Skip to content

S
starrygl-DynamicHistory
Commit 71cb0206 by zlj
Options

add examples

parent 16cff1d2
Showing with 1851 additions and 0 deletions
examples/a.out 0 → 100644
LOCAL RANK 0, RANK0
use cuda on 0
9228
get_neighbors consume: 0.0103759s
Epoch 0:
train loss:377.5712 train ap:0.903848 val ap:0.886584 val auc:0.904656
total time:11.40s prep time:9.88s
fetch time:0.00s write back time:0.00s
Epoch 1:
train loss:329.1190 train ap:0.920000 val ap:0.885216 val auc:0.904735
total time:11.32s prep time:9.79s
fetch time:0.00s write back time:0.00s
Epoch 2:
train loss:316.1359 train ap:0.924376 val ap:0.895123 val auc:0.912622
total time:11.49s prep time:9.95s
fetch time:0.00s write back time:0.00s
Epoch 3:
train loss:311.4889 train ap:0.926138 val ap:0.893922 val auc:0.912589
total time:11.50s prep time:9.97s
fetch time:0.00s write back time:0.00s
Epoch 4:
train loss:302.2057 train ap:0.929684 val ap:0.889695 val auc:0.909766
total time:11.48s prep time:9.95s
fetch time:0.00s write back time:0.00s
Epoch 5:
train loss:300.2464 train ap:0.931034 val ap:0.897774 val auc:0.916421
total time:11.48s prep time:9.95s
fetch time:0.00s write back time:0.00s
Epoch 6:
train loss:293.5465 train ap:0.934657 val ap:0.896159 val auc:0.914983
total time:11.55s prep time:10.02s
fetch time:0.00s write back time:0.00s
Epoch 7:
train loss:285.9396 train ap:0.937834 val ap:0.905351 val auc:0.922268
total time:11.52s prep time:9.99s
fetch time:0.00s write back time:0.00s
Epoch 8:
train loss:281.7048 train ap:0.941035 val ap:0.909690 val auc:0.924262
total time:11.51s prep time:9.98s
fetch time:0.00s write back time:0.00s
Epoch 9:
train loss:273.8330 train ap:0.945250 val ap:0.913860 val auc:0.928068
total time:11.56s prep time:10.00s
fetch time:0.00s write back time:0.00s
Epoch 10:
train loss:268.6164 train ap:0.947141 val ap:0.917379 val auc:0.930309
total time:11.77s prep time:10.19s
fetch time:0.00s write back time:0.00s
Epoch 11:
train loss:265.0121 train ap:0.949457 val ap:0.918648 val auc:0.931452
total time:11.62s prep time:10.08s
fetch time:0.00s write back time:0.00s
Epoch 12:
train loss:255.6320 train ap:0.953506 val ap:0.919272 val auc:0.932783
total time:11.50s prep time:9.98s
fetch time:0.00s write back time:0.00s
Epoch 13:
train loss:252.6296 train ap:0.954798 val ap:0.924649 val auc:0.936515
total time:11.50s prep time:9.96s
fetch time:0.00s write back time:0.00s
Epoch 14:
train loss:248.4476 train ap:0.956243 val ap:0.925952 val auc:0.938199
total time:11.53s prep time:10.00s
fetch time:0.00s write back time:0.00s
Epoch 15:
train loss:243.4459 train ap:0.958749 val ap:0.929440 val auc:0.940865
total time:11.54s prep time:10.01s
fetch time:0.00s write back time:0.00s
Epoch 16:
train loss:238.6286 train ap:0.960667 val ap:0.936339 val auc:0.946161
total time:17.48s prep time:15.12s
fetch time:0.00s write back time:0.00s
Epoch 17:
train loss:234.5283 train ap:0.961787 val ap:0.933828 val auc:0.944680
total time:18.09s prep time:15.69s
fetch time:0.00s write back time:0.00s
Epoch 18:
train loss:227.3527 train ap:0.964591 val ap:0.932110 val auc:0.943765
total time:18.17s prep time:15.46s
fetch time:0.00s write back time:0.00s
Epoch 19:
train loss:223.7772 train ap:0.965486 val ap:0.937780 val auc:0.947312
total time:17.80s prep time:15.43s
fetch time:0.00s write back time:0.00s
Epoch 20:
train loss:221.9428 train ap:0.966139 val ap:0.938104 val auc:0.948022
total time:18.31s prep time:15.82s
fetch time:0.00s write back time:0.00s
Epoch 21:
train loss:216.8870 train ap:0.968285 val ap:0.942088 val auc:0.950660
total time:18.14s prep time:15.48s
fetch time:0.00s write back time:0.00s
Epoch 22:
train loss:213.5077 train ap:0.968911 val ap:0.944023 val auc:0.951869
total time:18.09s prep time:15.56s
fetch time:0.00s write back time:0.00s
Epoch 23:
train loss:210.1412 train ap:0.970743 val ap:0.944840 val auc:0.952554
total time:17.74s prep time:15.47s
fetch time:0.00s write back time:0.00s
Epoch 24:
train loss:208.9109 train ap:0.971101 val ap:0.944029 val auc:0.952720
total time:18.47s prep time:15.73s
fetch time:0.00s write back time:0.00s
Epoch 25:
train loss:207.5198 train ap:0.970606 val ap:0.944518 val auc:0.952912
total time:17.97s prep time:15.66s
fetch time:0.00s write back time:0.00s
Epoch 26:
train loss:203.6585 train ap:0.971611 val ap:0.940218 val auc:0.949371
total time:17.70s prep time:15.42s
fetch time:0.00s write back time:0.00s
Epoch 27:
train loss:203.3531 train ap:0.972317 val ap:0.949000 val auc:0.956595
total time:18.01s prep time:15.33s
fetch time:0.00s write back time:0.00s
Epoch 28:
train loss:198.1525 train ap:0.973525 val ap:0.948420 val auc:0.955604
total time:17.78s prep time:15.31s
fetch time:0.00s write back time:0.00s
Epoch 29:
train loss:197.6365 train ap:0.973818 val ap:0.944911 val auc:0.953313
total time:17.74s prep time:15.49s
fetch time:0.00s write back time:0.00s
Epoch 30:
train loss:197.7800 train ap:0.973573 val ap:0.950356 val auc:0.958595
total time:18.24s prep time:15.60s
fetch time:0.00s write back time:0.00s
Epoch 31:
train loss:194.4391 train ap:0.974730 val ap:0.952775 val auc:0.959729
total time:17.84s prep time:15.23s
fetch time:0.00s write back time:0.00s
Epoch 32:
train loss:190.1150 train ap:0.976038 val ap:0.953111 val auc:0.959360
total time:17.72s prep time:15.46s
fetch time:0.00s write back time:0.00s
Epoch 33:
train loss:185.7417 train ap:0.976925 val ap:0.954769 val auc:0.961057
total time:18.04s prep time:15.56s
fetch time:0.00s write back time:0.00s
Epoch 34:
train loss:189.0004 train ap:0.976267 val ap:0.954641 val auc:0.961198
total time:17.89s prep time:15.12s
fetch time:0.00s write back time:0.00s
Epoch 35:
train loss:185.4487 train ap:0.977420 val ap:0.954675 val auc:0.960969
total time:17.65s prep time:15.13s
fetch time:0.00s write back time:0.00s
Epoch 36:
train loss:185.9187 train ap:0.977260 val ap:0.955284 val auc:0.961039
total time:17.67s prep time:15.36s
fetch time:0.00s write back time:0.00s
Epoch 37:
train loss:184.6686 train ap:0.977626 val ap:0.955124 val auc:0.961923
total time:17.90s prep time:15.42s
fetch time:0.00s write back time:0.00s
Epoch 38:
train loss:183.1190 train ap:0.977930 val ap:0.956069 val auc:0.962114
total time:18.10s prep time:15.26s
fetch time:0.00s write back time:0.00s
Epoch 39:
