search

guyulongcs / CIKM2020_DMT

Deep Multifaceted Transformers for Multi-objective Ranking in Large-Scale E-commerce Recommender Systems, CIKM 2020
109 stars 25 forks source link

There are too many bugs in the codes #10

Open LinaZhang opened 2 years ago

LinaZhang commented 2 years ago

I cannot run the codes successfully. When using the default configuration "ConcatOp : Dimensions of inputs should match" appears. While when I changed model_type to others all failed. Even if model_type = mlp, it failed as follow:

Traceback (most recent call last): File "./run_dnn.py", line 912, in train(wnd_conf, args['model_ckpt']) File "./run_dnn.py", line 154, in train tower_train_logits = inf.inference(tower_batch_features, is_train=True) File "/notebook/dmtfq/CIKM2020_DMT/DMT_code/model/inference_mlp.py", line 118, in inference return self.model.inference(inputs,is_train,is_predict) TypeError: inference() takes from 2 to 3 positional arguments but 4 were given

when I modify mlp inference() function parameters, it run next few steps but failed at last.

Yuqi-Zhou commented 1 year ago

I solve this bug by deleting some data about emb and attention_embed in conf/settings/conf, because there are some misssing data provided by the author.By the way, there are some bugs in ohter models like mmoe and mlp.