def main():
if len(sys.argv) != 3:
print "Usage: python run.py order"
sys.exit(2)
fcwd = dirname(realpath(__file__))
order = sys.argv[1]
paramfile = sys.argv[2]
with open(paramfile, "r") as f:
paramdata = f.readlines()
params = {}
params["order"] = order
params["predict_filelist"] = "data/test.txt"
for i in range(len(paramdata)):
line = paramdata[i].strip().split()
params.update({line[0]: line[1]})
print "####### Parameters Input #######"
for key in params.keys():
print key + ": " + params[key]
print "################################"
for code in [
"output_topdir",
"gpunum",
"caffemodel_topdir",
"model_name",
"batch_name",
"batch2predict",
"optimwrt",
"outputlayer",
]:
assert code in params.keys()
ctime = strftime("%Y-%m-%d-%H-%M-%S", localtime())
datadir = abspath(join(params["output_topdir"], "data"))
params["solver_file"] = join(params["caffemodel_topdir"], "solver.prototxt")
params["trainval_file"] = join(params["caffemodel_topdir"], "trainval.prototxt")
params["deploy_file"] = join(params["caffemodel_topdir"], "deploy.prototxt")
flag = False
if params["order"] == "train":
if not "batch_name" in params.keys():
params["batch_name"] = ctime
modeltopdir = abspath(join(params["output_topdir"], params["model_name"], params["batch_name"]))
if exists(modeltopdir):
print "model folder exists, will remove"
system("rm -r " + modeltopdir)
makedirs(modeltopdir)
cwd = getcwd()
for trial in range(int(params["trial_num"])):
modeldir = join(modeltopdir, "trial" + str(trial))
makedirs(modeldir)
system(" ".join(["cp ", params["solver_file"], modeldir]))
system(" ".join(["cp ", params["trainval_file"], modeldir]))
chdir(modeldir)
outlog = join(modeldir, "train.out")
errlog = join(modeldir, "train.err")
system(
" ".join(
[
"python",
join(fcwd, "train.py"),
basename(params["solver_file"]),
modeldir,
params["gpunum"],
"1 >",
outlog,
"2>",
errlog,
]
)
)
chdir(cwd)
### Get the best performing iteration among all the trials
best_trial, best_iter = getBestRunAll(modeltopdir, int(params["trial_num"]), "train.err", params["optimwrt"])
model_file = join(modeltopdir, "trial" + str(best_trial), "train_iter_" + best_iter + ".caffemodel")
testdir = join(modeltopdir, "best_trial")
if exists(testdir):
print "testdir " + testdir + " exists, will be removed"
system("rm -r " + testdir)
makedirs(testdir)
system(" ".join(["cp", model_file, join(testdir, "bestiter.caffemodel")]))
system(" ".join(["cp", params["deploy_file"], testdir]))
with open(join(testdir, "bestiter.info"), "w") as f:
f.write("best_trial\tbest_iter\n")
f.write("%d\t%s\n" % (best_trial, best_iter))
flag = True
if params["order"] == "test":
modeltopdir = abspath(join(params["output_topdir"], params["model_name"], params["batch2predict"]))
testdir = join(modeltopdir, "best_trial")
test(
params["deploy_file"],
join(testdir, "bestiter.caffemodel"),
join(params["output_topdir"], params["predict_filelist"]),
int(params["gpunum"]),
testdir,
params["outputlayer"],
)
flag = True
if params["order"] == "test_eval":
pred_topdir = abspath(
join(params["output_topdir"], params["model_name"], params["batch2predict"], "best_trial")
)
pred_f = join(pred_topdir, "bestiter.pred")
real_f = join(params["output_topdir"], params["predict_filelist"])
outfile = join(pred_topdir, "bestiter.pred.eval")
test_eval(pred_f, real_f, outfile)
flag = True
if params["order"] == "pred":
for code in ["deploy2predictW", "caffemodel2predictW", "data2predict", "predict_outdir"]:
