Next few videos will be about finetuning SLMs.❤️‍🔥 - Originally I wanted a 60 min vid covering model dev after pretraining (CPT, Midtraining, SFT, RLHF, RLVR) - But that would take me months, so instead I'm doing a mini-series of ~20 min vids. - Unsloth throughout with some HF and Pytorch where necessary. RL part will use Prime Labs. - Code aside, the major major major focus will be datasets. identifying OS datasets, preparing them, generating synthetic datasets for training, etc - Another major major will be about evaluation. Setting up harnesses for metrics. - Topics of later videos are not fully planned yet, screenshot is basic roadmap. I will play it by ear according to what questions I get from yall! - I will most definitely be working on other videos in between. If my mind wanders I'll let it. But I hope I finish it in the next 3 months or so. It all depends on my enjoyment, what else is happening in the world, and what my audience wants to engage with. That said, my next video will be 100% about continued pretraining on custom datasets.