train loss:179.3445 train ap:0.978350 val ap:0.958382 val auc:0.963833
total time:18.05s prep time:15.60s
fetch time:0.00s write back time:0.00s
Epoch 40:
train loss:174.6380 train ap:0.980014 val ap:0.956793 val auc:0.963013
total time:18.28s prep time:15.77s
fetch time:0.00s write back time:0.00s
Epoch 41:
train loss:178.2737 train ap:0.979067 val ap:0.958580 val auc:0.964227
total time:18.24s prep time:15.51s
fetch time:0.00s write back time:0.00s
Epoch 42:
train loss:175.7294 train ap:0.979611 val ap:0.960288 val auc:0.965754
total time:17.98s prep time:15.62s
fetch time:0.00s write back time:0.00s
Epoch 43:
train loss:173.2326 train ap:0.980324 val ap:0.960428 val auc:0.965867
total time:18.21s prep time:15.60s
fetch time:0.00s write back time:0.00s
Epoch 44:
train loss:172.3492 train ap:0.980196 val ap:0.962143 val auc:0.966774
total time:18.35s prep time:15.80s
fetch time:0.00s write back time:0.00s
Epoch 45:
train loss:168.8601 train ap:0.981180 val ap:0.963014 val auc:0.968132
total time:17.73s prep time:15.50s
fetch time:0.00s write back time:0.00s
Epoch 46:
train loss:169.5997 train ap:0.981473 val ap:0.961124 val auc:0.966405
total time:13.20s prep time:11.67s
fetch time:0.00s write back time:0.00s
Epoch 47:
train loss:167.5232 train ap:0.981394 val ap:0.961333 val auc:0.966534
total time:11.49s prep time:9.96s
fetch time:0.00s write back time:0.00s
Epoch 48:
train loss:165.6863 train ap:0.981684 val ap:0.960024 val auc:0.965201
total time:11.50s prep time:9.97s
fetch time:0.00s write back time:0.00s
Epoch 49:
train loss:165.3790 train ap:0.981795 val ap:0.962299 val auc:0.967019
total time:11.54s prep time:9.98s
fetch time:0.00s write back time:0.00s
Loading the best model at epoch 45
test AP:0.946485 test AUC:0.954197
test_dataset 23621 avg_time 13.31522078514099
examples/evaluate_link_predict.py 0 → 100644
import argparse
import os
import sys
from os.path import abspath, join, dirname
from starrygl.distributed.context import DistributedContext
from starrygl.distributed.utils import DistIndex
from starrygl.evaluation.get_evalute_data import get_link_prediction_data
from starrygl.module.modules import GeneralModel
from pathlib import Path
from pathlib import Path
from starrygl.module.utils import parse_config
from starrygl.sample.cache.fetch_cache import FetchFeatureCache
from starrygl.sample.graph_core import DataSet, DistributedGraphStore, TemporalNeighborSampleGraph
from starrygl.module.utils import parse_config, EarlyStopMonitor
from starrygl.sample.graph_core import DataSet, DistributedGraphStore, TemporalNeighborSampleGraph
from starrygl.sample.memory.shared_mailbox import SharedMailBox
from starrygl.sample.sample_core.EvaluateNegativeSampling import EvaluateNegativeSampling
from starrygl.sample.sample_core.base import NegativeSampling
from starrygl.sample.sample_core.neighbor_sampler import NeighborSampler
from starrygl.sample.part_utils.partition_tgnn import partition_load
import torch
import time
import torch
import torch.nn.functional as F
import torch.distributed as dist
import torch.multiprocessing as mp
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.distributed import init_process_group, destroy_process_group
import os
from starrygl.sample.data_loader import DistributedDataLoader
from starrygl.sample.batch_data import SAMPLE_TYPE
from starrygl.sample.stream_manager import getPipelineManger
parser = argparse.ArgumentParser(
description="RPC Reinforcement Learning Example",
formatter_class=argparse.ArgumentDefaultsHelpFormatter,
)
parser.add_argument('--rank', default=0, type=int, metavar='W',
help='name of dataset')
parser.add_argument('--patience', type=int, default=5, help='Patience for early stopping')
parser.add_argument('--world_size', default=1, type=int, metavar='W',
help='number of negative samples')
parser.add_argument('--dataname', default=1, type=str, metavar='W',
help='name of dataset')
parser.add_argument('--model', default='TGN', type=str, metavar='W',
help='name of model')
parser.add_argument('--negative_sample_strategy', default='random', type=str, metavar='W',
help='name of negative sample strategy')
parser.add_argument('--negative_sample_strategy', default='random', type=str, metavar='W',
help='name of negative sample strategy')
args = parser.parse_args()
from sklearn.metrics import average_precision_score, roc_auc_score
import torch
import time
import random
import dgl
import numpy as np
from sklearn.metrics import average_precision_score, roc_auc_score
from torch.nn.parallel import DistributedDataParallel as DDP
#os.environ['CUDA_VISIBLE_DEVICES'] = str(args.rank)
#os.environ["RANK"] = str(args.rank)
#os.environ["WORLD_SIZE"] = str(args.world_size)
#os.environ["LOCAL_RANK"] = str(0)
torch.cuda.set_device(int(os.environ["LOCAL_RANK"]))
os.environ["MASTER_ADDR"] = '10.214.211.187'
os.environ["MASTER_PORT"] = '9337'
def seed_everything(seed=42):
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
seed_everything(1234)
def main():
print('main')
use_cuda = True
sample_param, memory_param, gnn_param, train_param = parse_config('./config/{}.yml'.format(args.model))
torch.set_num_threads(12)
ctx = DistributedContext.init(backend="nccl", use_gpu=True)
device_id = torch.cuda.current_device()
pdata = partition_load("/mnt/data/part_data/evaluate/{}".format(args.dataname), algo="metis_for_tgnn")
graph = DistributedGraphStore(pdata = pdata,uvm_edge = False)
gnn_param['dyrep'] = True if args.model == 'DyRep' else False
use_src_emb = gnn_param['use_src_emb'] if 'use_src_emb' in gnn_param else False
use_dst_emb = gnn_param['use_dst_emb'] if 'use_dst_emb' in gnn_param else False
gnn_dim_node = 0 if graph.x is None else pdata.x.shape[1]
gnn_dim_edge = 0 if graph.edge_attr is None else pdata.edge_attr.shape[1]
print(gnn_dim_node,gnn_dim_edge)
avg_time = 0
MODEL_SAVE_PATH = f'./saved_models/{args.model}-{args.dataname}.pth'
if use_cuda:
model = GeneralModel(gnn_dim_node, gnn_dim_edge, sample_param, memory_param, gnn_param, train_param).cuda()
device = torch.device('cuda')
else:
model = GeneralModel(gnn_dim_node, gnn_dim_edge, sample_param, memory_param, gnn_param, train_param)
device = torch.device('cpu')
model.load_state_dict(torch.load(MODEL_SAVE_PATH))
sample_graph = TemporalNeighborSampleGraph(sample_graph = pdata.sample_graph,mode = 'full')
