add

DPM-GSP-fully-supervised/logs_train/cora/Gaussian_GCNConv/Feb17-04:36:56.log

@@ -0,0 +1,12 @@
+Feb17-04:36:56
+----------------------------------------------------------------------------------------------------
+[cora]   seed=42
+----------------------------------------------------------------------------------------------------
+lr=0.003 unweighted_MSE=False diffusion_steps=80 schedule=True weight_decay=0.003 cat=False skip=False
+nhid=16 layers=2 linears=2
+----------------------------------------------------------------------------------------------------
+001 | 1.02s | train acc: 1.645e-01 | val acc: 1.148e-01 | val graph: 3.000e-02 | test node: 9.083e-02 | test graph: 4.700e-02
+101 | 0.41s | train acc: 1.491e-01 | val acc: 1.556e-01 | val graph: 1.100e-01 | test node: 1.526e-01 | test graph: 1.070e-01
+201 | 0.40s | train acc: 1.272e-01 | val acc: 1.511e-01 | val graph: 1.080e-01 | test node: 1.598e-01 | test graph: 1.040e-01
+301 | 0.40s | train acc: 1.722e-01 | val acc: 2.900e-01 | val graph: 2.080e-01 | test node: 2.922e-01 | test graph: 2.080e-01
+401 | 0.40s | train acc: 1.542e-01 | val acc: 2.876e-01 | val graph: 2.180e-01 | test node: 2.895e-01 | test graph: 2.160e-01

DPM-GSP-fully-supervised/method_series/ddpm_trainer.py

@@ -3,7 +3,6 @@
 from tqdm import tqdm, trange
 import numpy as np
 import torch
-import neptune.new as neptune
 from utils.loader import load_seed, load_device, load_data, load_model_params, load_model_optimizer, load_batch, load_loss_fn
 from utils.logger import Logger, set_log, start_log, train_log
 

DPM-GSP-fully-supervised/method_series/ddpm_trainer_huge.py

@@ -3,7 +3,6 @@
 from tqdm import tqdm, trange
 import numpy as np
 import torch
-import neptune.new as neptune
 from utils.loader import load_seed, load_device, load_data, load_model_params, load_model_optimizer, \
                          load_batch, load_loss_fn
 from utils.logger import Logger, set_log, start_log, train_log

DPM-GSP-fully-supervised/method_series/ddpm_trainer_huge_ppi.py

@@ -3,7 +3,6 @@
 from tqdm import tqdm, trange
 import numpy as np
 import torch
-import neptune.new as neptune
 from utils.loader import load_seed, load_device, load_data, load_model_params, load_model_optimizer, \
                          load_batch, load_loss_fn
 from utils.logger import Logger, set_log, start_log, train_log

DPM-GSP-fully-supervised/method_series/gaussian_ddpm_losses.py

@@ -3,7 +3,6 @@
 import torch.nn.functional as F
 import math
 from torch.autograd import Variable
-import neptune.new as neptune
 import time
 from sklearn.metrics import f1_score
 from torch_scatter import scatter_mean

DPM-GSP-reasoning/method_series/gaussian_ddpm_losses.py

@@ -3,7 +3,6 @@
 import torch.nn.functional as F
 import math
 from torch.autograd import Variable
-import neptune.new as neptune
 
 
 def sum_except_batch(x, num_dims=1):

DPM-GSP-semi-supervised/logs_train/cora/Continuous_GATConv/Feb17-03:08:54.log

@@ -0,0 +1,7 @@
+Feb17-03:08:54
+----------------------------------------------------------------------------------------------------
+[cora]   seed=42
+----------------------------------------------------------------------------------------------------
+lr=0.01 diffusion_steps=80 temperature=0.3 schedule=True weight_decay=0.01  skip=True load start=2000
+nhid=8 layers=2 linears=1
+----------------------------------------------------------------------------------------------------

README.md

@@ -17,21 +17,27 @@ We provide the PyTorch implementation for DPM-GSP framework here. The repository
 ```
 
 ## Setting up the environment
-You can set up the environment by following commands.
+You can set up the environment by following commands. 
 
-### DPM-GSP-fully-supervised  
 ```sh
-conda create --name <env> --file requirements_supervised.txt
+conda create -n DPM-GSP python=3.10
+conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
+pip install tqdm
+pip install pyyaml
+pip install easydict
+pip install torch-sparse
+pip install torch-scatter==2.0.9
 ```
+You also need to install torch-geometric package. Each experiment requires a different version.
 
-### DPM-GSP-semi-supervised  
+### DPM-GSP for fully-supervised and reasoning  
 ```sh
-conda create --name <env> --file requirements_semi-supervised.txt
+pip install torch-geometric==1.7.1
 ```
 
-### DPM-GSP-reasoning
+### DPM-GSP-semi-supervised  
 ```sh
-conda create --name <env> --file requirements_supervised.txt
+pip install torch-geometric==2.1.0
 ```
 
 ## Running