Open Dranero opened 3 years ago
The Speedup options speed up an
{ "environment": "ReacherMemory-v0", "random_seed": 1, "number_generations": 100, "optimizer": { "type": "CMA_ES", "population_size": 100, "sigma": 1.0, "checkpoint_frequency": 0, "hof_size": 5, "mu": 25, "efficiency_weight": 0.0, "fix_seed": "false" }, "brain": { "type": "ELMAN", "hidden_layer_structure": [10], "diagonal_hidden_to_hidden": true, "use_bias": true, "optimize_initial_neuron_state": false }, "episode_runner": { "number_fitness_runs": 25, "environment_attributes": { "type": "ReacherMemoryAttr", "observation_frames": 20, "memory_frames": 20, "action_frames": 50 } } }
up by ~3%
Still in progress
Added Future based mp handler Consolidated multipreseesing tasks Refactored seed usage Added proper genom dicts for layered Brains
TODO *Messure speedup in comparison to master