/
combined_pair_loss_landscape.py
168 lines (147 loc) · 7.39 KB
/
combined_pair_loss_landscape.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
import os
os.chdir(os.path.dirname(os.path.realpath(__file__)))
import sys
import numpy as np
import logging
import fuxictr_version
from fuxictr import datasets
from datetime import datetime
from fuxictr.utils import load_config, set_logger, print_to_json, print_to_list
from fuxictr.features import FeatureMap
from fuxictr.pytorch.torch_utils import seed_everything
from fuxictr.pytorch.dataloaders import H5DataLoader
from fuxictr.preprocess import FeatureProcessor, build_dataset
import src as model_zoo
import gc
import argparse
import os
from pathlib import Path
import pandas as pd
import torch
from tqdm import tqdm
import matplotlib.pyplot as plt
import numpy as np
from mpl_toolkits.mplot3d import Axes3D
def set_logger(params):
dataset_id = params['dataset_id']
model_id = params.get('model_id', '')
log_dir = os.path.join(params.get('model_root', './checkpoints'), dataset_id)
os.makedirs(log_dir, exist_ok=True)
log_file = os.path.join(log_dir, model_id + '#inference.log')
for handler in logging.root.handlers[:]:
logging.root.removeHandler(handler)
logging.basicConfig(level=logging.INFO,
format='%(asctime)s P%(process)d %(levelname)s %(message)s',
handlers=[logging.FileHandler(log_file, mode='w'),
logging.StreamHandler()])
if __name__ == '__main__':
''' Usage: python inference.py --config {config_dir} --expid {experiment_id} --gpu {gpu_device_id}
'''
parser = argparse.ArgumentParser()
parser.add_argument('--config', type=str, default='./config/', help='The config directory.')
parser.add_argument('--expid', type=str, default='DeepFM_test', help='The experiment id to run.')
parser.add_argument('--gpu', type=int, default=-1, help='The gpu index, -1 for cpu')
parser.add_argument('--xmin', type=float, default=-5.)
parser.add_argument('--xmax', type=float, default=10.)
parser.add_argument('--xnum', type=int, default=30)
parser.add_argument('--ymin', type=float, default=-10.)
parser.add_argument('--ymax', type=float, default=5.)
parser.add_argument('--ynum', type=int, default=30)
args = vars(parser.parse_args())
xmin, xmax, xnum = args['xmin'], args['xmax'], args['xnum']
ymin, ymax, ynum = args['ymin'], args['ymax'], args['ynum']
experiment_id = args['expid']
params = load_config(args['config'], experiment_id)
params['gpu'] = args['gpu']
set_logger(params)
logging.info("Params: " + print_to_json(params))
seed_everything(seed=params['seed'])
data_dir = os.path.join(params['data_root'], params['dataset_id'])
feature_map_json = os.path.join(data_dir, "feature_map.json")
if params["data_format"] == "csv":
# Build feature_map and transform h5 data
feature_encoder = FeatureProcessor(**params)
params["train_data"], params["valid_data"], params["test_data"] = \
build_dataset(feature_encoder, **params)
feature_map = FeatureMap(params['dataset_id'], data_dir)
feature_map.load(feature_map_json, params)
logging.info("Feature specs: " + print_to_json(feature_map.features))
model_class = getattr(model_zoo, params['model'])
model = model_class(feature_map, **params)
model.count_parameters() # print number of parameters used in model
# model.load_weights(model.checkpoint)
params['shuffle'] = False
test_gen = H5DataLoader(feature_map, stage='test', **params).make_iterator()
model.to(model.device)
state_dict = torch.load(model.checkpoint, map_location="cpu")
model.load_state_dict(state_dict)
cat_weight = torch.cat([state_dict['fc.weight'],state_dict['fc.bias'].reshape(1,-1)],dim=1)
l2_norm = torch.norm(cat_weight, p=2)
fc_weight_shape = state_dict['fc.weight'].shape
fc_bias_shape = state_dict['fc.bias'].shape
def compute_loss(y_pred,logit,y_true):
weight = torch.where(y_true==1., model.pos_weight, 1)
index_pos = torch.nonzero(y_true.reshape(-1,)==1).reshape(-1,) # (1049,)
index_neg = torch.nonzero(y_true.reshape(-1,)==0).reshape(-1,)
cat = torch.cartesian_prod(index_pos, index_neg)
pos_logit = logit[cat[:,0]] # (3196303,)
neg_logit = logit[cat[:,1]] # (3196303,)
left = torch.cat((pos_logit,neg_logit),dim=0)
right = torch.cat((neg_logit,pos_logit),dim=0)
cmp_true = torch.cat((
torch.ones(len(pos_logit),1),
torch.zeros(len(pos_logit),1)
),dim=0).to(model.device)
pairwise_loss = torch.nn.functional.binary_cross_entropy(torch.sigmoid(left-right),cmp_true, reduction='mean')
pointwise_loss = torch.nn.functional.binary_cross_entropy(y_pred, y_true, reduction='mean',weight=weight)
loss = model.alpha* pairwise_loss+ (1-model.alpha)*pointwise_loss
return loss, pointwise_loss, pairwise_loss
def orthogonal_unit_vector(vec):
vector = vec.reshape(-1,)
unit_vector = vector / torch.norm(vector,p=2)
random_vector = torch.randn_like(vector)
dot_product = torch.dot(unit_vector, random_vector)
orthogonal_vector = random_vector - dot_product * unit_vector
orthogonal_unit_vector = orthogonal_vector / torch.norm(orthogonal_vector,p=2)
return orthogonal_unit_vector.reshape(1,-1)
alpha = np.linspace(xmin,xmax, xnum)
beta = np.linspace(ymin,ymax, ynum)
begin_vec = cat_weight #（1，891），最后一位是bias，前面890位是weight
delta_vec = cat_weight/l2_norm
niu_vec = orthogonal_unit_vector(delta_vec)
logging.info(torch.dot(niu_vec.reshape(-1,), delta_vec.reshape(-1,)))
def calculate_added_vector(wa,va,wb,vb):
return begin_vec + wa * va + wb * vb
if test_gen:
model.eval() # set to evaluation mode
alphas = []
betas= []
combined_pair_losses = []
pointwise_losses = []
pairwise_losses = []
with torch.no_grad():
y_pred = []
y_true = []
group_id = []
if model._verbose > 0:
data_generator = tqdm(test_gen, disable=False, file=sys.stdout)
for batch_data in data_generator:
# only consider the first batch
for wb in tqdm(beta):
for wa in alpha: # moving x-axis
vec = calculate_added_vector(wa, delta_vec, wb, niu_vec)
weight = vec[0,:-1]
bias = vec[0,-1]
model.fc.weight.data = weight.reshape(fc_weight_shape).to(model.device)
model.fc.bias.data = bias.reshape(fc_bias_shape).to(model.device)
return_dict = model.forward(batch_data)
y_true = model.get_labels(batch_data)
y_pred = return_dict["y_pred"]
logit = return_dict["logit"]
combined_pair, pointwise_loss, pairwise_loss = compute_loss(y_pred,logit,y_true)
alphas.append(wa)
betas.append(wb)
combined_pair_losses.append(float(combined_pair))
break
alpha_grid, beta_grid = np.meshgrid(alpha, beta)
np.save('./analysis_tools/loss_landscape/combined_pair_loss.npy',np.array(combined_pair_losses).reshape(alpha_grid.shape))