def cmd_setup_slave(args):
''' Setup proper slave environment and connect to master server '''
if not args.export:
cmd = ["mkt", "setup", "--no-dirs"]
if args.yes:
cmd.append("-y")
if args.force:
cmd.append("-f")
subprocess.call(cmd)
utils.init_config_file()
section = utils.load_config_file()
subprocess.call(["sudo", "mkdir", "-p", section['src']])
subprocess.call(["sudo", "chown", "-R", utils.username() + ":" + utils.group(), section['src']])
subprocess.call(["sudo", "mkdir", "-p", section['logs']])
subprocess.call(["sudo", "chown", "-R", utils.username() + ":" + utils.group(), section['logs']])
with tempfile.NamedTemporaryFile("w") as f:
f.write(open("/etc/fstab").read())
export = args.hostname[0] + ":" + \
section['src'][:-1] + "\t" + section['src'][:-1] + \
"\t" + "nfs" + "\t" + "ro,nolock 0 0" + "\n" + \
args.hostname[0] + ":" + \
section['logs'][:-1] + "\t" + section['logs'][:-1] + \
"\t" + "nfs" + "\t" + "rw,nolock 0 0" + "\n"
f.write(export)
f.flush()
subprocess.call(["sudo", "cp", f.name, "/etc/fstab"])
subprocess.call(["sudo", "mount", "-a", "-t", "nfs", "-o", "remount"])
def args_images(parser):
parser.add_argument(
"--no-pull",
dest="pull",
action="store_false",
help=
"Do not update the base docker images from the public docker registry",
default=True)
parser.add_argument(
"--only",
action="store",
help="Build only the image with the given docker image tag",
default=None)
parser.add_argument(
"--push",
action="store_true",
help=
"Upload created images to docker registry (need to have an account in harbor registry)",
default=False)
section = utils.load_config_file()
parser.add_argument("os",
nargs='?',
help="The image to build",
choices=sorted(supported_os),
default=section.get('os', default_os))
def cmd_build(args):
"""Smart build."""
from . import cmd_images
section = utils.load_config_file()
if not args.project:
set_args_project(args, section)
if args.project != 'custom' and args.with_kernel_headers:
exit("--with-kernel-headers is applicable for \"custom\" target only.")
build = Build(args.project)
recipe_dir = None
if args.build_recipe is not None:
args.build_recipe = os.path.realpath(args.build_recipe)
recipe_dir = os.path.dirname(args.build_recipe)
build.pickle['shell'] = args.run_shell
build.pickle["passwd"] = "%s:x:%s:%s:%s:%s:/bin/bash" % (
username(), os.getuid(), os.getgid(), username(), os.getenv("HOME"))
build.pickle["group"] = "%s:x:%s:" % (group(), os.getgid())
build.pickle["uid"] = int(os.getuid())
build.pickle["gid"] = int(os.getgid())
build.pickle["home"] = os.getenv("HOME")
build.pickle['clean'] = args.clean
build.pickle['build_recipe'] = args.build_recipe
if args.with_kernel_headers:
build.pickle['kernel'] = section.get('kernel', None)
do_cmd = ["python3", "/plugins/do-build.py"]
docker_exec(["run"] +
build.run_build_cmd(cmd_images.default_os, recipe_dir) +
do_cmd)
def cmd_setup_master(args):
''' Setup proper master environment and export it to slaves '''
if not args.export:
cmd = ["mkt", "setup"]
if args.yes:
cmd.append("-y")
if args.force:
cmd.append("-f")
subprocess.call(cmd)
utils.init_config_file()
section = utils.load_config_file()
with tempfile.NamedTemporaryFile("w") as f:
f.write(open("/etc/exports").read())
for host in args.hostnames:
export = section['src'] + "\t" + \
str(host) + "(ro,async,no_subtree_check,no_root_squash)" + \
"\n" + section['logs'] + "\t" + \
str(host) + "(rw,async,no_subtree_check,no_root_squash)" + "\n"
f.write(export)
f.flush()
subprocess.call(["sudo", "cp", f.name, "/etc/exports"])
subprocess.call(["sudo", "exportfs", "-r"])
def args_run(parser):
section = utils.load_config_file()
parser.add_argument("image",
nargs='?',
choices=sorted(utils.get_images()),
help="The IB card configuration to use")
kernel = parser.add_mutually_exclusive_group()
kernel.add_argument(
