Mia Wen

遇到了相同的问题，ffm跑出来的embedding是按特征计算的，请问这个问题后来解决了吗

you can use load_state function and set the number of features to use the model file. For example: **model2 = TFFMClassifier( order=2, rank=32, optimizer=tf.train.AdamOptimizer(learning_rate=0.001), n_epochs=3, batch_size=512, init_std=0.01, reg=0.01, input_type='sparse', seed=1...