-
Notifications
You must be signed in to change notification settings - Fork 4
/
inference_on_single_image.py
212 lines (180 loc) · 11.5 KB
/
inference_on_single_image.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
"""The inference script for DDPM model for a single image.
This script is not a part of the pipeline of the project. It
was used to generate plots and statistics for a single
data sample case.
"""
import argparse
import logging
import os
import warnings
from collections import OrderedDict
import numpy as np
import torch
from torch.nn.functional import mse_loss, l1_loss
from torch.utils.data import DataLoader
import model
from config import Config, get_current_datetime
from utils import dict2str, setup_logger, construct_and_save_wbd_plots, \
construct_mask, reverse_transform_candidates, set_seeds
from weatherbench_data import collate_wb_batch
from weatherbench_data.utils import reverse_transform, reverse_transform_tensor, load_object, prepare_test_data
warnings.filterwarnings("ignore")
if __name__ == "__main__":
set_seeds() # For reproducability.
parser = argparse.ArgumentParser()
parser.add_argument("-c", "--config", type=str, help="JSON file for configuration")
parser.add_argument("-p", "--phase", type=str, choices=["train", "val"],
help="Run either training or validation(inference).", default="train")
parser.add_argument("-gpu", "--gpu_ids", type=str, default=None)
args = parser.parse_args()
configs = Config(args)
torch.backends.cudnn.enabled = True
torch.backends.cudnn.benchmark = True
test_root = f"{configs.experiments_root}/test_on_single_image_{get_current_datetime()}"
os.makedirs(test_root, exist_ok=True)
setup_logger("test", test_root, "test", screen=True)
val_logger = logging.getLogger("test")
val_logger.info(dict2str(configs.get_hyperparameters_as_dict()))
# Preparing testing data.
transformations = load_object(configs.experiments_root, "transformations")
metadata = load_object(configs.experiments_root, "metadata")
if transformations and metadata:
val_logger.info("Transformations and Metadata are successfuly loaded.")
val_dataset = prepare_test_data(variables=configs.variables, val_min_date=configs.val_min_date,
val_max_date=configs.val_max_date, dataroot=configs.dataroot,
transformations=transformations)
val_logger.info(f"Dataset [{val_dataset.__class__.__name__} - Testing] is created.")
val_logger.info(f"""Created {val_dataset.__class__.__name__} dataset of length {len(val_dataset)},
containing data from {val_dataset.min_date} until {val_dataset.max_date}""")
val_logger.info(f"Group structure: {val_dataset.get_data_names()}")
val_logger.info(f"Channel count: {val_dataset.get_channel_count()}\n")
# The batch size is 1.
val_loader = DataLoader(val_dataset,
collate_fn=collate_wb_batch,
pin_memory=True, drop_last=True,
num_workers=configs.num_workers)
val_logger.info("Testing dataset is ready.")
# Defining the model.
diffusion = model.create_model(in_channel=configs.in_channel, out_channel=configs.out_channel,
norm_groups=configs.norm_groups, inner_channel=configs.inner_channel,
channel_multiplier=configs.channel_multiplier, attn_res=configs.attn_res,
res_blocks=configs.res_blocks, dropout=configs.dropout,
diffusion_loss=configs.diffusion_loss, conditional=configs.conditional,
gpu_ids=configs.gpu_ids, distributed=configs.distributed,
init_method=configs.init_method, train_schedule=configs.train_schedule,
train_n_timestep=configs.train_n_timestep,
train_linear_start=configs.train_linear_start,
train_linear_end=configs.train_linear_end,
val_schedule=configs.val_schedule, val_n_timestep=configs.val_n_timestep,
val_linear_start=configs.val_linear_start, val_linear_end=configs.val_linear_end,
finetune_norm=configs.finetune_norm, optimizer=None, amsgrad=configs.amsgrad,
learning_rate=configs.lr, checkpoint=configs.checkpoint,
resume_state=configs.resume_state, phase=configs.phase, height=configs.height)
val_logger.info("Model initialization is finished.")
current_step, current_epoch = diffusion.begin_step, diffusion.begin_epoch
val_logger.info(f"Testing the model at epoch: {current_epoch}, iter: {current_step}.")
diffusion.set_new_noise_schedule(schedule=configs.test_schedule,
n_timestep=configs.test_n_timestep,
linear_start=configs.test_linear_start,
linear_end=configs.test_linear_end)
accumulated_statistics = OrderedDict()
# Creating placeholder for storing validation metrics Mean Squared Error, Root MSE, Mean Residual.
val_metrics = OrderedDict({"MSE": 0.0, "RMSE": 0.0, "MAE": 0.0, "MR": 0.0,
"mean_bias_over_pixels": 0.0, "std_bias_over_pixels": 0.0})
idx = 0
result_path = f"{test_root}/results"
os.makedirs(result_path, exist_ok=True)
with torch.no_grad():
for val_data in val_loader:
idx += 1
# Works only for one image.
if idx % configs.val_vis_freq == 0:
diffusion.feed_data(val_data)
val_logger.info("Starting to generate SR images.")
diffusion.test(continuous=True) # Continues=False to return only the last timesteps's outcome.
val_logger.info("Finished generating SR images.")
