A brief roadmap for 2025
A brief overview of the roadmap for 2025, including key milestones and goals.
Welcome to MoaM
Learn
Intelligently
MoaM is directly attacking the final frontier of intelligence. We are training models as optimizers, and pursuing the elusive idea of an optimizer that learns to optimize itself.
# Load initial parameters mod_params = load_params("target_model.pkl") opt_params = load_params("learned_optimizer.pkl") # Training loop for epoch in range(num_epochs): env = get_env() # Get updates from learned optimizer delta = optimize(opt_params, mod_params, env) # Update model parameters mod_params = apply_updates(mod_params, delta)