-
Notifications
You must be signed in to change notification settings - Fork 8
/
evaluate.py
179 lines (152 loc) · 7.07 KB
/
evaluate.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
# fix for keras v3.0 update
import os
os.environ['TF_USE_LEGACY_KERAS'] = '1'
# python based
import tensorflow as tf
import random
from pathlib import Path
import pandas as pd
import argparse
import submitit
import json
import numpy as np
import shutil
# custom code
from dataloaders.OptimizedDataGenerator import OptimizedDataGenerator
from models import CreateModel
minval=1e-9
def evaluate(config):
# update %j with actual job number
try:
job_env = submitit.JobEnvironment()
config["outFileName"] = Path(str(config["outFileName"]).replace("%j", str(job_env.job_id)))
except:
config["outFileName"] = Path(str(config["outFileName"]).replace("%j", "%08x" % random.randrange(16**8)))
output_directory = config["outFileName"].parent
os.makedirs(output_directory, exist_ok=True)
print(output_directory)
# create tf records directory
tfrecords_dir = Path(output_directory, f"tfrecords_{'%08x' % random.randrange(16**8)}").resolve()
# data generator
test_generator = OptimizedDataGenerator(
data_directory_path = config["data_directory_path"],
labels_directory_path = config["labels_directory_path"],
is_directory_recursive = False,
file_type = "parquet",
data_format = "3D",
batch_size = config["val_batch_size"],
file_count = config["val_file_size"],
to_standardize= True,
include_y_local= False,
labels_list = ['x-midplane','y-midplane','cotAlpha','cotBeta'],
input_shape = (2,13,21), # (20,13,21),
transpose = (0,2,3,1),
files_from_end=True,
use_time_stamps = [0,19],
tfrecords_dir = tfrecords_dir,
)
# build model, load weights, predict
model=CreateModel((13,21,2), n_filters=config["n_filters"], pool_size=config["pool_size"])
model.load_weights(config["weightsPath"])
p_test = model.predict(test_generator)
complete_truth = None
for _, y in test_generator:
if complete_truth is None:
complete_truth = y
else:
complete_truth = np.concatenate((complete_truth, y), axis=0)
# creates df with all predicted values and matrix elements - 4 predictions, all 10 unique matrix elements
df = pd.DataFrame(p_test,columns=['x','M11','y','M22','cotA','M33','cotB','M44','M21','M31','M32','M41','M42','M43'])
# stores all true values in same matrix as xtrue, ytrue, etc.
df['xtrue'] = complete_truth[:,0]
df['ytrue'] = complete_truth[:,1]
df['cotAtrue'] = complete_truth[:,2]
df['cotBtrue'] = complete_truth[:,3]
df['M11'] = minval+tf.math.maximum(df['M11'], 0)
df['M22'] = minval+tf.math.maximum(df['M22'], 0)
df['M33'] = minval+tf.math.maximum(df['M33'], 0)
df['M44'] = minval+tf.math.maximum(df['M44'], 0)
# calculates residuals for x, y, cotA, cotB
residuals = df['xtrue'] - df['x']
residualsy = df['ytrue'] - df['y']
residualsA = df['cotAtrue'] - df['cotA']
residualsB = df['cotBtrue'] - df['cotB']
# stores results as csv
df.to_csv(config["outFileName"], header=True, index=False)
# clean up tf records
shutil.rmtree(tfrecords_dir)
if __name__ == "__main__":
# set up command line arguments
parser = argparse.ArgumentParser()
parser.add_argument("-i", "--inFolder", help="path to the training results", default=None)
parser.add_argument("--query", help="path to json file containing query", default=None)
parser.add_argument("--njobs", help="number of jobs to actually launch. default is all", default=-1, type=int)
parser.add_argument("--doOverwrite", help="overwrite any existing evaluation files", action="store_true")
args = parser.parse_args()
# read in query
if Path(args.query).resolve().exists():
query_path = Path(args.query).resolve()
else:
# throw
raise ValueError(f"Could not locate {args.query} in query directory or as absolute path")
with open(query_path) as f:
query = json.load(f)
# data paths and configs
data_directory_path = "/net/projects/particlelab/smartpix/dataset8/unflipped/" # "/net/scratch/badea/dataset8/unflipped/"
labels_directory_path = "/net/projects/particlelab/smartpix/dataset8/unflipped/" # "/net/scratch/badea/dataset8/unflipped/"
val_batch_size = 5000
val_file_size = 16
# figure out which weight to use
inFolder = Path(args.inFolder).resolve()
weightsFolders = list(inFolder.glob("*/weights*"))
top_dir = Path(inFolder, "eval", f'{"%08x" % random.randrange(16**8)}', "%j").resolve()
# weightsFolder = Path("/home/badea/smartpix/semiparametric/timeslices-2/neurips-3x3-2conv/results/training-a9a85a9d/16367_0/weights-nFilters1-poolSize1-checkpoints").resolve()
# configurations
confs = []
for weightsFolder in weightsFolders:
n_filters = int(weightsFolder.parts[-1].split("-")[1].split("nFilters")[1])
pool_size = int(weightsFolder.parts[-1].split("-")[2].split("poolSize")[1])
# files = os.listdir(weightsFolder)
files = [str(f) for f in weightsFolder.glob("*.hdf5")]
vlosses = [float(f.split("-v")[1].split(".hdf5")[0]) for f in files]
bestfile = files[np.argmin(vlosses)]
weightsPath = Path(weightsFolder, bestfile).resolve()
# outFileName = Path(str(weightsPath).replace(".hdf5", "_eval.csv")).resolve()
outFileName = Path(top_dir, weightsPath.parts[-3], weightsPath.parts[-1].replace(".hdf5", "_eval.csv")).resolve()
if outFileName.exists() and not args.doOverwrite:
print(f"Warning: {outFileName} exists. If you want to overwrite it pass in --doOverwrite.")
continue
confs.append({
"weightsPath" : weightsPath,
"outFileName" : outFileName,
"data_directory_path" : data_directory_path,
"labels_directory_path" : labels_directory_path,
"n_filters" : n_filters,
"pool_size" : pool_size,
"val_batch_size" : val_batch_size,
"val_file_size" : val_file_size
})
# if submitit false then just launch job
if not query.get("submitit", False):
for iC, conf in enumerate(confs):
# only launch a single job
if args.njobs != -1 and (iC+1) > args.njobs:
continue
print(conf)
evaluate(conf)
else:
# submission
executor = submitit.AutoExecutor(folder=top_dir)
executor.update_parameters(**query.get("slurm", {}))
# the following line tells the scheduler to only run at most 2 jobs at once. By default, this is several hundreds
# executor.update_parameters(slurm_array_parallelism=2)
# loop over configurations
jobs = []
with executor.batch():
for iC, conf in enumerate(confs):
# only launch a single job
if args.njobs != -1 and (iC+1) > args.njobs:
continue
print(conf)
job = executor.submit(evaluate, conf)
jobs.append(job)