How leonardosabattini/amc_oven works
amc-oven iteratively searches for neural-network parameters that define a bounded temperature schedule. Running opt_step.py samples parameters, converts them into a protocol, and saves optimizer state; an external experiment evaluates the protocol and writes a score for the next run. The README documents this user loop. The optimizer and protocol-generation code are excerpted; the external evaluator is not part of the listed repository.
Main entry points: opt_step.py Optimization runner · aMC_optimizer.py Adaptive Monte Carlo · nets.py Neural networks
Components 9
Experiment orchestration
- Optimization runner
opt_step.py - Experiment score
Parameter search
- Adaptive Monte Carlo
aMC_optimizer.py - Sampled parameters
- Saved optimizer state
Protocol generation
- Neural networks
nets.py - Temperature protocol
Other components
- Experiment operator
- External experiment
Connections 11
- Optimization runner → Adaptive Monte Carlo: calls sampling and step
opt_step.py - Optimization runner → Neural networks: instantiates
opt_step.py - Adaptive Monte Carlo → Sampled parameters: writes samples
aMC_optimizer.py - Optimization runner → Sampled parameters: loads parameters
opt_step.py - Optimization runner → Neural networks: sets weights
opt_step.py - Optimization runner → Temperature protocol: writes protocol
opt_step.py - Experiment operator → External experiment: runs protocol
README.md - External experiment → Experiment score: provides score
README.md - Adaptive Monte Carlo → Experiment score: reads score
aMC_optimizer.py - Optimization runner → Saved optimizer state: saves state
opt_step.py - Optimization runner → Adaptive Monte Carlo: restores state
opt_step.py