To turn the CPRS baseline model

[seongeunryu]

Hello, thank you for introducing a great model.

I would like to run the CPRS baseline model, and I'm wondering if I just need to run 'python CPRS.py'. If not, please let me know the correct way to run it. Also, I couldn't find the data needed to build embdict[word], which is /home/sansa/dataset/no.tsv. Could you share it with me? I am trying to run it using glove.840B.300d.txt, but various errors occur.

Thank you for your reply.

[summmeer]

Yes, I guess you can directly run it. The missing file is here.

[seongeunryu]

Thank you very much for your prompt response. Would it be possible to also receive the '/home/sansa/dataset/Adressa/articles_category.pkl' file necessary for running the CPRS.py file? I apologize for taking up your time.

Also, in the process of trying to run CPRS, I replaced the 'articles_titles_2.pkl' file mentioned in the code as data for using articles_content with 'articles_titles_4.pkl'. Is there any problem with this? I couldn't obtain the 'articles_titles_2.pkl' file.

[summmeer]

The category file is easy to obtain, plz refer to https://github.com/summmeer/session-based-news-recommendation/issues/8#issuecomment-1334697445 The second question: yes, it's ok to do the replace

[seongeunryu]

Thank you so much for explaining kindly. Sorry to bother you again. Thanks to the information you provided, I was able to run the CPRS.py code. However, an Out Of Memory error occurred and the training did not proceed.

I have two TITAN Xp 12GB graphics cards, but even when I set the batch size to 1, the following error occurred...

2023-04-23 06:50:20.307285: W tensorflow/c/c_api.cc:300] Operation '{name:'loss/AddN' id:3198 op device:{requested: '', assigned: ''} def:{{{node loss/AddN}} = AddN[N=2, T=DT_FLOAT, has_manual_control_dependencies=true](loss/mul, loss/mul_1)}}' was changed by setting attribute after it was run by a session. This mutation will have no effect, and will trigger an error in the future. Either don't modify nodes after running them or create a new session. 2023-04-23 06:50:20.399976: W tensorflow/c/c_api.cc:300] Operation '{name:'training/Adam/dense_2/kernel/v/Assign' id:4898 op device:{requested: '', assigned: ''} def:{{{node training/Adam/dense_2/kernel/v/Assign}} = AssignVariableOp[has_manual_control_dependencies=true, dtype=DT_FLOAT, validate_shape=false](training/Adam/dense_2/kernel/v, training/Adam/dense_2/kernel/v/Initializer/zeros)}}' was changed by setting attribute after it was run by a session. This mutation will have no effect, and will trigger an error in the future. Either don't modify nodes after running them or create a new session. 2023-04-23 06:50:21.799618: W tensorflow/tsl/framework/cpu_allocator_impl.cc:83] Allocation of 6913350000 exceeds 10% of free system memory. 2023-04-23 06:50:23.295928: W tensorflow/tsl/framework/cpu_allocator_impl.cc:83] Allocation of 6913350000 exceeds 10% of free system memory. 2023-04-23 06:50:24.794740: W tensorflow/tsl/framework/cpu_allocator_impl.cc:83] Allocation of 6913350000 exceeds 10% of free system memory. 2023-04-23 06:50:26.830885: W tensorflow/tsl/framework/cpu_allocator_impl.cc:83] Allocation of 6913350000 exceeds 10% of free system memory. 2023-04-23 06:50:29.420800: W tensorflow/tsl/framework/cpu_allocator_impl.cc:83] Allocation of 6913350000 exceeds 10% of free system memory. 2023-04-23 06:50:50.141729: W tensorflow/tsl/framework/bfc_allocator.cc:485] Allocator (GPU_0_bfc) ran out of memory trying to allocate 6.44GiB (rounded to 6913350144)requested by op training/Adam/gradients/gradients/zeros If the cause is memory fragmentation maybe the environment variable 'TF_GPU_ALLOCATOR=cuda_malloc_async' will improve the situation. Current allocation summary follows. Current allocation summary follows. 2023-04-23 06:50:50.141804: I tensorflow/tsl/framework/bfc_allocator.cc:1039] BFCAllocator dump for GPU_0_bfc 2023-04-23 06:50:50.141845: I tensorflow/tsl/framework/bfc_allocator.cc:1046] Bin (256): Total Chunks: 84, Chunks in use: 84. 21.0KiB allocated for chunks. 21.0KiB in use in bin. 344B client-requested in use in bin. 2023-04-23 06:50:50.141876: I tensorflow/tsl/framework/bfc_allocator.cc:1046] Bin (512): Total Chunks: 20, Chunks in use: 20. 10.0KiB allocated for chunks. 10.0KiB in use in bin. 7.8KiB client-requested in use in bin. . . . 2023-04-23 06:50:50.145528: I tensorflow/tsl/framework/bfc_allocator.cc:1107] Sum Total of in-use chunks: 6.53GiB 2023-04-23 06:50:50.145548: I tensorflow/tsl/framework/bfc_allocator.cc:1109] Total bytes in pool: 11920474112 memory_limit: 11920474112 available bytes: 0 curr_region_allocation_bytes: 23840948224 2023-04-23 06:50:50.145580: I tensorflow/tsl/framework/bfc_allocator.cc:1114] Stats: Limit: 11920474112 InUse: 7013281536 MaxInUse: 7013281536 NumAllocs: 147 MaxAllocSize: 6913350144 Reserved: 0 PeakReserved: 0 LargestFreeBlock: 0