if memory_param['type'] != 'none':
mailbox = SharedMailBox(pdata.ids.shape[0], memory_param, dim_edge_feat = pdata.edge_attr.shape[1] if pdata.edge_attr is not None else 0)
else:
mailbox = None
fanout = []
num_layers = sample_param['layer'] if 'layer' in sample_param else 1
fanout = sample_param['neighbor'] if 'neighbor' in sample_param else [10]
policy = sample_param['strategy'] if 'strategy' in sample_param else 'recent'
sampler = NeighborSampler(num_nodes=graph.num_nodes, num_layers=num_layers, fanout=fanout,graph_data=sample_graph, workers=10,policy = policy, graph_name = "wiki_train")
train_data = torch.masked_select(graph.edge_index,pdata.train_mask.to(device)).reshape(2,-1)
train_ts = torch.masked_select(graph.edge_ts,pdata.train_mask.to(device))
val_data = torch.masked_select(graph.edge_index,pdata.val_mask.to(device)).reshape(2,-1)
val_ts = torch.masked_select(graph.edge_ts,pdata.val_mask.to(device))
test_data = torch.masked_select(graph.edge_index,pdata.test_mask.to(device)).reshape(2,-1)
test_ts = torch.masked_select(graph.edge_ts,pdata.test_mask.to(device))
#print(train_data.shape[1],val_data.shape[1],test_data.shape[1])
train_data = DataSet(edges = train_data,ts =train_ts,eids = torch.nonzero(pdata.train_mask).view(-1))
test_data = DataSet(edges = test_data,ts =test_ts,eids = torch.nonzero(pdata.test_mask).view(-1))
val_data = DataSet(edges = val_data,ts = val_ts,eids = torch.nonzero(pdata.val_mask).view(-1))
new_node_val_data = torch.masked_select(graph.edge_index,pdata.new_node_val_mask.to(device)).reshape(2,-1)
new_node_val_ts = torch.masked_select(graph.edge_ts,pdata.new_node_val_mask.to(device))
new_node_test_data = torch.masked_select(graph.edge_index,pdata.new_node_test_mask.to(device)).reshape(2,-1)
new_node_test_ts = torch.masked_select(graph.edge_ts,pdata.new_node_test_mask.to(device))
new_node_val_data = DataSet(edges = new_node_val_data, ts = new_node_val_ts, edis = torch.nonzero(pdata.new_node_val_mask).view(-1))
new_node_test_data = DataSet(edges = new_node_test_data, ts = new_node_test_ts, edis = torch.nonzero(pdata.new_node_test_mask).view(-1))
if args.negative_sample_strategy != 'random':
val_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=graph.edge_index[0,:], dst_node_ids=graph.edge_index[1,:],
interact_times=graph.edge_ts, last_observed_time=train_data.ts[-1],
negative_sample_strategy=args.negative_sample_strategy, seed=0)
new_node_val_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=new_node_val_data.edges[0,:], dst_node_ids=new_node_val_data.edges[1,:],
interact_times=new_node_val_data.ts, last_observed_time=train_data.ts[-1],
negative_sample_strategy=args.negative_sample_strategy, seed=1)
test_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=graph.edge_index[0,:], dst_node_ids=graph.edge_index[1,:],
interact_times=graph.edge_ts, last_observed_time=val_data.ts[-1],
negative_sample_strategy=args.negative_sample_strategy, seed=2)
new_node_test_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=new_node_test_data.edges[0,:], dst_node_ids=new_node_test_data.edges[1,:],
interact_times=new_node_test_data.ts, last_observed_time=val_data.ts[-1],
negative_sample_strategy=args.negative_sample_strategy, seed=3)
else:
val_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=graph.edge_index[0,:], dst_node_ids=graph.edge_index[1,:], seed=0)
new_node_val_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=new_node_val_data.edges[0,:], dst_node_ids=new_node_val_data.edges[1,:], seed=1)
test_neg_edge_sampler =EvaluateNegativeSampling(src_node_ids=graph.edge_index[0,:], dst_node_ids=graph.edge_index[1,:], seed=2)
new_node_test_neg_edge_sampler = EvaluateNegativeSampling(src_node_ids=new_node_test_data.edges[0,:], dst_node_ids=new_node_test_data.edges[1,:], seed=3)
examples/preprocess/pre_train.py 0 → 100644
import itertools
import torch
import torch.nn as nn
import torch.nn.functional as F
import dgl.function as fn
from dgl.nn import SAGEConv
import numpy as np
class TimeEncode(torch.nn.Module):
def __init__(self, dim):
super(TimeEncode, self).__init__()
self.dim = dim
self.w = torch.nn.Linear(1, dim)
self.w.weight = torch.nn.Parameter((torch.from_numpy(1 / 10 ** np.linspace(0, 9, dim, dtype=np.float32))).reshape(dim, -1))
self.w.bias = torch.nn.Parameter(torch.zeros(dim))
def forward(self, t):
output = torch.cos(self.w(t.float().reshape((-1, 1))))
return output
class GraphSAGE(nn.Module):
def __init__(self, in_feats, h_feats):
super(GraphSAGE, self).__init__()
self.conv1 = SAGEConv(in_feats, h_feats, "mean")
self.conv2 = SAGEConv(h_feats, h_feats, "mean")
def forward(self, g, in_feat):
h = self.conv1(g, in_feat)
h = F.relu(h)
h = self.conv2(g, h)
return h
class TSAGELayer(nn.Module):
def __init__(self, in_dim = 0, edge_dim = 0, time_dim = 0, h_feats = 0):
super(TSAGELayer, self).__init__()
assert in_dim + time_dim != 0 and h_feats != 0
self.time_dim = time_dim
self.time_enc = TimeEncode(time_dim)
self.sage = SAGEConv(in_dim,h_feats,"mean"),
def forward(self,b):
time_f = self.time_enc(b.edata['dt'])
time_f = torch.cat((torch.zeros(b.num_dst_nodes(),self.time_dim,dtype = time_f.dtype,
device = time_f.device),time_f),dim = 1)
if 'f' in b.edata:
edge_f = torch.cat((torch.zeros(b.num_dst_nodes(),b.edata['f'].shape[1],
dtype = b.edata['f'].dtype,device = b.edata['f'].device),
b.edata['f']),dim = 1)
if 'h' in b.srcdata:
b.srcdata['h'] = torch.cat((b.srcnode['h'],edge_f,time_f),dim = 1)
else:
b.srcdata['h'] = torch.cat((edge_f,time_f),dim = 1)
else:
if 'h' in b.srcdata:
b.srcdata['h'] = torch.cat((b.srcnode['h'],time_f),dim = 1)
else:
b.srcdata['h'] = time_f
return F.relu(self.sage(b,b.src_data['h']))
class TSAGEModel(nn.Module):
def __init__(self, num_layer, node_dim, edge_dim, time_dim, h_dim):
super(TSAGEModel, self).__init__()
self.num_layer = num_layer
layer = []
for i in range(num_layer):
if i != 0:
layer.append(TSAGELayer(h_dim,edge_dim,time_dim,h_dim))
else:
layer.append(TSAGELayer(node_dim,edge_dim,time_dim,h_dim))
self.layers = layer
def forward(self,mfgs):
for l in range(len(mfgs)):
for h in range(len(mfgs[l])):
if l < self.num_layer - 1:
mfgs[l+1][h].srcdata['h'] = self.layers[l](self,mfgs[l][h])
else:
return self.layers[l](self,mfgs[l][h])
class DotPredictor(nn.Module):
def forward(self, g, h):
with g.local_scope():
