-
Notifications
You must be signed in to change notification settings - Fork 5
/
run_offline_hgt_tokenizer.py
173 lines (148 loc) · 7.2 KB
/
run_offline_hgt_tokenizer.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
# Copyright 2023 Haotian Liu
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import List, Optional, Tuple, Union
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import CrossEntropyLoss
from transformers import AutoConfig, AutoModelForCausalLM, \
LlamaConfig, LlamaModel, LlamaForCausalLM, \
CLIPVisionModel, CLIPImageProcessor
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
from higpt.model.graph_layers import MPNN, GNN, CLIP, graph_transformer
from higpt.model.meta_hgt import MetaHGTConvCfg, MetaHGTConv
from higpt.model.heteclip_models import Transformer, LayerNorm, CLIPTextCfg
from torch_geometric.data import Data
import json
import os.path as osp
import glob
from tqdm import tqdm
from lightning.pytorch import LightningModule, Trainer, seed_everything
import re
import os
import argparse
DEFAULT_GRAPH_TOKEN = "<graph>"
DEFAULT_GRAPH_PATCH_TOKEN = "<g_patch>"
DEFAULT_G_START_TOKEN = "<g_start>"
DEFAULT_G_END_TOKEN = "<g_end>"
class HeteroLlamaConfig(LlamaConfig):
model_type = "HeteroLlama"
device = 'cuda:2'
def load_metahgt_pretrained(model_name, pretrain_model_path):
# load conig json
assert osp.exists(osp.join(pretrain_model_path, 'graph_config.json')), 'graph_config.json missing'
with open(osp.join(pretrain_model_path, 'graph_config.json'), 'r') as f:
graph_config_dict = json.load(f)
graph_cfg = MetaHGTConvCfg(**graph_config_dict)
assert osp.exists(osp.join(pretrain_model_path, 'text_config.json')), 'text_config.json missing'
with open(osp.join(pretrain_model_path, 'text_config.json'), 'r') as f:
text_config_dict = json.load(f)
text_cfg = CLIPTextCfg(**text_config_dict)
assert model_name == MetaHGTConv
model = model_name(in_channels = graph_cfg.in_channels,
out_channels = graph_cfg.out_channels,
heads = graph_cfg.heads,
dynamic = graph_cfg.dynamic,
text_cfg = text_cfg,)
pkl_files = glob.glob(osp.join(pretrain_model_path, '*.ckpt'))
state_dict = torch.load(pkl_files[0], map_location = 'cpu')['state_dict']
print('loading graph pre train model ...')
gnn_state_dict = {}
for key, value in state_dict.items():
if key.startswith('model.graph_encoder'):
new_key = key.split('model.graph_encoder.')[1]
gnn_state_dict[new_key] = value
model.load_state_dict(gnn_state_dict, strict=False)
return model
def check_offline_hgt():
parser = argparse.ArgumentParser(description='check data')
parser.add_argument('--pretrained_gnn_path', default='/root/paddlejob/workspace/env_run/output/HeteGPT/MetaHGT', type=str)
parser.add_argument('--graph_root', default='/root/paddlejob/workspace/env_run/output/HetBaseline/data/DBLP', type=str)
parser.add_argument('--dsname', default='instruct_ds', type=str)
parser.add_argument('--data_type', default='dblp', type=str)
args = parser.parse_args()
hgnn_name = args.pretrained_gnn_path.split('/')[-1]
hgt = load_metahgt_pretrained(MetaHGTConv, args.pretrained_gnn_path)
hgt = hgt.to(device)
graph_root = args.graph_root
dsname = args.dsname
node_feas_dict = torch.load(f'/root/paddlejob/workspace/env_run/output/HeteGPT/hetegpt/model/meta_hgt/meta_dict/{args.data_type}/node_type.pt')
for k, v in node_feas_dict.items():
node_feas_dict[k] = v.to(device)
edge_feas_dict = torch.load(f'/root/paddlejob/workspace/env_run/output/HeteGPT/hetegpt/model/meta_hgt/meta_dict/{args.data_type}/edge_type.pt')
for k, v in edge_feas_dict.items():
edge_feas_dict[k] = v.to(device)
ann_data_path = []
# dsname = dsname.split(',')
# for i, path in enumerate(dsname):
ann_file = glob.glob(osp.join(graph_root, dsname, 'ann', '**/*.json'), recursive=True)
# assert len(ann_file) == 1, f"Need to have one ann file for each graph"
ann_data_path.extend(ann_file)
# json_file = osp.join(graph_root, dsname, 'ann', 'DBLP_train_std_0_400.json')
data_list_dict = {}
for ann_file in (ann_data_path):
file_name = ann_file.split('/')[-1]
print(file_name)
with open(ann_file, 'r', encoding='utf-8') as f:
data_item = json.load(f)
data_list_dict[file_name] = data_item
processed_data_list = []
count_cnt = 0
for file_name, data_list in data_list_dict.items():
for data_item in tqdm(data_list):
# if count_cnt > 0:
# break
graph_path = osp.join(graph_root, data_item['graph']['graph'])
graph_dict = torch.load(graph_path)
graph_dict = graph_dict.to(device)
if args.data_type == 'dblp':
new_conf_feas = torch.ones([graph_dict['conference'].num_nodes, 768])
new_conf_feas = new_conf_feas.to(device)
graph_dict['conference'].x = new_conf_feas
with torch.no_grad():
res = hgt(x_dict = graph_dict.x_dict,
edge_index_dict = graph_dict.edge_index_dict, # Support both.
node_type_feas_dict = node_feas_dict,
edge_type_feas_dict = edge_feas_dict)
for k, v in res.items():
res[k] = v.cpu()
if torch.any(torch.isnan(res[k])):
print(k, res[k])
raise ValueError
processed_graph_path = re.sub('graph_data', f'graph_data_processed_{hgnn_name}', data_item['graph']['graph'])
processed_data_item = data_item.copy()
processed_data_item['graph']['graph'] = processed_graph_path
processed_data_list.append(processed_data_item.copy())
save_path = osp.join(graph_root, processed_graph_path)
if osp.exists(osp.dirname(save_path)) is False:
os.makedirs(osp.dirname(save_path), exist_ok=True)
# print(graph_dict.x_dict['author'])
# for k, v in graph_dict.items():
# graph_dict.x_dict[k] = res[k]
graph_dict = graph_dict.to('cpu')
graph_dict.x_dict = res
# print(graph_dict.x_dict['author'])
# print(res['author'])
torch.save(graph_dict, save_path)
count_cnt += 1
# time.sleep(1)
processed_json_file = osp.join(graph_root, dsname, f'ann_processed_{hgnn_name}', file_name)
if osp.exists(osp.dirname(processed_json_file)) is False:
os.makedirs(osp.dirname(processed_json_file), exist_ok=True)
with open(processed_json_file, 'w', encoding='utf-8') as f:
json.dump(processed_data_list, f)
if __name__ == '__main__':
seed_everything(42)
check_offline_hgt()