I am facing an error while running demo on local 'NoneType' object has no attribute 'loc

[sainiudit]

Table topic exception: dtype('float32')

AttributeError Traceback (most recent call last)
in ()
----> 1 res = evaluate(df.head(10))

1 frames
/content/drive/MyDrive/columns_classification/sato/demo/predict.py in evaluate(df)
83 def evaluate(df):
84
---> 85 feature_dic, labels, mask = extract(df)
86
87 emissions = classifier(feature_dic).view(1, MAX_COL_COUNT, -1)

/content/drive/MyDrive/columns_classification/sato/demo/predict.py in extract(df)
64 # topic vectors
65 topic_features = extract_topic_features(df_dic)
---> 66 topic_vec = pad_vec(topic_features.loc[0,'table_topic'])
67 feature_dic['topic'] = torch.FloatTensor(np.vstack((np.tile(topic_vec,(n,1)), np.zeros((MAX_COL_COUNT - n, topic_dim)))))
68

AttributeError: 'NoneType' object has no attribute 'loc'

[Face1essboy]

I also met the same problem,is it solved now?

[varlen]

In case anyone is trying to run it, I've successfully ran it using an older version of Python within a docker container. I've created a Dockerfile to make it reproducible. Since my use case includes calling the model through a HTTP endpoint, the server.py invocation is hardcoded in the CMD from the Dockerfile but it is simple to adapt for other purposes.

https://github.com/varlen/sato-env

Invoke the ./build.sh script then ./run.sh <sato-repo>

[taziksh]

@varlen many thanks