g.ndata["h"] = h
# Compute a new edge feature named 'score' by a dot-product between the
# source node feature 'h' and destination node feature 'h'.
g.apply_edges(fn.u_dot_v("h", "h", "score"))
# u_dot_v returns a 1-element vector for each edge so you need to squeeze it.
return g.edata["score"][:, 0]
# Thumbnail credits: Link Prediction with Neo4j, Mark Needham
# sphinx_gallery_thumbnail_path = '_static/blitz_4_link_predict.png'
\ No newline at end of file
examples/test.sh 0 → 100644
#torchrun --standalone --nproc-per-node 1 train_tgnn.py --dataname tgbl-wiki --model TGN > tgbl_wiki_train.out &
#wait
#torchrun --standalone --nproc-per-node 4 train_tgnn.py --dataname tgbl-wiki --model TGN > tgbl_wiki_train_4.out &
#wait
#torchrun --standalone --nproc-per-node 1 train_tgnn.py --dataname tgbl-review --model TGN > tgbl_review_train.out &
#wait
#torchrun --standalone --nproc-per-node 4 train_tgnn.py --dataname tgbl-review --model TGN > tgbl_review_train_4.out &
#wait
torchrun --standalone --nproc-per-node 1 train_tgnn.py --dataname tgbl-coin --model TGN_600 > tgbl_coin_train.out &
wait
torchrun --standalone --nproc-per-node 4 train_tgnn.py --dataname tgbl-coin --model TGN_600 > tgbl_coin_train_4.out &
wait
torchrun --standalone --nproc-per-node 1 train_tgnn.py --dataname tgbl-comment --model TGN_600 > tgbl_comment_train.out &
wait
torchrun --standalone --nproc-per-node 4 train_tgnn.py --dataname tgbl-comment --model TGN_600 > tgbl_comment_train_4.out &
wait
torchrun --standalone --nproc-per-node 1 train_tgnn.py --dataname tgbl-flight --model TGN_600 > tgbl_flight_train.out &
wait
torchrun --standalone --nproc-per-node 4 train_tgnn.py --dataname tgbl-flight --model TGN_600 > tgbl_flight_train_4.out &
wait
#torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-wiki --model TGN --train_world_size 1 > tgbl_wiki_4.out &
#wait
#torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-wiki --model TGN --train_world_size 4 > tgbl_wiki.out &
#wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-review --model TGN --train_world_size 1 > tgbl_review.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-review --model TGN --train_world_size 4 > tgbl_review_4.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-coin --model TGN_600 --train_world_size 1 > tgbl_coin.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-coin --model TGN_600 --train_world_size 4 > tgbl_coin_4.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-comment --model TGN_600 --train_world_size 1 > tgbl_comment.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-comment --model TGN_600 --train_world_size 4 > tgbl_comment_4.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-flight --model TGN_600 --train_world_size 1 > tgbl_flight.out &
wait
torchrun --standalone --nproc-per-node 1 evaluate_tgbl_predict.py --dataname tgbl-flight --model TGN_600 --train_world_size 4 > tgbl_flight_4.out &
wait
\ No newline at end of file
examples/tgbl_coin.out 0 → 100644
LOCAL RANK 0, RANK0
initlize distributed
use cuda on 0
638486
get_neighbors consume: 4.52747s
raw file found, skipping download
Dataset directory is /home/zlj/.miniconda3/envs/dgnn-3.10/lib/python3.10/site-packages/tgb/datasets/tgbl_coin
loading processed file
tensor([1648811421, 1648811421, 1648811424, ..., 1667278439, 1667278439,
1667278439], device='cuda:0') tensor([1648811421, 1648811421, 1648811424, ..., 1667278439, 1667278439,
1667278439])
tensor([1662096249, 1662096249, 1662096249, 1662096249, 1662096249, 1662096249,
1662096249, 1662096249, 1662096249, 1662096249, 1662096254, 1662096254,
1662096254, 1662096254, 1662096254, 1662096254, 1662096254, 1662096254,
1662096254, 1662096254, 1662096254, 1662096254, 1662096254, 1662096254,
1662096254, 1662096254, 1662096254, 1662096254, 1662096254, 1662096254,
1662096254, 1662096254, 1662096254, 1662096254, 1662096254, 1662096254,
1662096276, 1662096276, 1662096276, 1662096276, 1662096276, 1662096286,
1662096290, 1662096290, 1662096290, 1662096290, 1662096290, 1662096290,
1662096290, 1662096290, 1662096293, 1662096293, 1662096293, 1662096293,
1662096293, 1662096293, 1662096297, 1662096297, 1662096297, 1662096297,
1662096297, 1662096297, 1662096297, 1662096297, 1662096297, 1662096297,
1662096297, 1662096297, 1662096297, 1662096297, 1662096297, 1662096297,
1662096297, 1662096297, 1662096297, 1662096297, 1662096297, 1662096297,
1662096297, 1662096297, 1662096297, 1662096297, 1662096297, 1662096297,
1662096297, 1662096297, 1662096325, 1662096325, 1662096325, 1662096325,
1662096325, 1662096325, 1662096325, 1662096325, 1662096325, 1662096325,
1662096325, 1662096325, 1662096325, 1662096325], device='cuda:0') tensor([1664482319, 1664482319, 1664482319, ..., 1667278439, 1667278439,
1667278439], device='cuda:0') tensor([1648811421, 1648811421, 1648811424, ..., 1662096217, 1662096217,
1662096217], device='cuda:0')
examples/tgbl_coin_4.out 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
examples/tgbl_coin_train.out 0 → 100644
LOCAL RANK 0, RANK0
use cuda on 0
638486
get_neighbors consume: 4.12395s
Epoch 0:
examples/tgbl_comment.out 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
examples/tgbl_comment_4.out 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
examples/tgbl_comment_train.out 0 → 100644
LOCAL RANK 0, RANK0
use cuda on 0
994790
get_neighbors consume: 6.18508s
Epoch 0:
train loss:12236.7578 train ap:0.986976 val ap:0.934674 val auc:0.946284
total time:630.37s prep time:545.79s
fetch time:0.00s write back time:0.00s
Epoch 1:
train loss:11833.1818 train ap:0.987815 val ap:0.960581 val auc:0.965728
total time:628.44s prep time:542.56s
fetch time:0.00s write back time:0.00s
Epoch 2:
train loss:11622.9559 train ap:0.988244 val ap:0.956752 val auc:0.963083
total time:622.89s prep time:538.77s
fetch time:0.00s write back time:0.00s
Epoch 3:
train loss:11679.1400 train ap:0.988072 val ap:0.929351 val auc:0.943797
total time:681.88s prep time:569.50s
fetch time:0.00s write back time:0.00s
Epoch 4:
train loss:11676.1710 train ap:0.988098 val ap:0.936353 val auc:0.948531
total time:849.98s prep time:741.47s
