting VRAM optimization with Unsloth. Great approach!
Stellar work Keith! Props to you and the Unsloth AI team for your collective perserverance.
Great work!
this is huge for the employed