-
Notifications
You must be signed in to change notification settings - Fork 0
/
_script_embed_all_no_paral.py
227 lines (201 loc) · 12 KB
/
_script_embed_all_no_paral.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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
from _script_overlap_computation import *
import tqdm
import matplotlib.pyplot as plt
def run_experiment_instance(model_file: str, table_dict: dict | str=None, graph_dataset: str | dict=None, embeddings: dict=None) -> dict:
"""Given a table dict this function computes some performance measures about the speed of the framework in the embedding generation
Args:
model_file (str): path to a model checkpoint of GNNTE
table_dict_path (dict): instanced table_dict
embeddings (dict): dictionary containing the computed embeddings
Returns:
dict: dictionary containing the results
"""
print('Loading model....')
model = GNNTE(model_file=model_file)
print('Model loaded')
if graph_dataset == None:
if isinstance(table_dict, str):
with open(table_dict, 'rb') as f:
table_dict = pickle.load(f)
in_channels = model.in_channels
print('Loading embedding_buffer....')
if in_channels == 300:
embedding_buffer = FasttextEmbeddingBuffer(model='fasttext-wiki-news-subwords-300')
print('embedding_buffer loaded....')
print('Loading string_token_preprocessor....')
string_token_preprocessor = String_token_preprocessor()
print('string_token_preprocessor loaded')
else:
embedding_buffer = Hash_embedding_buffer()
print('embedding_buffer loaded....')
experiment_data = {}
for k in tqdm.tqdm(table_dict.keys()):
t = table_dict[k]
start = time.time()
#gen graph
try:
if in_channels == 300:
g = {k:Graph(table_dict[k], k, embedding_buffer, string_token_preprocessor, token_length_limit=None)}
else:
g = {k:Graph_Hashed_Node_Embs(table_dict[k], k)}
except:
continue
end_graph = time.time()
#gen emb
start_emb = time.time()
gd = GraphsDataset(g)
dataloader = DataLoader(gd, batch_size=1, num_workers=0, shuffle=False)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
embedding = embed(model, dataloader, device)
end = time.time()
data = {
'n_rows' : t.shape[0],
'n_cols' : t.shape[1],
'area' : t.shape[0]*t.shape[1],
't_graph_gen' : (end_graph-start),
't_emb_gen' : (end-start_emb),
't_tot' : (end-start)
}
experiment_data[k] = data
if embeddings != None:
embeddings[k] = embedding
else:
print(f'Loading graph dict from "{graph_dataset}"')
if isinstance(graph_dataset, str):
with open(graph_dataset, 'rb') as f:
graph_dataset = pickle.load(f)
experiment_data = {}
for k in tqdm.tqdm(graph_dataset.keys()):
g = {k:graph_dataset[k]}
#gen emb
start_emb = time.time()
gd = GraphsDataset(g)
dataloader = DataLoader(gd, batch_size=1, num_workers=0, shuffle=False)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
embedding = embed(model, dataloader, device)
end = time.time()
data = None
experiment_data[k] = data
if embeddings != None:
embeddings[k] = embedding
if embeddings != None:
return experiment_data, embeddings
else:
return experiment_data
def update_table_dict(table_dict: dict, experiment_data: dict) -> dict:
outliers = {}
for k in experiment_data.keys():
try:
if experiment_data[str(k)]['t_tot'] > 1000:
outliers[str(k)] = table_dict[str(k)]
except:
if experiment_data[k]['t_tot'] > 1000:
outliers[k] = table_dict[k]
return outliers
def run_experiment(model_file: str, table_dict_path: str | dict=None, graphs_path: str|dict=None, experiment_data_file_path: str=None, iters: int=1, embedding_file: str=None) -> dict:
print('Loading table_dict....')
if type(table_dict_path) is dict:
table_dict = table_dict_path
print('table_dict loaded')
elif isinstance(table_dict_path, str):
with open(table_dict_path, 'rb') as f:
table_dict = pickle.load(f)
print('table_dict loaded')
if isinstance(graphs_path, str):
print(f'Loading graph dict from "{graphs_path}"')
with open(graphs_path, 'rb') as f:
graphs_path = pickle.load(f)
print('Graph dict loaded')
experiment_data = {}
embeddings = {}
for _ in range(iters):
if len(experiment_data.values()) != 0:
table_dict = update_table_dict(table_dict, experiment_data)
new_exp_data = run_experiment_instance(model_file=model_file, table_dict=table_dict)
for k in new_exp_data.keys():
try:
if new_exp_data[str(k)]['t_tot'] < experiment_data[str(k)]['t_tot']:
experiment_data[str(k)] = new_exp_data[str(k)]
except:
if new_exp_data[k]['t_tot'] < experiment_data[k]['t_tot']:
experiment_data[k] = new_exp_data[k]
else:
experiment_data, embeddings = run_experiment_instance(model_file=model_file, graph_dataset=graphs_path, table_dict=table_dict_path, embeddings=embeddings)
if embedding_file != None:
with open(embedding_file,'wb') as f:
pickle.dump(embeddings, f)
if experiment_data_file_path:
with open(experiment_data_file_path, 'wb') as f:
pickle.dump(experiment_data,f)
if __name__ == '__main__':