fetch time:0.00s write back time:0.00s
Epoch 5:
train loss:11745.6001 train ap:0.987897 val ap:0.950828 val auc:0.958958
total time:862.77s prep time:750.90s
fetch time:0.00s write back time:0.00s
Epoch 6:
Early stopping at epoch 6
Loading the best model at epoch 1
0.9248434901237488 0.929413378238678
0.8653780221939087 0.861071765422821
test AP:0.847958 test AUC:0.837159
test_dataset 6647176 avg_time 87.00003329753876
examples/tgbl_comment_train_4.out 0 → 100644
LOCAL RANK 0, RANK0
LOCAL RANK 2, RANK2LOCAL RANK 1, RANK1
LOCAL RANK 3, RANK3
use cuda on 0
use cuda on 1
use cuda on 2
use cuda on 3
994790
994790
994790
994790
get_neighbors consume: 6.11692s
get_neighbors consume: 6.12671s
get_neighbors consume: 6.03983s
get_neighbors consume: 6.05302s
num_batchs: tensor([17384], device='cuda:0')
num_batchs: tensor([3931], device='cuda:2')
num_batchs: tensor([16139], device='cuda:1')
num_batchs: tensor([14244], device='cuda:3')
num_batchs: num_batchs: tensor([1915], device='cuda:3')
num_batchs: tensor([1395], device='cuda:1')
num_batchs: tensor([6920], device='cuda:2')
tensor([850], device='cuda:0')
num_batchs:num_batchs: num_batchs: num_batchs: tensor([5545], device='cuda:2')
tensor([1015], device='cuda:0')tensor([2785], device='cuda:3')
tensor([1736], device='cuda:1')
Epoch 0:
Epoch 0:
Epoch 0:
Epoch 0:
train loss:1331.9399 train ap:0.977517 val ap:0.961959 val auc:0.965566
train loss:1162.0666 train ap:0.981900 val ap:0.961959 val auc:0.965566
train loss:1244.0312 train ap:0.978548 val ap:0.961959 val auc:0.965566
train loss:1308.8701 train ap:0.979221 val ap:0.961959 val auc:0.965566
total time:125.18s prep time:60.37s
total time:125.18s prep time:60.37s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
total time:125.18s prep time:60.37s
total time:125.18s prep time:60.37s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
Epoch 1:
Epoch 1:Epoch 1:
Epoch 1:
train loss:1169.3326 train ap:0.981283 val ap:0.965914 val auc:0.968475
train loss:1227.6728 train ap:0.981686 val ap:0.965914 val auc:0.968475
train loss:1226.7282 train ap:0.980509 val ap:0.965914 val auc:0.968475
train loss:1078.3342 train ap:0.984551 val ap:0.965914 val auc:0.968475
total time:125.97s prep time:61.45s
total time:125.97s prep time:61.45s
total time:125.97s prep time:61.45s
total time:125.97s prep time:61.45s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
Epoch 2:
Epoch 2:
Epoch 2:Epoch 2:
train loss:1157.2600 train ap:0.981707 val ap:0.967135 val auc:0.969484
train loss:1227.8567 train ap:0.981577 val ap:0.967135 val auc:0.969484
train loss:1224.2131 train ap:0.980388 val ap:0.967135 val auc:0.969484
total time:125.33s prep time:60.54s
total time:125.33s prep time:60.54s
total time:125.33s prep time:60.54s
fetch time:0.00s write back time:0.00s
train loss:1071.5106 train ap:0.984690 val ap:0.967135 val auc:0.969484
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
total time:125.33s prep time:60.54s
fetch time:0.00s write back time:0.00s
Epoch 3:Epoch 3:
Epoch 3:
Epoch 3:
train loss:1154.4245 train ap:0.981666 val ap:0.939874 val auc:0.947249
train loss:1221.6654 train ap:0.981759 val ap:0.939874 val auc:0.947249
train loss:1217.4941 train ap:0.980394 val ap:0.939874 val auc:0.947249
train loss:1064.5069 train ap:0.984769 val ap:0.939874 val auc:0.947249
total time:124.82s prep time:60.42s
total time:124.82s prep time:60.42s
total time:124.82s prep time:60.42s
fetch time:0.00s write back time:0.00s
total time:124.82s prep time:60.42s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
Epoch 4:
Epoch 4:
Epoch 4:
Epoch 4:
train loss:1206.6170 train ap:0.980821 val ap:0.958336 val auc:0.962538
train loss:1058.5493 train ap:0.984994 val ap:0.958336 val auc:0.962538
train loss:1153.8455 train ap:0.981657 val ap:0.958336 val auc:0.962538
train loss:1214.2795 train ap:0.981938 val ap:0.958336 val auc:0.962538
total time:124.91s prep time:60.22s
total time:124.91s prep time:60.22s
total time:124.91s prep time:60.22s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
total time:124.91s prep time:60.22s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
Epoch 5:
Epoch 5:
Epoch 5:
Epoch 5:
train loss:1211.5216 train ap:0.982009 val ap:0.949435 val auc:0.953953
train loss:1140.1948 train ap:0.982149 val ap:0.949435 val auc:0.953953
train loss:1050.7336 train ap:0.985195 val ap:0.949435 val auc:0.953953
total time:124.92s prep time:60.40s
train loss:1205.8990 train ap:0.980842 val ap:0.949435 val auc:0.953953
total time:124.92s prep time:60.40s
fetch time:0.00s write back time:0.00s
total time:124.92s prep time:60.40s
total time:124.92s prep time:60.40s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
Epoch 6:Epoch 6:
Epoch 6:
Epoch 6:
train loss:1180.1657 train ap:0.981860 val ap:0.960236 val auc:0.963957
train loss:1120.2221 train ap:0.982891 val ap:0.960236 val auc:0.963957
train loss:1180.3385 train ap:0.983051 val ap:0.960236 val auc:0.963957
total time:124.69s prep time:60.01s
total time:124.69s prep time:60.01s
fetch time:0.00s write back time:0.00s
total time:124.69s prep time:60.01s
fetch time:0.00s write back time:0.00s
train loss:1026.1377 train ap:0.985808 val ap:0.960236 val auc:0.963957
fetch time:0.00s write back time:0.00s
total time:124.69s prep time:60.01s
fetch time:0.00s write back time:0.00s
Epoch 7:
Epoch 7:
Epoch 7:
Epoch 7:
Early stopping at epoch 7
Early stopping at epoch 7
Early stopping at epoch 7
Early stopping at epoch 7
Loading the best model at epoch 2
Loading the best model at epoch 2
Loading the best model at epoch 2
Loading the best model at epoch 2
0.9759191870689392 0.977138340473175
0.9759191870689392 0.977138340473175
0.9759191870689392 0.977138340473175
0.9759191870689392 0.977138340473175
0.9553558826446533 0.9581618309020996
0.9553558826446533 0.9581618309020996
0.9553558826446533 0.9581618309020996
0.9553558826446533 0.9581618309020996
test AP:0.940169 test AUC:0.942460
test AP:0.940169 test AUC:0.942460
test AP:0.940169 test AUC:0.942460
test AP:0.940169 test AUC:0.942460
test_dataset 836763 avg_time 9.689588661193847
test_dataset 509738 avg_time 9.689606781005859
