kirdiekirdie

Magic the Gathering (MTG) podcasts could really benefit from a video version where the discussed cards are shown but creating such a video takes hours or even days.

I made a slide show generator which only takes minutes.

Input is lines of text with MTG card names separated by the | character.

Load one of the presets and click "Start Slideshow", then you can screen record and navigate slides with the arrow characters.

It would suck to be included in the trial and then be randomly selected into the placebo group.

I never understand those articles. I have been doing everything with a normal bike my whole life, including shopping and bringing the kids to school and kindergarden, and I see no reason to let my body atrophy by pointlessly buying an expensive bike with a motor. Then I probably need to spend even more money and time on a bike machine in a fitness studio and more money on doctors after suffering from weight related illnesses. Those articles always assume that car is the god given default state. It's great for the elderly and for my colleague who is suffering from long covid but everyone else I know is doing perfectly fine with a bike without motor including my grandpa until he got around 85 years old.

How do you finetune a dataset on LLaMA2-7b with an A10 with 24 GB? I thought you need to load the dataset 4 times for training, so with 16 bit weights this would be 7 * 10^9 * 2 bytes * 4 = 56 GB GPU RAM. Have you found a way to train with 4 bit weights? And would a single textbook be enough text for a meaningful finetuning? I am worried that a model that is trained on billions of documents will not adapt strongly enough to a comparatively small document.