almightygosu/anr

Code for our CIKM 2018 paper titled "ANR: Aspect-based Neural Recommender"

anr's Issues

Tensor dimension does not match

Hello, I am trying to run the code. I have successfully run the step 1 and processed the data. However, in step 2, when I run the following command.: python3 PyTorchTEST.py -d "amazon_instant_video" -m "ANRS" -e 10 -p 1 -rs 1337 -gpu 0 -vb 1 -sm "amazon_instant_video_ANRS"

I get and error saying tensor dimensionality miss match, Please see attached. Could you please help me to resolve the issue?

Thanks

code error

Hi！ very brilliant job, but i can't run the code

When i use pytorch==0.4.1
Loading pretrained word embeddings from "./datasets/amazon_instant_video/amazon_instant_video_wid_wordEmbed.npy"..
Traceback (most recent call last):
File "PyTorchTEST.py", line 101, in
mdl = mdlZoo.createAndInitModel()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 49, in createAndInitModel
self.initModel()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 81, in initModel
self.initANR()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 109, in initANR
self.loadWordEmbeddings()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 187, in loadWordEmbeddings
self.mdl.wid_wEmbed.weight.data.copy_(torch.from_numpy(np_wid_wEmbed))
RuntimeError: The expanded size of the tensor (50002) must match the existing size (37458) at non-singleton dimension 0

When i use pytorch==0.3.1
Loading pretrained word embeddings from "./datasets/amazon_instant_video/amazon_instant_video_wid_wordEmbed.npy"..
Traceback (most recent call last):
File "PyTorchTEST.py", line 101, in
mdl = mdlZoo.createAndInitModel()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 49, in createAndInitModel
self.initModel()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 81, in initModel
self.initANR()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 109, in initANR
self.loadWordEmbeddings()
File "/data1/sunrui/py_project/ANR/model/ModelZoo.py", line 187, in loadWordEmbeddings
self.mdl.wid_wEmbed.weight.data.copy_(torch.from_numpy(np_wid_wEmbed))
RuntimeError: invalid argument 2: sizes do not match at /pytorch/torch/lib/THC/generic/THCTensorCopy.c:52

thanks !

Some experimental details about parameter setting

Hi, thanks for sharing your implementation.
I tried to reuse your code on 5-core Amazon datasets, but I didn't get the performance as expected.
I set the maximum length of each document to be 500 and padded it if necessary.
And I tried run "ANR" model with and without pretraining, but the result was better without pretraining which is out of my expectation.
I doubt whether I made some mistakes in parameter settings.

DAttn model; Error: self.mdl.cuda(): NoneType' object has no attribute 'cuda'

Hello,

I have tried to compare DAttn with ANR (ANR works fine!) and got this error:

Command: -d reviews_electronics -m DAttn -e 10 -p 1 -rs 98 -gpu 2 -vb 1

Loading 'info' from "/data/anr_files/datasets/reviews_electronics/reviews_electronics_info.pkl"..
'info' loaded!

[INFO] # of Users: 2,586,767, # of Items: 362,819

Creating model (Selected Model: DAttn)..
Traceback (most recent call last):
File "PyTorchTEST.py", line 101, in
mdl = mdlZoo.createAndInitModel()
File "/root/ANR/model/ModelZoo.py", line 39, in createAndInitModel
self.mdl.cuda()
AttributeError: 'NoneType' object has no attribute 'cuda'

Could you please fix it? Thank you in advance!

torch (0.3.0)
Python 3.5.2 (default, Nov 12 2018, 13:43:14)

Variable Defination

hello, I can't find where the following initialized in ModelZoo.

1. at line 151: fp_uid_userDoc
2. at line 152: fp_iid_itemDoc
3. at line 180: fp_wid_wordEmbed
4. at line 182: fp_wordEmbed
   Can you please explain where those variables were defined?
   Thanks