test_dataset 1148929 avg_time 9.689572229385377
test_dataset 4151746 avg_time 9.689626097679138
examples/tgbl_flight.out 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
examples/tgbl_flight_4.out 0 → 100644
LOCAL RANK 0, RANK0
initlize distributed
use cuda on 0
18143
get_neighbors consume: 5.92616s
raw file found, skipping download
Dataset directory is /home/zlj/.miniconda3/envs/dgnn-3.10/lib/python3.10/site-packages/tgb/datasets/tgbl_flight
loading processed file
tensor([1546318800, 1546318800, 1546318800, ..., 1667188800, 1667188800,
1667188800], device='cuda:0') tensor([1546318800, 1546318800, 1546318800, ..., 1667188800, 1667188800,
1667188800])
tensor([1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000, 1638162000, 1638162000,
1638162000, 1638162000, 1638162000, 1638162000], device='cuda:0') tensor([1653796800, 1653796800, 1653796800, ..., 1667188800, 1667188800,
1667188800], device='cuda:0') tensor([1546318800, 1546318800, 1546318800, ..., 1638075600, 1638075600,
1638075600], device='cuda:0')
examples/tgbl_flight_train.out 0 → 100644
LOCAL RANK 0, RANK0
use cuda on 0
18143
get_neighbors consume: 5.32513s
Epoch 0:
train loss:17328.4542 train ap:0.989251 val ap:0.991792 val auc:0.992826
total time:985.43s prep time:855.49s
fetch time:0.00s write back time:0.00s
Epoch 1:
train loss:14294.4074 train ap:0.992597 val ap:0.987955 val auc:0.988882
total time:981.59s prep time:851.99s
fetch time:0.00s write back time:0.00s
Epoch 2:
train loss:14297.0658 train ap:0.992655 val ap:0.991572 val auc:0.992562
total time:980.14s prep time:851.62s
fetch time:0.00s write back time:0.00s
Epoch 3:
train loss:14622.0160 train ap:0.992315 val ap:0.989207 val auc:0.990117
total time:1115.17s prep time:951.63s
fetch time:0.00s write back time:0.00s
Epoch 4:
train loss:14551.7185 train ap:0.992456 val ap:0.986963 val auc:0.988173
total time:1225.09s prep time:1063.97s
fetch time:0.00s write back time:0.00s
Epoch 5:
Early stopping at epoch 5
Loading the best model at epoch 0
0.975067675113678 0.9767603874206543
0.9743184447288513 0.9764328598976135
test AP:0.970799 test AUC:0.973111
test_dataset 10026943 avg_time 112.53466749191284
examples/tgbl_flight_train_4.out 0 → 100644
LOCAL RANK 0, RANK0
LOCAL RANK 2, RANK2
LOCAL RANK 1, RANK1
LOCAL RANK 3, RANK3
use cuda on 1
use cuda on 0
use cuda on 3
use cuda on 2
1814318143
18143
18143
get_neighbors consume: 5.95983s
get_neighbors consume: 6.05613s
get_neighbors consume: 6.14142s
num_batchs: tensor([19750], device='cuda:3')
num_batchs: tensor([17645], device='cuda:2')
num_batchs: tensor([20015], device='cuda:1')
get_neighbors consume: 6.29594s
num_batchs: tensor([20994], device='cuda:0')
num_batchs: num_batchs:num_batchs: num_batchs: tensor([3877], device='cuda:0')tensor([3605], device='cuda:2')
tensor([4524], device='cuda:3')
tensor([4708], device='cuda:1')
num_batchs:num_batchs:num_batchs: num_batchs: tensor([4149], device='cuda:0')tensor([4748], device='cuda:3')tensor([4298], device='cuda:1')tensor([3638], device='cuda:2')
Epoch 0:
Epoch 0:
Epoch 0:
Epoch 0:
train loss:8041.7524 train ap:0.963479 val ap:0.976991 val auc:0.979452
train loss:5974.9788 train ap:0.976456 val ap:0.976991 val auc:0.979452
train loss:6391.0674 train ap:0.969010 val ap:0.976991 val auc:0.979452
total time:369.28s prep time:310.18s
total time:369.28s prep time:310.18s
fetch time:0.00s write back time:0.00s
total time:369.28s prep time:310.18s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
train loss:6098.5427 train ap:0.979030 val ap:0.976991 val auc:0.979452
total time:369.28s prep time:310.18s
fetch time:0.00s write back time:0.00s
Epoch 1:
Epoch 1:
Epoch 1:
Epoch 1:
train loss:6785.4083 train ap:0.974335 val ap:0.981245 val auc:0.983090
train loss:5226.3762 train ap:0.979351 val ap:0.981245 val auc:0.983090
train loss:4392.9569 train ap:0.986932 val ap:0.981245 val auc:0.983090
total time:369.13s prep time:309.22s
total time:369.13s prep time:309.22s
total time:369.13s prep time:309.22s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
train loss:4860.7541 train ap:0.986432 val ap:0.981245 val auc:0.983090
total time:369.13s prep time:309.22s
fetch time:0.00s write back time:0.00s
Epoch 2:
Epoch 2:
Epoch 2:
Epoch 2:
train loss:4129.9429 train ap:0.988314 val ap:0.980251 val auc:0.982577
train loss:6498.5192 train ap:0.976384 val ap:0.980251 val auc:0.982577
train loss:4974.4880 train ap:0.981104 val ap:0.980251 val auc:0.982577
total time:365.46s prep time:306.60s
total time:365.46s prep time:306.60s
total time:365.46s prep time:306.60s
train loss:4656.6260 train ap:0.987497 val ap:0.980251 val auc:0.982577
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
total time:365.46s prep time:306.60s
fetch time:0.00s write back time:0.00s
Epoch 3:Epoch 3:
Epoch 3:
Epoch 3:
train loss:6346.4929 train ap:0.977303 val ap:0.978576 val auc:0.980634
train loss:4068.8824 train ap:0.988540 val ap:0.978576 val auc:0.980634
train loss:4918.4225 train ap:0.981662 val ap:0.978576 val auc:0.980634
total time:362.13s prep time:303.87s
train loss:4560.1019 train ap:0.987865 val ap:0.978576 val auc:0.980634
total time:362.13s prep time:303.87s
total time:362.13s prep time:303.87s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
total time:362.13s prep time:303.87s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
Epoch 4:Epoch 4:Epoch 4:
Epoch 4:
train loss:6310.2800 train ap:0.977605 val ap:0.974950 val auc:0.978057
train loss:3919.8300 train ap:0.989225 val ap:0.974950 val auc:0.978057
train loss:4854.9052 train ap:0.981859 val ap:0.974950 val auc:0.978057
total time:363.65s prep time:304.71s
total time:363.65s prep time:304.71s
fetch time:0.00s write back time:0.00s
total time:363.65s prep time:304.71s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
train loss:4467.4543 train ap:0.988407 val ap:0.974950 val auc:0.978057
total time:363.65s prep time:304.71s
fetch time:0.00s write back time:0.00s
Epoch 5:
Epoch 5:
Epoch 5:
Epoch 5:
train loss:6215.2998 train ap:0.978244 val ap:0.970081 val auc:0.973433
train loss:3909.4435 train ap:0.989068 val ap:0.970081 val auc:0.973433