'--kernel',
help="Path to the top of a compiled kernel source tree to boot",
default=None)
kernel.add_argument('--kernel-rpm',
help="Path to a kernel RPM to boot",
default=None)
parser.add_argument('--dir',
action="append",
help="Other paths to map",
default=[])
parser.add_argument('--simx',
metavar='SIMX_DEV',
action="append",
default=[],
choices=sorted(get_simx_rdma_devices()),
help="Run using simx to create a mlx5 IB device")
parser.add_argument(
'--run-shell',
action="store_true",
default=False,
help="Run a shell inside the container instead of invoking kvm")
parser.add_argument(
'--pci',
metavar="PCI_BDF",
action="append",
default=[],
choices=sorted(get_pci_rdma_devices().keys()),
help="Pass a given PCI bus/device/function to the guest")
parser.add_argument(
'--virt',
metavar="VIRT_DEV",
action="append",
default=[],
choices=sorted(get_virt_rdma_devices()),
help="Pass a virtual device type-interface format to the guest")
parser.add_argument(
'--boot-script',
help="Path to the custom boot script which will be executed after boot",
default=None)
parser.add_argument('--gdbserver',
metavar='PORT',
type=int,
help="TCP port for QEMU's GDB server",
default=None)
parser.add_argument('--nested_pci',
metavar='NESTED_PCI',
action="append",
default=[],
help="Provide PCI list for the nested VM")
def test_load_config_file_error(monkeypatch):
def mock_get_config_file_name():
return 'invalid_path'
monkeypatch.setattr(utils, 'get_config_file_name',
mock_get_config_file_name)
with pytest.raises(FileNotFoundError):
dictionary = utils.load_config_file()
def build_content(base_path: str) -> List[Post]:
"""
Convert all the markdown files to html files.
Returns a list of:
- destination file path
- html transformed text
- meta information
"""
# Load markdown with the meta extension
configuration = load_config_file(base_path)
html_path = os.path.join(base_path, required_folders["public"])
result = []
# iterate over pairs file name and file path
for filen, file_path in list_content(base_path,
required_folders["content"]):
# Open the file
with open(file_path, "r") as f:
# load markdown
md = markdown.Markdown(
extensions=['meta', 'tables', 'sane_lists', 'attr_list'])
# Read document
data = f.read()
# Convert markdown to html
html = md.convert(data)
# Get file extension
filenn, extension = os.path.splitext(filen)
# If it's not md skip file
if extension != '.md':
continue
if 'index' in filen:
result.append(
Post(path=os.path.join(html_path, "index.html"),
html=html,
meta=md.Meta,
config=configuration,
index=True))
else:
if 'slug' in md.Meta.keys():
slug = md.Meta['slug'][0]
else:
slug = filen.split(".")[0]
md.Meta['slug'] = slug
result.append(
Post(path=os.path.join(html_path, slug, "index.html"),
html=html,
meta=md.Meta,
config=configuration))
return result
def get_params(file):
params = utils.load_config_file(file)
# Convert from JSON format to DataLoader format
params["training_dataset_paths"] = utils.format_dataset_path(
params["training_dataset_paths"])
params["test_dataset_paths"] = utils.format_dataset_path(
params["test_dataset_paths"])
return params
def main():
parser = create_parser(True)
args = parser.parse_args()
spec = load_config_file(args.yaml) # spec is a dict of dicts of dicts
trainer = Trainer(args, args.yaml, add_timestamp=True)
xval_merge = XvalMerge(args, spec["data"], trainer)
data_pair, val_results = run_on_split(args, split=None, trainer=trainer)
xval_merge.add(1, data_pair, val_results)
xval_merge.finalize()
xval_merge.save(xval_merge.trainer.tb_log_dir)
def main(args):
print("Loading config file: ", args.config)
params = utils.load_config_file(args.config)
params["dataset_paths"] = utils.format_dataset_path(
params["dataset_paths"])
if "nyu" not in params:
params["nyu"] = False
# Data loading code
print("Creating data loaders...")
