| date | section | title |
|---|---|---|
2021年 九月 1日 星期三 |
3 |
galileo.framework.python.base_unsupervised.BaseUnsupervised |
galileo.framework.python.base_unsupervised.BaseUnsupervised - base unsupervised model for graph
\
继承自 metaclass , 以及 ABCMeta .
被 galileo.framework.pytorch.python.unsupervised.Unsupervised , 以及 galileo.framework.tf.python.unsupervised.Unsupervised 继承.
def __init__ (self, *args, **kwargs)\
def target_encoder (self, inputs)
unsupervised target encoder
def context_encoder (self, inputs)
unsupervised context encoder
def compute_logits (self, target, context)
compute logits
def loss_and_metrics (self, logits, negative_logits)\
def convert_ids_tensor (self, inputs)
convert ids tensor
def convert_features_tensor (self, inputs)
convert features tensor
def unpack_sample (self, inputs)
unpack sample
def __call__ (self, inputs, **kwargs)\
base unsupervised model for graph
including target embedding and context embedding
参数
inputs dict of tensors,
参见
unpack_sample
返回
a dict of ids and embeddings if only_embedding is True, otherwise return a dict of loss and metrics
def galileo.framework.python.base_unsupervised.BaseUnsupervised.loss_and_metrics ( self, logits, negative_logits)
返回
a dict of loss and metrics
被 galileo.framework.tf.python.unsupervised.Unsupervised , 以及 galileo.framework.pytorch.python.unsupervised.Unsupervised 重载.
unpack sample
参数
inputs dict, keys of dict:
case 1: (target,) for save embedding, target and target_ids are same
case 2: (target, target_ids) for save embedding, target is features of target_ids
case 3: (target, context, negative) for train and evaluate, all is features
返回
target, target_ids, None, only_embedding=True
target, context, negative, only_embedding=False
由 Doyxgen 通过分析 Galileo 的 源代码自动生成.