train loss:4805.6621 train ap:0.982218 val ap:0.970081 val auc:0.973433
total time:367.40s prep time:307.71s
total time:367.40s prep time:307.71s
total time:367.40s prep time:307.71s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
fetch time:0.00s write back time:0.00s
train loss:4363.8407 train ap:0.988786 val ap:0.970081 val auc:0.973433
total time:367.40s prep time:307.71s
fetch time:0.00s write back time:0.00s
Epoch 6:
Epoch 6:
Epoch 6:
Epoch 6:
examples/tgbl_review.out 0 → 100644
LOCAL RANK 0, RANK0
initlize distributed
use cuda on 0
352637
get_neighbors consume: 1.38667s
raw file found, skipping download
Dataset directory is /home/zlj/.miniconda3/envs/dgnn-3.10/lib/python3.10/site-packages/tgb/datasets/tgbl_review
loading processed file
tensor([ 929232000, 930787200, 931824000, ..., 1538524800, 1538611200,
1538611200], device='cuda:0') tensor([ 929232000, 930787200, 931824000, ..., 1538524800, 1538611200,
1538611200])
tensor([1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000, 1464912000, 1464912000,
1464912000, 1464912000, 1464912000, 1464912000], device='cuda:0') tensor([1488844800, 1488844800, 1488844800, ..., 1538524800, 1538611200,
1538611200], device='cuda:0') tensor([ 929232000, 930787200, 931824000, ..., 1464825600, 1464825600,
1464825600], device='cuda:0')
val metric
metric hits@ is 0.545051806698618
metric mrr is 0.24419529163892895
metric ap is 0.05492426247254426
metric auc is 0.8336828640087952
test metric
metric hits@ is 0.50441967217847
metric mrr is 0.20610860927523572
metric ap is 0.03974634949228331
metric auc is 0.8243427933946369
test_dataset 728919 avg_time 0.0
examples/tgbl_review_4.out 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
examples/tgbl_review_train.out 0 → 100644
LOCAL RANK 0, RANK0
use cuda on 0
352637
get_neighbors consume: 1.137s
Epoch 0:
train loss:13642.4319 train ap:0.904599 val ap:0.877952 val auc:0.875205
total time:237.42s prep time:206.95s
fetch time:0.00s write back time:0.00s
Epoch 1:
train loss:13355.5800 train ap:0.909874 val ap:0.893142 val auc:0.885588
total time:254.75s prep time:224.74s
fetch time:0.00s write back time:0.00s
Epoch 2:
train loss:13198.5205 train ap:0.912843 val ap:0.880398 val auc:0.881145
total time:254.27s prep time:223.80s
fetch time:0.00s write back time:0.00s
Epoch 3:
train loss:13217.6608 train ap:0.912531 val ap:0.900547 val auc:0.891135
total time:246.83s prep time:217.90s
fetch time:0.00s write back time:0.00s
Epoch 4:
train loss:13175.3821 train ap:0.913456 val ap:0.896569 val auc:0.888247
total time:254.71s prep time:222.90s
fetch time:0.00s write back time:0.00s
Epoch 5:
train loss:13129.5668 train ap:0.914303 val ap:0.901932 val auc:0.891772
total time:256.61s prep time:225.98s
fetch time:0.00s write back time:0.00s
Epoch 6:
train loss:13110.9403 train ap:0.914760 val ap:0.899841 val auc:0.890011
total time:258.74s prep time:228.35s
fetch time:0.00s write back time:0.00s
Epoch 7:
train loss:13174.3921 train ap:0.913513 val ap:0.899906 val auc:0.890777
total time:246.85s prep time:217.08s
fetch time:0.00s write back time:0.00s
Epoch 8:
train loss:13082.4202 train ap:0.915001 val ap:0.896229 val auc:0.888280
total time:248.71s prep time:218.14s
fetch time:0.00s write back time:0.00s
Epoch 9:
train loss:13044.7357 train ap:0.915478 val ap:0.900687 val auc:0.891358
total time:251.59s prep time:220.47s
fetch time:0.00s write back time:0.00s
Epoch 10:
Early stopping at epoch 10
Loading the best model at epoch 5
0.7501348853111267 0.711144745349884
0.8450031280517578 0.8455020189285278
test AP:0.803709 test AUC:0.829060
test_dataset 728919 avg_time 48.60917709827423
examples/tgbl_wiki.out 0 → 100644
LOCAL RANK 0, RANK0
initlize distributed
use cuda on 0
9227
get_neighbors consume: 0.0126036s
raw file found, skipping download
Dataset directory is /home/zlj/.miniconda3/envs/dgnn-3.10/lib/python3.10/site-packages/tgb/datasets/tgbl_wiki
loading processed file
998
998
val metric
metric hits@ is 0.6441952827610548
metric mrr is 0.4429445469178228
metric ap is 0.13983213193432092
metric auc is 0.9535034098327113
998
998
test metric
metric hits@ is 0.5930238971918947
metric mrr is 0.4000823857097686
metric ap is 0.11770197066761524
metric auc is 0.9466253653660359
test_dataset 23621 avg_time 0.0
examples/tgbl_wiki_4.out 0 → 100644
LOCAL RANK 0, RANK0
initlize distributed
use cuda on 0
9227
get_neighbors consume: 0.0121001s
raw file found, skipping download
Dataset directory is /home/zlj/.miniconda3/envs/dgnn-3.10/lib/python3.10/site-packages/tgb/datasets/tgbl_wiki
loading processed file
998
998
val metric
metric hits@ is 0.5399417177408556
metric mrr is 0.2723149118104061
metric ap is 0.05431445631127528
metric auc is 0.934435814512715
998
998
test metric
metric hits@ is 0.4802390880222562
metric mrr is 0.23669700951704495
metric ap is 0.040643292031104734
metric auc is 0.923300786458167
test_dataset 23621 avg_time 0.0
examples/tgbl_wiki_train.out 0 → 100644
LOCAL RANK 0, RANK0
use cuda on 0
9227
get_neighbors consume: 0.0119422s
Epoch 0:
train loss:438.1820 train ap:0.905229 val ap:0.929162 val auc:0.928815
total time:10.91s prep time:9.85s
fetch time:0.00s write back time:0.00s
Epoch 1:
train loss:372.5819 train ap:0.926295 val ap:0.929234 val auc:0.926704
total time:8.88s prep time:7.75s
fetch time:0.00s write back time:0.00s
Epoch 2:
train loss:348.3494 train ap:0.939406 val ap:0.951674 val auc:0.948898
total time:9.72s prep time:8.57s
fetch time:0.00s write back time:0.00s
Epoch 3:
train loss:311.1370 train ap:0.952314 val ap:0.959255 val auc:0.956934
total time:9.88s prep time:8.77s
fetch time:0.00s write back time:0.00s
Epoch 4:
train loss:288.5227 train ap:0.959499 val ap:0.967144 val auc:0.964830
total time:9.83s prep time:8.71s
fetch time:0.00s write back time:0.00s
Epoch 5:
train loss:270.2249 train ap:0.964637 val ap:0.971685 val auc:0.969840
total time:9.53s prep time:8.41s
fetch time:0.00s write back time:0.00s
Epoch 6:
train loss:262.1692 train ap:0.966888 val ap:0.972277 val auc:0.970054
total time:9.58s prep time:8.50s
fetch time:0.00s write back time:0.00s
Epoch 7:
