I would like to get some help here. I want to use the "weights" of my trained appliance (in this case the boiler of House 4). The dictionary seems to be well defined, but whatever path file I write in the "pre-trained" line is accepted, and my API runs anyway, what makes me think it is actually to taking the trained h5 file. This is very urgent for me since I am running the experiments in a standard PC (no GPU), and otherwise the experiments take forever.
ukdale = {
'power': {
'mains': ['apparent','active'],
'appliance': ['apparent','active']
},
'sample_rate': 6,
#'appliances': ['television','kettle', 'boiler', 'freezer', 'washing machine'],
'appliances': ['boiler'],
'pre-trained': {'Seq2Point': './Weights/boiler_seq2point-temp-weights-45573.h5'},
'methods': {
#'WindowGRU':WindowGRU({'n_epochs':1,'batch_size':32}),
#'RNN':RNN({'n_epochs':1,'batch_size':32}),
#'DAE':DAE({'n_epochs':1,'batch_size':32}),
'Seq2Point':Seq2Point({'n_epochs':1,'batch_size':32, 'sequence_length': 99}),
#'Seq2Seq':Seq2Seq({'n_epochs':1,'batch_size':32}),
'Mean': Mean({}),
},
'train': {
'datasets': {
'UK-DALE': {
'path': './Dataset/ukdale.h5',
#'path': './Weights/boiler_seq2point-temp-weights-45573.h5',
'buildings': {
4: {
'start_time': '2013-04-01 00:00:00',
'end_time': '2013-04-30 00:00:00'
},
# 56: {
# 'start_time': '2015-01-28',
# 'end_time': '2015-01-30'
# },
}
}
}
},
'test': {
'datasets': {
'UK-DALE': {
'path': './Dataset/ukdale.h5',
'buildings': {
4: {
'start_time': '2013-09-02 00:00:00',
'end_time': '2013-09-04 00:00:00'
},
}
}
},
'metrics':['mae', 'rmse']
}
}
I would like to get some help here. I want to use the "weights" of my trained appliance (in this case the boiler of House 4). The dictionary seems to be well defined, but whatever path file I write in the "pre-trained" line is accepted, and my API runs anyway, what makes me think it is actually to taking the trained h5 file. This is very urgent for me since I am running the experiments in a standard PC (no GPU), and otherwise the experiments take forever.
I took the "pre-trained" line from a closed issue in the NILMTK repo here.
Appreciate your support!