2023-04-23 06:50:50.145649: W tensorflow/tsl/framework/bfc_allocator.cc:497] ***********************************************************_________________________________________ 2023-04-23 06:50:50.145724: W tensorflow/core/framework/op_kernel.cc:1807] OP_REQUIRES failed at constant_op.cc:81 : RESOURCE_EXHAUSTED: OOM when allocating tensor of shape [512100,15,225] and type float Traceback (most recent call last): File "CPRS.py", line 515, in model.fit_generator(traingen, epochs = 3, steps_per_epoch = len(train_label)//batch_size) File "/home/yyko/.local/lib/python3.8/site-packages/keras/engine/training_v1.py", line 1356, in fit_generator return self.fit( File "/home/yyko/.local/lib/python3.8/site-packages/keras/engine/training_v1.py", line 856, in fit return func.fit( File "/home/yyko/.local/lib/python3.8/site-packages/keras/engine/training_generator_v1.py", line 647, in fit return fit_generator( File "/home/yyko/.local/lib/python3.8/site-packages/keras/engine/training_generator_v1.py", line 282, in model_iteration batch_outs = batch_function(*batch_data) File "/home/yyko/.local/lib/python3.8/site-packages/keras/engine/training_v1.py", line 1181, in train_on_batch outputs = self.train_function(ins) File "/home/yyko/.local/lib/python3.8/site-packages/keras/backend.py", line 4606, in call self._make_callable(feed_arrays, feed_symbols, symbol_vals, session) File "/home/yyko/.local/lib/python3.8/site-packages/keras/backend.py", line 4531, in _make_callable callable_fn = session._make_callable_from_options(callable_opts) File "/home/yyko/.local/lib/python3.8/site-packages/tensorflow/python/client/session.py", line 1514, in _make_callable_from_options return BaseSession._Callable(self, callable_options) File "/home/yyko/.local/lib/python3.8/site-packages/tensorflow/python/client/session.py", line 1472, in init self._handle = tf_session.TF_SessionMakeCallable( tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor of shape [512100,15,225] and type float [[{{node training/Adam/gradients/gradients/zeros}}]]

Lastly, if it's not too much trouble, could I receive information on the minimum system requirements to train this code? What kind of envirenment did the author use to run it? I would appreciate any advice on how to make the training work in my environment. Thank you.

[seongeunryu]

Could you please check if a lot of memory is used to turn the CPRS?

[summmeer]

No, I think the memory usage is totally tolerable.

[amengpa]

Hello, thank you very much for providing a great source code. I tried to configure the file '/home/sansa/dataset/Addressa/articls_category.pkl', but I couldn't create it. I'm sorry, but can I get the code that makes up the file? Your offer would be greatly appreciated.

[amengpa]

Could not find where 'category_id' is. I'd really appreciate your help.

[summmeer]

Maybe this file can help you. @amengpa

articles_embeddings+titles.zip Originally posted by @summmeer in https://github.com/summmeer/session-based-news-recommendation/issues/7#issuecomment-1441383978