-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
182 lines (146 loc) · 6.41 KB
/
train.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
import os
import torch
import random
import time
from math import log, log2
import matplotlib.pyplot as plt
import math
import numpy as np
import pandas as pd
from tqdm import tqdm
import pickle
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torch_geometric
from torch_geometric.nn.conv import MessagePassing
from torch_geometric.utils import degree
from torch_scatter import scatter_mean, scatter_max, scatter_sum, scatter_add
from torch_geometric.utils import to_dense_adj, to_undirected, remove_self_loops
import wandb
import argparse
from datetime import datetime
parser = argparse.ArgumentParser()
parser.add_argument('--epochs', type=int, help='Number of epochs')
parser.add_argument('--dataset', type=str, default='ml-100k', choices = ['ml-100k', 'ml-1m','facebook', 'yahoo'], help='Choice of the dataset')
parser.add_argument('--K_list', nargs='+', type=int, default= [1,3,5,10,20,50,100], help='First value of K')
parser.add_argument('--wandb', type=bool, default = False, help='Choose if you want to use Wandb or not')
parser.add_argument('--run_name', type=str, help = 'Name of the run for logging')
parser.add_argument('--layers', type=int, help = 'Number of layers')
parser.add_argument('--seed', type=int, default=42, help = 'Seed')
parser.add_argument('--gpu_id', type=str, default= '0', help = 'Id of the gpu')
parser.add_argument('--learning_rate', default=0.001, type=float, help = 'Learning rate')
parser.add_argument('--entity_name', default='sheaf_nn_recommenders', type=str, help = 'Entity name for shared projects in Wandb. If there is no shared project, default there is no shared project (0).')
parser.add_argument('--project_name', default='Recommendation', type=str, help = 'Project name for Wandb')
parser.add_argument('--model', default='sheaf', type=str, help = 'Name of the model')
parser.add_argument('--log_metrics', type=bool, default=False, help = 'Log for statistical tests')
parser.add_argument('--latent_dim', type=int, default=64, help = 'Number of latent dimensions')
args = parser.parse_args()
latent_dim = args.latent_dim
n_layers = args.layers
EPOCHS = args.epochs
SEED = args.seed
BATCH_SIZE = 1024
DECAY = 0.0001
LR = args.learning_rate
K_list = args.K_list
DATASET = args.dataset
def store_params(gpu_id, dataset_name, model):
params = {'gpu_id' : gpu_id, 'dataset_name': dataset_name, 'model': model, 'seed': args.seed,
'run_name' : args.run_name}
with open(os.getcwd() + '/params.pickle', 'wb') as handle:
pickle.dump(params, handle)
store_params(args.gpu_id, args.dataset, args.model)
from dataset import *
from models import *
from evaluation import *
if args.wandb:
wandb.init(
entity = args.entity_name if args.entity_name != '0' else None ,
project= args.project_name,
name=args.run_name,
config={
"learning_rate": args.learning_rate,
"dataset": args.dataset,
"epochs": args.epochs,
"seed": SEED,
"layers": args.layers,
})
torch.manual_seed(SEED)
device = torch.device("cuda:" + str(args.gpu_id) if torch.cuda.is_available() else "cpu")
def eval(model, train_df, data_df, split_name = "val"):
model.eval()
with torch.no_grad():
_, out = model(train_edge_index)
final_user_Embed, final_item_Embed = torch.split(out, (n_users, n_items))
all_metrics = get_metrics(
final_user_Embed, final_item_Embed, n_users, n_items, train_df, data_df, K_list,
return_mean_values=True, log_metrics=args.log_metrics, device=device)
#We have to log the metrics for each value of K.
#We have to log if is for train or test by using split_name
if args.wandb:
for k in K_list:
den = all_metrics[f'precision@{k}'] + all_metrics[f'recall@{k}']
if den != 0:
f1 = 2 * all_metrics[f'precision@{k}'] * all_metrics[f'recall@{k}'] / den
else:
f1 = 0
wandb.log({"{} Top Recall@{}".format(split_name, k): all_metrics[f'recall@{k}'],
"{} Top Precision@{}".format(split_name, k): all_metrics[f'precision@{k}'],
"{} Top F1@{}".format(split_name, k): f1,
"{} Top NDGC@{}".format(split_name, k): all_metrics[f'ndcg@{k}'],
"{} Top MRR@{}".format(split_name, k): all_metrics[f'mrr@{k}']})
def train_and_eval(model, optimizer, train_df):
'''
model: input of the training method
optimizer: selected optimizer, to compute the BPR loss and the evaluation metrics
train_df: data taken as input
This is the main method of this project. It trains the network and then computes all the metrics.
'''
loss_list_epoch = []
bpr_loss_list_epoch = []
reg_loss_list_epoch = []
for epoch in tqdm(range(EPOCHS)):
n_batch = int(len(train)/BATCH_SIZE)
final_loss_list = []
bpr_loss_list = []
reg_loss_list = []
model.train()
for batch_idx in tqdm(range(n_batch)):
optimizer.zero_grad()
users, pos_items, neg_items = data_loader(train_df, BATCH_SIZE, n_users, n_items)
users_emb, pos_emb, neg_emb, userEmb0, posEmb0, negEmb0 = model.encode_minibatch(users, pos_items, neg_items, train_edge_index)
bpr_loss, reg_loss = compute_bpr_loss(
users, users_emb, pos_emb, neg_emb, userEmb0, posEmb0, negEmb0
)
final_loss = bpr_loss + reg_loss
final_loss.backward()
optimizer.step()
final_loss_list.append(final_loss.item())
bpr_loss_list.append(bpr_loss.item())
reg_loss_list.append(reg_loss.item())
eval(model, train_df, val_df, "val")
eval(model, train_df, test_df, "test")
if args.wandb:
wandb.log({"Loss":round(np.mean(final_loss_list),4)})
loss_list_epoch.append(round(np.mean(final_loss_list),4))
bpr_loss_list_epoch.append(round(np.mean(bpr_loss_list),4))
reg_loss_list_epoch.append(round(np.mean(reg_loss_list),4))
return (
loss_list_epoch,
bpr_loss_list_epoch,
reg_loss_list_epoch,
)
sheafnn = RecSysGNN(
latent_dim=latent_dim,
num_layers=n_layers,
num_users=n_users,
num_items=n_items,
model='sheaf'
)
sheafnn.to(device)
optimizer = torch.optim.Adam(sheafnn.parameters(), lr=LR)
sheafnn_loss, sheafnn_bpr, sheafnn_reg = train_and_eval(sheafnn, optimizer, train_df)
if args.wandb:
wandb.finish()