experiment_name='VarNaming_subtoken_edge_ablation',
experiment_run_log_id=experiment_run_log_id,
model_name='VarNamingFixedVocabGGNN',
model_label='all_edge',
n_workers=8,
n_batch=250 * 4,
evaluation_metrics=('evaluate_full_name_accuracy',
'evaluate_subtokenwise_accuracy',
'evaluate_edit_distance'),
model_params_to_load='best.params',
skip_s3_sync=skip_s3_sync,
test=test),
dict(seed=5145,
gpu_ids=(0, 1, 2, 3),
dataset_name='18_popular_mavens',
experiment_name='VarNaming_subtoken_edge_ablation',
experiment_run_log_id=experiment_run_log_id,
model_name='VarNamingNameGraphVocabGGNN',
model_label='syntax_edge',
n_workers=8,
n_batch=250 * 4,
evaluation_metrics=('evaluate_full_name_accuracy',
'evaluate_subtokenwise_accuracy',
'evaluate_edit_distance'),
model_params_to_load='best.params',
skip_s3_sync=skip_s3_sync,
test=test),
])
run_command_on_remote('local', evaluate_models_for_experiment,
list_of_kwargs)
debug=False)),
(aws_config['remote_ids']['box1'],
dict(dataset_name='18_popular_mavens',
experiment_name='FITB_vocab_comparison',
experiment_run_log_id=experiment_run_log_id,
seed=5145,
gpu_ids=(0, 1, 2, 3),
model_name='FITBGSCVocabGGNN',
model_label='syntax_edge',
model_kwargs=dict(hidden_size=64,
type_emb_size=30,
name_emb_size=31,
n_msg_pass_iters=8),
init_fxn_name='Xavier',
init_fxn_kwargs=dict(),
loss_fxn_name='FITBLoss',
loss_fxn_kwargs=dict(),
optimizer_name='Adam',
optimizer_kwargs={'learning_rate': .0002},
val_fraction=0.15,
n_workers=8,
n_epochs=200,
evaluation_metrics=('evaluate_FITB_accuracy', ),
n_batch=250 * 4,
skip_s3_sync=False,
debug=False)),
]
for instance_id, kwargs in instance_ids_train_kwargs:
time.sleep(1)
run_command_on_remote(instance_id, train_model_for_experiment, kwargs)
# Copyright 2019 Amazon.com, Inc. or its affiliates. All Rights Reserved.
from data.AugmentedAST import syntax_only_excluded_edge_types
from experiments import aws_config
from experiments.make_tasks_and_preprocess_for_experiment import make_tasks_and_preprocess
from experiments.run_command_on_remote import run_command_on_remote
if __name__ == '__main__':
kwargs = dict(seed=515,
dataset_name='18_popular_mavens',
experiment_name='VarNaming_subtoken_edge_ablation',
task_names=['VarNamingTask'],
n_jobs=30,
model_names_labels_and_prepro_kwargs=[
('VarNamingGSCVocabGGNN', 'all_edge_except_subtoken',
frozenset(),
dict(max_name_encoding_length=30,
add_edges=False), dict(max_nodes_per_graph=500)),
('VarNamingGSCVocabGGNN', 'syntax_edge_except_subtoken',
syntax_only_excluded_edge_types,
dict(max_name_encoding_length=30,
add_edges=False), dict(max_nodes_per_graph=500)),
],
skip_make_tasks=False)
run_command_on_remote(aws_config['remote_ids']['box1'],
make_tasks_and_preprocess, kwargs)
'evaluate_subtokenwise_accuracy',
'evaluate_edit_distance',
'evaluate_length_weighted_edit_distance',
'evaluate_top_5_full_name_accuracy',
),
model_params_to_load='model_checkpoint_epoch_23.params',
skip_s3_sync=skip_s3_sync,
test=test),
dict(seed=5145,
gpu_ids=(0, 1, 2, 3),
dataset_name='18_popular_mavens',
experiment_name='VarNaming_subtoken_edge_ablation',
experiment_run_log_id=experiment_run_log_id,
model_name='VarNamingGSCVocabGGNN',
model_label='syntax_edge_except_subtoken',
n_workers=8,
n_batch=250 * 4,
evaluation_metrics=(
'evaluate_full_name_accuracy',
'evaluate_subtokenwise_accuracy',
'evaluate_edit_distance',
'evaluate_length_weighted_edit_distance',
'evaluate_top_5_full_name_accuracy',
),
model_params_to_load='model_checkpoint_epoch_36.params',
skip_s3_sync=skip_s3_sync,
test=test),
])
run_command_on_remote(aws_config['remote_ids']['box1'],
evaluate_models_for_experiment, list_of_kwargs)