assert code in params.keys()
test(
params["deploy2predictW"],
params["caffemodel2predictW"],
params["data2predict"],
int(params["gpunum"]),
params["predict_outdir"],
params["outputlayer"],
)
flag = True
if not flag:
print "Cannot recognize order; Exit"
exit(2)
def main():
if len(sys.argv) != 3:
print 'Usage: python run.py order'
sys.exit(2)
fcwd = dirname(realpath(__file__))
order = sys.argv[1]
paramfile = sys.argv[2]
with open(paramfile, 'r') as f:
paramdata = f.readlines()
params = {}
params['order'] = order
params['predict_filelist'] = 'data/test.txt'
for i in range(len(paramdata)):
line = paramdata[i].strip().split()
params.update({line[0]: line[1]})
print '####### Parameters Input #######'
for key in params.keys():
print key + ': ' + params[key]
print '################################'
for code in [
'output_topdir', 'gpunum', 'caffemodel_topdir', 'model_name',
'batch_name', 'batch2predict', 'optimwrt', 'outputlayer'
]:
assert (code in params.keys())
ctime = strftime("%Y-%m-%d-%H-%M-%S", localtime())
datadir = abspath(join(params['output_topdir'], 'data'))
params['solver_file'] = join(params['caffemodel_topdir'],
'solver.prototxt')
params['trainval_file'] = join(params['caffemodel_topdir'],
'trainval.prototxt')
params['deploy_file'] = join(params['caffemodel_topdir'],
'deploy.prototxt')
flag = False
if params['order'] == 'train':
if not 'batch_name' in params.keys():
params['batch_name'] = ctime
modeltopdir = abspath(
join(params['output_topdir'], params['model_name'],
params['batch_name']))
if exists(modeltopdir):
print 'model folder exists, will remove'
system('rm -r ' + modeltopdir)
makedirs(modeltopdir)
cwd = getcwd()
for trial in range(int(params['trial_num'])):
modeldir = join(modeltopdir, 'trial' + str(trial))
makedirs(modeldir)
system(' '.join(['cp ', params['solver_file'], modeldir]))
system(' '.join(['cp ', params['trainval_file'], modeldir]))
chdir(modeldir)
outlog = join(modeldir, 'train.out')
errlog = join(modeldir, 'train.err')
system(' '.join([
'python',
join(fcwd, 'train.py'),
basename(params['solver_file']), modeldir, params['gpunum'],
'1 >', outlog, '2>', errlog
]))
chdir(cwd)
### Get the best performing iteration among all the trials
best_trial, best_iter = getBestRunAll(modeltopdir,
int(params['trial_num']),
'train.err', params['optimwrt'])
model_file = join(modeltopdir, 'trial' + str(best_trial),
'train_iter_' + best_iter + '.caffemodel')
testdir = join(modeltopdir, 'best_trial')
if exists(testdir):
print 'testdir ' + testdir + ' exists, will be removed'
system('rm -r ' + testdir)
makedirs(testdir)
system(' '.join(
['cp', model_file,
join(testdir, 'bestiter.caffemodel')]))
system(' '.join(['cp', params['deploy_file'], testdir]))
with open(join(testdir, 'bestiter.info'), 'w') as f:
f.write('best_trial\tbest_iter\n')
f.write('%d\t%s\n' % (best_trial, best_iter))
flag = True
if params['order'] == 'test':
modeltopdir = abspath(
join(params['output_topdir'], params['model_name'],
params['batch2predict']))
testdir = join(modeltopdir, 'best_trial')
test(params['deploy_file'], join(testdir, 'bestiter.caffemodel'),
join(params['output_topdir'], params['predict_filelist']),
int(params['gpunum']), testdir, params['outputlayer'])
flag = True
if params['order'] == 'test_eval':
pred_topdir = abspath(
join(params['output_topdir'], params['model_name'],
params['batch2predict'], 'best_trial'))
pred_f = join(pred_topdir, 'bestiter.pred')