train loss:255.6310 train ap:0.968568 val ap:0.972964 val auc:0.971109
total time:9.37s prep time:8.28s
fetch time:0.00s write back time:0.00s
Epoch 8:
train loss:248.8265 train ap:0.970266 val ap:0.975745 val auc:0.974002
total time:9.68s prep time:8.53s
fetch time:0.00s write back time:0.00s
Epoch 9:
train loss:243.9315 train ap:0.971279 val ap:0.976228 val auc:0.974509
total time:9.86s prep time:8.69s
fetch time:0.00s write back time:0.00s
Epoch 10:
train loss:239.1431 train ap:0.972743 val ap:0.977532 val auc:0.975514
total time:9.80s prep time:8.62s
fetch time:0.00s write back time:0.00s
Epoch 11:
train loss:237.8836 train ap:0.972920 val ap:0.977701 val auc:0.976333
total time:9.93s prep time:8.76s
fetch time:0.00s write back time:0.00s
Epoch 12:
train loss:231.7141 train ap:0.973696 val ap:0.978296 val auc:0.976799
total time:9.90s prep time:8.73s
fetch time:0.00s write back time:0.00s
Epoch 13:
train loss:230.5749 train ap:0.974243 val ap:0.978770 val auc:0.977488
total time:9.77s prep time:8.62s
fetch time:0.00s write back time:0.00s
Epoch 14:
train loss:227.9846 train ap:0.974771 val ap:0.978397 val auc:0.977168
total time:9.80s prep time:8.63s
fetch time:0.00s write back time:0.00s
Epoch 15:
train loss:224.3624 train ap:0.975223 val ap:0.980206 val auc:0.979011
total time:9.78s prep time:8.68s
fetch time:0.00s write back time:0.00s
Epoch 16:
train loss:223.1655 train ap:0.975816 val ap:0.981120 val auc:0.979900
total time:9.62s prep time:8.51s
fetch time:0.00s write back time:0.00s
Epoch 17:
train loss:219.1989 train ap:0.976670 val ap:0.981726 val auc:0.980641
total time:9.79s prep time:8.69s
fetch time:0.00s write back time:0.00s
Epoch 18:
train loss:215.7983 train ap:0.977476 val ap:0.981537 val auc:0.980316
total time:10.03s prep time:8.84s
fetch time:0.00s write back time:0.00s
Epoch 19:
train loss:217.2757 train ap:0.976921 val ap:0.981455 val auc:0.980277
total time:9.82s prep time:8.66s
fetch time:0.00s write back time:0.00s
Epoch 20:
train loss:219.2030 train ap:0.976782 val ap:0.981089 val auc:0.980011
total time:10.38s prep time:9.28s
fetch time:0.00s write back time:0.00s
Epoch 21:
train loss:219.9309 train ap:0.976416 val ap:0.981670 val auc:0.980690
total time:9.43s prep time:8.32s
fetch time:0.00s write back time:0.00s
Epoch 22:
train loss:214.2197 train ap:0.977587 val ap:0.982226 val auc:0.981129
total time:9.75s prep time:8.53s
fetch time:0.00s write back time:0.00s
Epoch 23:
train loss:208.9837 train ap:0.978911 val ap:0.982907 val auc:0.981704
total time:11.65s prep time:10.40s
fetch time:0.00s write back time:0.00s
Epoch 24:
train loss:210.4146 train ap:0.978243 val ap:0.982097 val auc:0.980691
total time:10.86s prep time:9.73s
fetch time:0.00s write back time:0.00s
Epoch 25:
train loss:210.4207 train ap:0.978632 val ap:0.982267 val auc:0.981537
total time:9.95s prep time:8.77s
fetch time:0.00s write back time:0.00s
Epoch 26:
train loss:205.5232 train ap:0.979174 val ap:0.983918 val auc:0.982727
total time:9.51s prep time:8.42s
fetch time:0.00s write back time:0.00s
Epoch 27:
train loss:204.8931 train ap:0.979227 val ap:0.983066 val auc:0.982013
total time:9.13s prep time:8.08s
fetch time:0.00s write back time:0.00s
Epoch 28:
train loss:199.6552 train ap:0.980440 val ap:0.982168 val auc:0.981335
total time:9.11s prep time:8.04s
fetch time:0.00s write back time:0.00s
Epoch 29:
train loss:202.9698 train ap:0.979732 val ap:0.981972 val auc:0.981716
total time:9.21s prep time:8.11s
fetch time:0.00s write back time:0.00s
Epoch 30:
train loss:201.1851 train ap:0.980338 val ap:0.983631 val auc:0.982533
total time:9.40s prep time:8.30s
fetch time:0.00s write back time:0.00s
Epoch 31:
train loss:202.0885 train ap:0.979852 val ap:0.984241 val auc:0.983195
total time:9.11s prep time:7.96s
fetch time:0.00s write back time:0.00s
Epoch 32:
train loss:195.8186 train ap:0.981171 val ap:0.985042 val auc:0.984066
total time:9.39s prep time:8.17s
fetch time:0.00s write back time:0.00s
Epoch 33:
train loss:195.5999 train ap:0.980943 val ap:0.984088 val auc:0.983118
total time:9.81s prep time:8.62s
fetch time:0.00s write back time:0.00s
Epoch 34:
train loss:195.3828 train ap:0.981070 val ap:0.984907 val auc:0.983951
total time:10.13s prep time:8.90s
fetch time:0.00s write back time:0.00s
Epoch 35:
train loss:194.4766 train ap:0.981191 val ap:0.985196 val auc:0.984022
total time:10.03s prep time:8.93s
fetch time:0.00s write back time:0.00s
Epoch 36:
train loss:194.5252 train ap:0.981201 val ap:0.984551 val auc:0.983756
total time:9.26s prep time:8.19s
fetch time:0.00s write back time:0.00s
Epoch 37:
train loss:193.6458 train ap:0.981244 val ap:0.985164 val auc:0.984187
total time:9.13s prep time:8.07s
fetch time:0.00s write back time:0.00s
Epoch 38:
train loss:195.7096 train ap:0.981002 val ap:0.983946 val auc:0.982756
total time:9.05s prep time:7.99s
fetch time:0.00s write back time:0.00s
Epoch 39:
train loss:195.2296 train ap:0.981059 val ap:0.985010 val auc:0.983845
total time:9.11s prep time:8.04s
fetch time:0.00s write back time:0.00s
Epoch 40:
train loss:191.4868 train ap:0.981932 val ap:0.985390 val auc:0.984618
total time:9.09s prep time:8.03s
fetch time:0.00s write back time:0.00s
Epoch 41:
train loss:189.6005 train ap:0.981906 val ap:0.984920 val auc:0.984112
total time:9.05s prep time:7.98s
fetch time:0.00s write back time:0.00s
Epoch 42:
train loss:191.7600 train ap:0.981801 val ap:0.984937 val auc:0.984145
total time:9.18s prep time:8.12s
fetch time:0.00s write back time:0.00s
Epoch 43:
train loss:193.1520 train ap:0.981423 val ap:0.984112 val auc:0.983052
total time:9.07s prep time:7.98s
fetch time:0.00s write back time:0.00s
Epoch 44:
train loss:193.6521 train ap:0.981466 val ap:0.984804 val auc:0.983935
total time:9.35s prep time:8.24s
fetch time:0.00s write back time:0.00s
Epoch 45:
Early stopping at epoch 45
Loading the best model at epoch 40
0.9833124876022339 0.9827097654342651
0.9832088351249695 0.982053816318512
test AP:0.978572 test AUC:0.977290
test_dataset 23621 avg_time 7.839827566146851
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment