huangfu/InteractionImitation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

More policy iterations per epoch

Browse Source
This commit is contained in:
ebuehrle
2022-02-28 22:39:28 +01:00
parent a9314c4657
commit 261e35eae8
1 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
shail-experiment.py
View File

@@ -185,11 +185,11 @@ if __name__ == '__main__':
'learning_rate': 3e-4,
'learning_rate_decay': 1.0,
'clip_ratio': 0.2,
'iterations_per_epoch': 100,
'hidden_layer_size': tune.grid_search([20, 40]),
'n_hidden_layers': tune.grid_search([2, 3]),
'iterations_per_epoch': tune.grid_search([250, 500, 750]),
'hidden_layer_size': 40, #tune.grid_search([20, 40]),
'n_hidden_layers': 3, #tune.grid_search([2, 3]),
'activation':0,
'option': tune.grid_search(list(range(len(option_list))))
'option': 0, #tune.grid_search(list(range(len(option_list))))
},
'value': {
'learning_rate': 1e-3,
@@ -199,8 +199,8 @@ if __name__ == '__main__':
'learning_rate': 1e-3,
'weight_decay': 1e-4,
'iterations_per_epoch': 100,
'n_hidden_layers_element': tune.grid_search([3,4]),
'n_hidden_layers_global': tune.grid_search([1,2]),
'n_hidden_layers_element': 3, #tune.grid_search([3,4]),
'n_hidden_layers_global': 2, #tune.grid_search([1,2]),
'hidden_layer_size': 10,
'activation': 0,
},