real_f = join(params['output_topdir'], params['predict_filelist'])
outfile = join(pred_topdir, 'bestiter.pred.eval')
test_eval(pred_f, real_f, outfile)
flag = True
if params['order'] == 'pred':
for code in [
'deploy2predictW', 'caffemodel2predictW', 'data2predict',
'predict_outdir'
]:
assert (code in params.keys())
test(params['deploy2predictW'],
params['caffemodel2predictW'], params['data2predict'],
int(params['gpunum']), params['predict_outdir'],
params['outputlayer'])
flag = True
if not flag:
print 'Cannot recognize order; Exit'
exit(2)
def main():
if len(sys.argv)!=3:
print 'Usage: python run.py order'
sys.exit(2)
order = sys.argv[1]
paramfile = sys.argv[2]
with open(paramfile,'r') as f:
paramdata = f.readlines()
params = {}
params['order'] = order
for i in range(len(paramdata)):
line = paramdata[i].strip().split()
params.update({line[0]:line[1]})
print '####### Parameters Input #######'
for key in params.keys():
print key + ': ' + params[key]
print '################################'
ctime = strftime("%Y-%m-%d-%H-%M-%S", localtime())
datadir = os.path.abspath(os.path.join(params['model_topdir'],'data'))
flag = False;
if params['order']=='getdata':
if os.path.exists(datadir):
print 'data folder exists, will remove'
os.system('rm -r '+datadir)
os.makedirs(datadir)
cmd = ' '.join(['scp -r ',params['data_src']+'/*',datadir])
os.system(cmd)
flag = True
if params['order']=='train':
if not 'modelname' in params.keys():
params['model_batchname'] = ctime
modeltopdir = abspath(join(params['model_topdir'],params['modelname'],params['model_batchname']))
if os.path.exists(modeltopdir):
print 'model folder exists, will remove'
os.system('rm -r ' + modeltopdir)
os.makedirs(modeltopdir)
cwd = os.getcwd()
for trial in range(int(params['trial_num'])):
modeldir = join(modeltopdir,'trial'+str(trial))
makedirs(modeldir)
cmd = ' '.join(['cp ',params['solver_file'], modeldir])
os.system(cmd)
cmd = ' '.join(['cp ',params['trainval_file'], modeldir])
os.system(cmd)
os.chdir(modeldir)
#train(os.path.basename(params['solver_file']),modeldir)
outlog = os.path.join(modeldir,'train.out')
errlog = os.path.join(modeldir,'train.err')
os.system(' '.join(['python',os.path.join(params['codedir'],'train.py'),os.path.basename(params['solver_file']),modeldir,params['gpunum'],'1 >',outlog,'2>',errlog]))
os.chdir(cwd)
flag = True
if params['order']=='test':
modeltopdir = os.path.abspath(os.path.join(params['model_topdir'],params['modelname'],params['predictmodel_batch']))
testdir = join(modeltopdir,'best_trial')
if exists(testdir):
print 'testdir '+testdir+' exists, will be removed'
system('rm -r ' + testdir)
makedirs(testdir)
cmd = ' '.join(['cp',params['deploy_file'], testdir])
os.system(cmd)
os.chdir(testdir)
test(os.path.basename(params['deploy_file']),modeltopdir,os.path.join(params['model_topdir'],params['predict_filelist']),int(params['gpunum']),int(params['trial_num']),testdir,params['optimwrt'])
flag = True
if params['order']=='test_eval':
modeldir = os.path.abspath(os.path.join(params['model_topdir'],params['modelname'],params['predictmodel_batch'],'output'))
pred_topdir = abspath(join(params['model_topdir'],params['modelname'],params['predictmodel_batch'],'best_trial'))
pred_f = join(pred_topdir,'bestiter.pred')
real_f = join(params['model_topdir'],params['predict_filelist'])
outfile = join(pred_topdir,'bestiter.pred.eval')
test_eval(pred_f,real_f,outfile)
flag = True
if not flag:
print 'Cannot recognize order; Exit'
exit(2)
