Reuse entity, item and user embeddings

[lfomendes]

I'm experimenting with KGAT using my own user-item and KG data.
But I would like to reuse all the embeddings generated during the recommendations in other models and to visualize/clusterize the items.

How can I do that?
Looking at the KGAT.py I see that you keep

all_weights['user_embed']
all_weights['entity_embed']
all_weights['relation_embed']
all_weights['trans_W']

Does it make sense to get the embeddings from the 'entity embed' or since you are using TransR they only make sense when I trasnform with a specific relation?

Thanks

[lfomendes]

And is it possible to save these vectors?
The only output for this code is the recommendation metrics?

[xiangwang1223]

Yes, you can save these embedding vectors via the codes like:

knowledge_graph_attention_network/Model/Main.py

Lines 170 to 177 in 530327a

	user_embed, entity_embed, relation_embed = sess.run(
	[model.weights['user_embed'], model.weights['entity_embed'], model.weights['relation_embed']],
	feed_dict={})
	temp_save_path = '%spretrain/%s/%s.npz' % (args.proj_path, args.dataset, args.model_type)
	ensureDir(temp_save_path)
	np.savez(temp_save_path, user_embed=user_embed, entity_embed=entity_embed, relation_embed=relation_embed)
	print('save the weights of kgat in path: ', temp_save_path)

Hope it can be helpful.

[lfomendes]

I will try that
Thank you =D

Does the relation and entity make sense alone? Or do I have to combine them in some way?
In TransR a projection matrix is used, correct?

[xiangwang1223]

I think that storing the relation matrices and entity embeddings separately is reasonable, such that you can combine them later in flexible ways.

[lfomendes]

Thank you for your response =D I got it running and saved the embeddings like you showed me. But now I'm trying to run with "real" data but I got some errors I'm using my own KG data I'm running with the following command !python Main.py --model_type kgat --alg_type bi --dataset bh15 --regs

[1e-5,1e-5] --layer_size [64,32,16] --embed_size 64 --lr 0.001 --epoch 100 --verbose 50 --save_flag 1 --pretrain 1 --batch_size 256 --node_dropout [0.1] --mess_dropout [0.1,0.1,0.1] --use_att True --use_kge True

And this is the data [n_users, n_items]=[252346, 462485]

[n_train, n_test]=[722370, 80211] [n_entities, n_relations, n_triples]=[483176, 5, 1593039] [batch_size, batch_size_kg]=[256, 564]

And the error: without pretraining.

Traceback (most recent call last): File "/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1365, in _do_call return fn(*args) File "/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1350, in _run_fn target_list, run_metadata) File "/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1443, in _call_tf_sessionrun run_metadata) tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor with shape[735522,32] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc [[{{node gradients/SparseTensorDenseMatMul_251/SparseTensorDenseMatMul_grad/SparseTensorDenseMatMul}}]] Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

I will try to understand what is happening but if you have any tips I would appreciate it Thanks ps: I'm using tensorflow 2 but i dont think this is the problem

…

On Mon, Jun 22, 2020 at 10:21 PM Xiang Wang ***@***.***> wrote: I think that storing the relation matrices and entity embeddings separately is reasonable, such that you can combine them later in flexible ways. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#29 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB3D5F232X3GPFUSD6GNOCLRX77SXANCNFSM4OC63TRA> .

[FrankChengGD]

Thank you for your response =D I got it running and saved the embeddings like you showed me. But now I'm trying to run with "real" data but I got some errors I'm using my own KG data I'm running with the following command !python Main.py --model_type kgat --alg_type bi --dataset bh15 --regs
[1e-5,1e-5] --layer_size [64,32,16] --embed_size 64 --lr 0.001 --epoch 100 --verbose 50 --save_flag 1 --pretrain 1 --batch_size 256 --node_dropout [0.1] --mess_dropout [0.1,0.1,0.1] --use_att True --use_kge True
And this is the data [n_users, n_items]=[252346, 462485]
[n_train, n_test]=[722370, 80211] [n_entities, n_relations, n_triples]=[483176, 5, 1593039] [batch_size, batch_size_kg]=[256, 564]
And the error: without pretraining.
Traceback (most recent call last): File "/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1365, in _do_call return fn(args) File "/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1350, in _run_fn target_list, run_metadata) File "/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1443, in _call_tf_sessionrun run_metadata) tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor with shape[735522,32] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc [[{{node gradients/SparseTensorDenseMatMul_251/SparseTensorDenseMatMul_grad/SparseTensorDenseMatMul}}]] Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.
I will try to understand what is happening but if you have any tips I would appreciate it Thanks ps: I'm using tensorflow 2 but i dont think this is the problem
…
On Mon, Jun 22, 2020 at 10:21 PM Xiang Wang @.**> wrote: I think that storing the relation matrices and entity embeddings separately is reasonable, such that you can combine them later in flexible ways. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#29 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB3D5F232X3GPFUSD6GNOCLRX77SXANCNFSM4OC63TRA .

@lfomendes
Hi! I am also facing the same error as this 'ResourceExhaustedError'. Is it means the GPU is short of memory?
May I get some help on how did you fix this error?

Thanks a lot!