AttributeError: module 'graphlearn' has no attribute 'encoders'

[zdx]

Describe the bug
A clear and concise description of what the bug is.
tensorflow :2.2.0

run node_classification_on_citation.ipynb :
config = {"class_num": 349, # output dimension
"features_num": 130, # 128 dimension + kcore + triangle count
"batch_size": 500,
"val_batch_size": 100,
"test_batch_size":100,
"categorical_attrs_desc": "",
"hidden_dim": 256,
"in_drop_rate": 0.5,
"hops_num": 2,
"neighs_num": [5, 10],
"full_graph_mode": False,
"agg_type": "gcn", # mean, sum
"learning_algo": "adam",
"learning_rate": 0.0005,
"weight_decay": 0.000005,
"epoch": 20,
"node_type": "paper",
"edge_type": "cites"}

train(config, lg)

the error:

AttributeError Traceback (most recent call last)
in
18 "edge_type": "cites"}
19
---> 20 train(config, lg)

in train(config, graph)
27 config["learning_algo"],
28 config["learning_rate"],
---> 29 config["weight_decay"]))
30 trainer.train_and_evaluate()

~/anaconda3/envs/graphscope/lib/python3.7/site-packages/graphscope/learning/graphlearn/python/model/tf/trainer.py in init(self, model_func, epoch, optimizer)
161 epoch,
162 optimizer)
--> 163 self.model = self._model_func()
164
165 def init(self, **kwargs):

in model_fn()
21 node_type=config["node_type"],
22 edge_type=config["edge_type"],
---> 23 full_graph_mode=config["full_graph_mode"])
24 trainer = LocalTFTrainer(model_fn,
25 epoch=config["epoch"],

~/anaconda3/envs/graphscope/lib/python3.7/site-packages/graphscope/learning/examples/tf/gcn/gcn.py in init(self, graph, output_dim, features_num, batch_size, val_batch_size, test_batch_size, categorical_attrs_desc, hidden_dim, in_drop_rate, hops_num, neighs_num, full_graph_mode, node_type, edge_type)
88 self.src_ego_spec = gl.EgoSpec(src_spec, hops_spec=[hop_spec] * self.hops_num)
89 # encoders.
---> 90 self.encoders = self._encoders()
91
92 def _sample_seed(self):

~/anaconda3/envs/graphscope/lib/python3.7/site-packages/graphscope/learning/examples/tf/gcn/gcn.py in _encoders(self)
128
129 depth = self.hops_num
--> 130 feature_encoders = [gl.encoders.IdentityEncoder()] * (depth + 1)
131 conv_layers = []
132 for i in range(depth):

AttributeError: module 'graphlearn' has no attribute 'encoders'

Environment (please complete the following information):

• GraphScope version: main
• OS: Linux
• Version centos7
• Kubernetes Version:1.20.1

Additional context
Add any other context about the problem here.

[sighingnow]

Looks like you encounter a tensorflow's ImportError first, then install tensorflow and continue to run the demo script.

To use models from examples/ in graphscope.learning.examples, you need to install tensorflow first then "import graphscope", i.e., you need to restart the notebook after install tensorflow.

[sighingnow]

Hi @zdx I think that has already been fixed in your environment. Feel free to re-open if you have any further comment.