thanks to the Thinking Machines team, we used Tinker to prototype our reward models and train the prompt expander via RL.
for more information, read the full technical report on the data, architecture, and training behind Krea 2 👇 https://t.co/okdxlqJuzA
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
Why it matters
Krea trained the prompt expander for its image model with reinforcement learning and prototyped the reward models on Tinker, a worked example of applying an RL fine-tuningTaking a trained model and training it a bit more on your own examples so it gets better at one specific job.Full definition → service inside an image generation pipeline.
