def __init__(self, name):
self._ffconfig = ff.FFConfig()
print("Python API batchSize(%d) workersPerNodes(%d) numNodes(%d)" %
(self._ffconfig.batch_size, self._ffconfig.workers_per_node,
self._ffconfig.num_nodes))
self._ffmodel = None
self._name = name
self._ffoptimizer = None
self._layers = []
self._nb_layers = 0
self._input_layers = []
self._input_tensors = []
self._output_tensor = 0
self._label_tensor = 0
self._num_samples = 0
self._input_dataloaders = []
self._input_dataloaders_dim = []
self._label_dataloader = 0
self._label_dataloader_dim = 0
self._loss = None
self._metrics = []
self._label_type = ff.DataType.DT_FLOAT
self._layer_inited = False
global tracing_id
self.__tracing_id = tracing_id
tracing_id += 1
def __init__(self, name):
self._ffconfig = ff.FFConfig()
self._ffconfig.parse_args()
print("Python API batchSize(%d) workersPerNodes(%d) numNodes(%d)" %
(self._ffconfig.get_batch_size(),
self._ffconfig.get_workers_per_node(),
self._ffconfig.get_num_nodes()))
self._ffmodel = None
self._name = name
self._ffoptimizer = None
self._layers = []
self._nb_layers = 0
self._input_layers = []
self._input_tensors = []
self._output_tensor = 0
self._label_tensor = 0
self._full_input_tensors = []
self._full_label_tensor = 0
self._num_samples = 0
self._input_dataloaders = []
self._input_dataloaders_dim = []
self._label_dataloader = 0
self._label_dataloader_dim = 0
self._loss = None
self._metrics = []
global tracing_id
self.__tracing_id = tracing_id
tracing_id += 1
def __init__(self, inputs, onnx_model):
self._ffconfig = ff.FFConfig()
self._ffconfig.parse_args()
print("Python API batchSize(%d) workersPerNodes(%d) numNodes(%d)" %
(self._ffconfig.get_batch_size(),
self._ffconfig.get_workers_per_node(),
self._ffconfig.get_num_nodes()))
self._ffmodel = None
self._onnx_model = onnx_model
for node in onnx_model.graph.node:
print(node)
for input in onnx_model.graph.initializer:
print(input.name, input.dims, len(input.dims))
# for input in onnx_model.graph.input:
# print(input)
self._input_tensors = []
for key in inputs:
input_tensor = inputs[key]
t = Tensor(ffconfig=self._ffconfig,
key=key,
shape=input_tensor.shape,
dtype=input_tensor.dtype)
self._input_tensors.append(t)
self._loss = None
self._label_type = None
self._metrics = []
self._label_type = ff.DataType.DT_FLOAT
self._my_onnx_model = None
self._output_tensor = None
self._full_input_tensors = []
self._full_label_tensor = 0
self._num_samples = 0
self._input_dataloaders = []
self._input_dataloaders_dim = []
self._label_dataloader = 0
self._label_dataloader_dim = 0
global tracing_id
self.__tracing_id = tracing_id
tracing_id += 1
def __init__(self):
self.ffconfig = ff.FFConfig()
self.ffconfig.parse_args()
print("Python API batchSize(%d) workersPerNodes(%d) numNodes(%d)" %
(self.ffconfig.get_batch_size(),
self.ffconfig.get_workers_per_node(),
self.ffconfig.get_num_nodes()))
self.ffmodel = ff.FFModel(self.ffconfig)
self.ffoptimizer = 0
self._layers = dict()
self._nb_layers = 0
self.input_tensors = []
self.output_tensor = 0
self.label_tensor = 0
self.full_input_tensors = []
self.full_label_tensor = 0
self.num_samples = 0
self.input_dataloaders = []
self.input_dataloaders_dim = []
self.label_dataloader = 0
self.label_dataloader_dim = 0
def __init__(self): super(Module, self).__init__() self._ffconfig = ff.FFConfig() self._ffconfig.parse_args() self._ffmodel = ff.FFModel(self._ffconfig) self._graph = None