# Computing metrics on vlaidation data..
visuals = diffusion.get_current_visuals()
# When continuous is True, visuals["SR"] has [T, C, H, W] dimension
# where T is the number of diffusion timesteps.
inv_visuals = reverse_transform(visuals, transformations,
configs.variables, diffusion.get_months(),
configs.tranform_monthly)
# Computing MSE and RMSE on original data.
mse_value = mse_loss(inv_visuals["HR"], inv_visuals["SR"][-1])
val_metrics["MSE"] += mse_value
val_metrics["RMSE"] += torch.sqrt(mse_value)
val_metrics["MAE"] += l1_loss(inv_visuals["HR"], inv_visuals["SR"][-1])
# Computing residuals for visualization.
residuals = inv_visuals["SR"][-1] - inv_visuals["HR"]
val_metrics["MR"] += residuals.mean()
path = f"{result_path}/{idx}/"
os.makedirs(path, exist_ok=True)
path = f"{path}{idx}"
val_logger.info("Started generating multiple SR candidates.")
sr_candidates = diffusion.generate_multiple_candidates(n=configs.sample_size)
reverse_transform_candidates(sr_candidates, reverse_transform_tensor,
transformations, configs.variables,
"hr", diffusion.get_months(),
configs.tranform_monthly)
val_logger.info("Finished generating multiple SR candidates.")
mean_candidate = sr_candidates.mean(dim=0)
std_candidate = sr_candidates.std(dim=0)
bias = mean_candidate - inv_visuals["HR"]
mean_bias_over_pixels = bias.mean()
std_bias_over_pixels = bias.std()
val_metrics["mean_bias_over_pixels"] += mean_bias_over_pixels
val_metrics["std_bias_over_pixels"] += std_bias_over_pixels
# Computing min and max measures to set a fixed colorbar for all visualizations.
vmin = min(inv_visuals["HR"].min(),
inv_visuals["LR"].min(),
inv_visuals["INTERPOLATED"].min(),
mean_candidate.min())
vmax = max(inv_visuals["HR"].max(),
inv_visuals["LR"].max(),
inv_visuals["INTERPOLATED"].max(),
mean_candidate.max())
vmin, vmax = np.floor(vmin), np.ceil(vmax)
val_logger.info(f"[{idx // configs.val_vis_freq}] Visualizing and storing some examples.")
# Choosing the first n_val_vis number of samples to visualize.
construct_and_save_wbd_plots(latitude=metadata.hr_lat, longitude=metadata.hr_lon,
data=inv_visuals["HR"],
path=f"{path}_hr.png", vmin=vmin, vmax=vmax)
construct_and_save_wbd_plots(latitude=metadata.hr_lat, longitude=metadata.hr_lon,
data=inv_visuals["SR"],
path=f"{path}_sr.png", vmin=vmin, vmax=vmax)
construct_and_save_wbd_plots(latitude=metadata.lr_lat, longitude=metadata.lr_lon,
data=inv_visuals["LR"],
path=f"{path}_lr.png", vmin=vmin, vmax=vmax)
construct_and_save_wbd_plots(latitude=metadata.hr_lat, longitude=metadata.hr_lon,
data=inv_visuals["INTERPOLATED"],
path=f"{path}_interpolated.png", vmin=vmin, vmax=vmax)
construct_and_save_wbd_plots(latitude=metadata.hr_lat, longitude=metadata.hr_lon,
data=construct_mask(residuals),
path=f"{path}_residual.png", vmin=-1, vmax=1,
costline_color="red", cmap="binary",
label="Signum(SR - HR)")
construct_and_save_wbd_plots(latitude=metadata.hr_lat, longitude=metadata.hr_lon,
data=mean_candidate,
path=f"{path}_mean_sr.png", vmin=vmin, vmax=vmax)
construct_and_save_wbd_plots(latitude=metadata.hr_lat, longitude=metadata.hr_lon,
data=std_candidate,
path=f"{path}_std_sr.png", vmin=0.0, cmap="Greens")
normalizing_constant = idx // configs.val_vis_freq
val_metrics["MSE"] /= normalizing_constant
val_metrics["RMSE"] /= normalizing_constant
val_metrics["MR"] /= normalizing_constant
val_metrics["MAE"] /= normalizing_constant
val_metrics["mean_bias_over_pixels"] /= normalizing_constant
val_metrics["std_bias_over_pixels"] /= normalizing_constant
message = f"Epoch: {current_epoch:5} | Iteration: {current_step:8}"
for metric, value in val_metrics.items():
message = f"{message} | {metric:s}: {value:.5f}"
val_logger.info(message)
torch.save(inv_visuals["LR"], f"{path}_LR.pt")
torch.save(inv_visuals["HR"], f"{path}_HR.pt")
torch.save(inv_visuals["INTERPOLATED"], f"{path}_INTERPOLATED.pt")
torch.save(inv_visuals["SR"], f"{path}_SR.pt")
torch.save(mean_candidate, f"{path}_mean_sr.pt")
torch.save(std_candidate, f"{path}_std.pt")
val_logger.info("End of testing.")