# run_experiment(
# #model_file='/home/francesco.pugnaloni/GNNTE/models/model_wikidata_450k_GraphSAGE_50ep.pth',
# model_file='/home/francesco.pugnaloni/GNNTE/models/wikidata/model_wikidata_450k_GraphSAGE_50ep.pth',
# #table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/just_1k_tables.pkl',
# table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/full_table_dict_with_id.pkl',
# #experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/Datasets/just_1k_tables_stats.pkl",
# experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/tmp/tmp.pkl",
# embedding_file = '/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/embeddings/emb_wikifull_450k_15-02.pkl'
# )
# dd = pd.read_csv('/home/francesco.pugnaloni/GNNTE/test_data/t_exec/end_2_end_overlap_comparison/t_execs_compared_seconds_full_100tokens.csv')
# with open('/home/francesco.pugnaloni/GNNTE/Datasets/1_Gittables/table_dict_796970_good.pkl','rb') as f:
# table_dict = pickle.load(f)
# table_list = []
# for r in tqdm.tqdm(range(dd.shape[0])):
# table_list.append(dd.iloc[r]['r_id'])
# table_list.append(dd.iloc[r]['s_id'])
# table_list = set(table_list)
# table_dd = {}
# for k in table_list:
# table_dd[k] = table_dict[k]
# with open('/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/full_table_dict_with_id.pkl', 'rb') as f:
# table_dd = pickle.load(f)
# run_experiment(
# #model_file='/home/francesco.pugnaloni/GNNTE/models/model_wikidata_450k_GraphSAGE_50ep.pth',
# #model_file='/home/francesco.pugnaloni/GNNTE/models/wikidata/wikidata_06-03-24_GraphSAGE_50_ep_max_1000_tokens.pth',
# model_file = '/home/francesco.pugnaloni/GNNTE/model_wikitables.pth',
# #table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/just_1k_tables.pkl',
# table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/1_Gittables/table_dict_796970_good.pkl',
# iters=3,
# #table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/debug_files/tables.pkl',
# #graphs_path='/home/francesco.pugnaloni/GNNTE/Datasets/1_Gittables/balanced_datasets/graph_dict.pkl',
# #experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/t_exec/gen_emb_seq/gittables/embedding_time_gittables_sha256_64_epochs.pkl",
# #experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/tmp/tmp.pkl",
# experiment_data_file_path='/home/francesco.pugnaloni/GNNTE/efficiency_embedding_gen/experiments_efficiency_on_gittables_using_wikidata_arm.pkl',
# #embedding_file = '/home/francesco.pugnaloni/GNNTE/test_data/t_exec/end_2_end_overlap_comparison/embeddings_100token_test_gittables.pkl'
# # embedding_file='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/embeddings/emb_wiki_20_03_sha256.pkl'
# embedding_file='/home/francesco.pugnaloni/GNNTE/Datasets/1_Gittables/embeddings/embeddings_made_with_arm_trained_on_wikidata.pkl'
# )
print('Generating embeddings with wikitables for wikitabels')
run_experiment(
#model_file='/home/francesco.pugnaloni/GNNTE/models/model_wikidata_450k_GraphSAGE_50ep.pth',
#model_file='/home/francesco.pugnaloni/GNNTE/models/wikidata/wikidata_06-03-24_GraphSAGE_50_ep_max_1000_tokens.pth',
model_file = '/home/francesco.pugnaloni/GNNTE/model_wikitables.pth',
#table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/just_1k_tables.pkl',
table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/table_dict.pkl',
iters=3,
#table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/debug_files/tables.pkl',
#graphs_path='/home/francesco.pugnaloni/GNNTE/Datasets/1_Gittables/balanced_datasets/graph_dict.pkl',
#experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/t_exec/gen_emb_seq/gittables/embedding_time_gittables_sha256_64_epochs.pkl",
#experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/tmp/tmp.pkl",
experiment_data_file_path='/home/francesco.pugnaloni/GNNTE/efficiency_embedding_gen/experiments_efficiency_on_wikitables_using_wikitables_armadillo.pkl',
#embedding_file = '/home/francesco.pugnaloni/GNNTE/test_data/t_exec/end_2_end_overlap_comparison/embeddings_100token_test_gittables.pkl'
# embedding_file='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/embeddings/emb_wiki_20_03_sha256.pkl'
embedding_file='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/embeddings/embeddings_made_with_arm_trained_on_wikitables.pkl'
)
print('Generating embeddings with gittables for wikitabels')
run_experiment(
#model_file='/home/francesco.pugnaloni/GNNTE/models/model_wikidata_450k_GraphSAGE_50ep.pth',
#model_file='/home/francesco.pugnaloni/GNNTE/models/wikidata/wikidata_06-03-24_GraphSAGE_50_ep_max_1000_tokens.pth',
model_file = '/home/francesco.pugnaloni/GNNTE/best_model_gittables.pth',
#table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/just_1k_tables.pkl',
#table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/1_Gittables/table_dict_796970_good.pkl',
iters=1,
#table_dict_path='/home/francesco.pugnaloni/GNNTE/Datasets/debug_files/tables.pkl',
graphs_path='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/graph_dict.pkl',
#experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/t_exec/gen_emb_seq/gittables/embedding_time_gittables_sha256_64_epochs.pkl",
#experiment_data_file_path="/home/francesco.pugnaloni/GNNTE/test_data/tmp/tmp.pkl",
#embedding_file = '/home/francesco.pugnaloni/GNNTE/test_data/t_exec/end_2_end_overlap_comparison/embeddings_100token_test_gittables.pkl'
# embedding_file='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/embeddings/emb_wiki_20_03_sha256.pkl'
embedding_file='/home/francesco.pugnaloni/GNNTE/Datasets/2_WikiTables/embeddings/embeddings_made_with_arm_trained_on_gittables.pkl'
)