HR, NDCG = [], []
set_all = set(range(item_num))
#spend 461s
test_start_time = time.time()
for u_i in testing_user_set:
item_i_list = list(testing_user_set[u_i])
index_end_i = len(item_i_list)
item_j_list = list(set_all - training_user_set[u_i] -
testing_user_set[u_i])
item_i_list.extend(item_j_list)
pre_one = all_pre[u_i][item_i_list]
indices = largest_indices(pre_one, top_k)
indices = list(indices[0])
hr_t, ndcg_t = evaluate.hr_ndcg(indices, index_end_i, top_k)
elapsed_time = time.time() - test_start_time
HR.append(hr_t)
NDCG.append(ndcg_t)
hr_test = round(np.mean(HR), 4)
ndcg_test = round(np.mean(NDCG), 4)
# test_loss,hr_test,ndcg_test = evaluate.metrics(model,testing_loader,top_k,num_negative_test_val,batch_size)
str_print_evl = "epoch:" + str(epoch) + 'time:' + str(
round(elapsed_time, 2)) + "\t test" + " hit:" + str(
hr_test) + ' ndcg:' + str(ndcg_test)
print(str_print_evl)
result_file.write(gcn_user_emb)
result_file.write('\n')
result_file.write(gcn_item_emb)
result_file.write('\n')
def test_amazons(args):
print("start testing")
dataset_base_path = '../../data/AmazonBooks/amazonbooks_x0'
##gowalla
user_num = 52643
item_num = 91599
factor_num = args.embed_size
batch_size = 2048 * 512
top_k = 20
num_negative_test_val = -1 ##all
start_i_test = 130
end_i_test = 200
setp = 5
run_id = args.run_id
print(run_id)
dataset = 'amazon-book'
path_save_base = './log/' + dataset + '/newloss' + run_id
if (os.path.exists(path_save_base)):
print('has results save path')
else:
print('error')
pdb.set_trace()
result_file = open(path_save_base + '/results_hdcg_hr.txt',
'a') #('./log/results_gcmc.txt','w+')
copyfile('../../external/LR_gccf/test_amazons.py',
path_save_base + '/test_amazos' + run_id + '.py')
path_save_model_base = './newlossModel/' + dataset + '/s' + run_id
if (os.path.exists(path_save_model_base)):
print('has model save path')
else:
pdb.set_trace()
training_user_set, training_item_set, training_set_count = np.load(
dataset_base_path + '/datanpy/training_set.npy', allow_pickle=True)
testing_user_set, testing_item_set, testing_set_count = np.load(
dataset_base_path + '/datanpy/testing_set.npy', allow_pickle=True)
user_rating_set_all = np.load(dataset_base_path +
'/datanpy/user_rating_set_all.npy',
allow_pickle=True).item()
u_d = readD(training_user_set, user_num)
i_d = readD(training_item_set, item_num)
d_i_train = u_d
d_j_train = i_d
sparse_u_i = readTrainSparseMatrix(training_user_set, True, u_d, i_d)
sparse_i_u = readTrainSparseMatrix(training_item_set, False, u_d, i_d)
test_batch = 52 #int(batch_size/32)
testing_dataset = data_utils.resData(train_dict=testing_user_set,
batch_size=test_batch,
num_item=item_num,
all_pos=training_user_set)
testing_loader = DataLoader(testing_dataset,
batch_size=1,
shuffle=False,
num_workers=0)
model = BPR(user_num, item_num, factor_num, sparse_u_i, sparse_i_u,
d_i_train, d_j_train)
model = model.to('cuda')
# lr = 0.001
optimizer_bpr = torch.optim.Adam(model.parameters(),
lr=args.lr) #, betas=(0.5, 0.99))
print('--------test processing-------')
count, best_hr = 0, 0
for epoch in range(start_i_test, end_i_test, setp):
model.train()
PATH_model = path_save_model_base + '/epoch' + str(epoch) + '.pt'
#torch.save(model.state_dict(), PATH_model)
model.load_state_dict(torch.load(PATH_model))
model.eval()
# ######test and val###########
gcn_users_embedding, gcn_items_embedding, gcn_user_emb, gcn_item_emb = model(
torch.cuda.LongTensor([0]), torch.cuda.LongTensor([0]),
torch.cuda.LongTensor([0]))
user_e = gcn_users_embedding.cpu().detach().numpy()
item_e = gcn_items_embedding.cpu().detach().numpy()
all_pre = np.matmul(user_e, item_e.T)
HR, NDCG, RC = [], [], []
set_all = set(range(item_num))
#spend 461s
test_start_time = time.time()
for u_i in testing_user_set:
item_i_list = list(testing_user_set[u_i])
index_end_i = len(item_i_list)
item_j_list = list(set_all - training_user_set[u_i] -
testing_user_set[u_i])
item_i_list.extend(item_j_list)
pre_one = all_pre[u_i][item_i_list]
indices = largest_indices(pre_one, top_k)
indices = list(indices[0])
hr_t, ndcg_t, rc_t = evaluate.hr_ndcg(indices, index_end_i, top_k)
elapsed_time = time.time() - test_start_time
HR.append(hr_t)
NDCG.append(ndcg_t)
RC.append(rc_t)
hr_test = round(np.mean(HR), 4)
ndcg_test = round(np.mean(NDCG), 4)
rc_test = round(np.mean(RC), 4)
# test_loss,hr_test,ndcg_test = evaluate.metrics(model,testing_loader,top_k,num_negative_test_val,batch_size)
str_print_evl = "epoch:" + str(epoch) + 'time:' + str(
round(elapsed_time,
2)) + "\t test" + " hit:" + str(hr_test) + ' ndcg:' + str(
ndcg_test) + ' recall:' + str(rc_test)
print(str_print_evl)
result_file.write(gcn_user_emb)
result_file.write('\n')
result_file.write(gcn_item_emb)
result_file.write('\n')
result_file.write(str_print_evl)
result_file.write('\n')
result_file.flush()