if params["nyu"]:
from dataloaders.nyu import NYUDataset
val_dataset = NYUDataset(params["dataset_paths"], split='val')
else:
val_dataset = Datasets.FastDepthDataset(params["dataset_paths"],
split='val',
depth_min=params["depth_min"],
depth_max=params["depth_max"],
input_shape_model=(224, 224),
random_crop=False)
# set batch size to be 1 for validation
data_loader = torch.utils.data.DataLoader(
val_dataset,
batch_size=1,
shuffle=False,
num_workers=params["num_workers"],
pin_memory=True)
# Set GPU
params["device"] = torch.device(
"cuda:{}".format(params["device"])
if params["device"] >= 0 and torch.cuda.is_available() else "cpu")
print("Using device", params["device"])
print("Loading models...")
models = []
model_names = []
for model_dict in params["models"]:
model_names.append(Path(model_dict["model_path"]).stem)
model, _ = utils.load_model(model_dict, model_dict["model_path"],
params["device"])
model.to(params["device"])
models.append(model)
# Create output directory
output_directory = os.path.join(params["save_folder"],
".".join(model_names))
if not os.path.exists(output_directory):
os.makedirs(output_directory)
params["output_directory"] = output_directory
print("Saving results to " + output_directory)
compare_models(params, data_loader, models)
def _get_blast_version(self,command):
configs = utils.load_config_file()
p = subprocess.Popen([configs[command], "-h"], stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
output, err = p.communicate(b"input data that is passed to subprocess' stdin")
rc = p.returncode
output = output.split()
for i in range(0,len(output)):
if "version" == output[i] and command=="makeblastdb":
return ("BLAST-" + output[i+1])
elif "version" in output[i] and command == "blastp":
return ("BLAST-" + output[i+4])
return ("BLAST version not found\n")
def __init__(self,
prep_id,
scan_id,
hmm_db,
out_dir = ".",
cpu = 2):
self.args = {"prep_dir" : os.path.join(os.path.abspath(out_dir),prep_id + "_PREPARE") ,
"scan_dir" : os.path.join(os.path.abspath(out_dir),scan_id + "_SCAN") ,
"cpu" : cpu,
"hmm_db" : hmm_db,
"configs" : utils.load_config_file() ,
"prep_id": prep_id,
"scan_id": scan_id}
def test_load_config(tmpdir, monkeypatch):
yaml_string = """
a: 1
b: 2
"""
yaml_file = tmpdir.mkdir("sub").join("test.yaml")
yaml_file.write(yaml_string)
def mock_get_config_file_name():
return yaml_file.strpath
monkeypatch.setattr(utils, 'get_config_file_name',
mock_get_config_file_name)
dictionary = utils.load_config_file()
assert dictionary == {'a': 1, 'b': 2}
def get_configs(cfg_type, **kwargs):
cfg_file = load_config_file()
config = None
if kwargs.get('cfg_name') is not None:
result = list(
filter(lambda x: x.get(kwargs.get('cfg_name')) is not None,
cfg_file.get(cfg_type)))
if result:
config = result[0][kwargs.get('cfg_name')]
else:
config = cfg_file.get(cfg_type)
return config
def main():
parser = create_parser(True)
args = parser.parse_args()
spec = utils.load_config_file(
args.yaml) # spec is a dict of dicts of dicts
data_settings = procdata.apply_defaults(spec["data"])
para_settings = utils.apply_defaults(spec["params"])
xval_merge = XvalMerge(args, data_settings)
data_pair, val_results = run_on_split(args,
data_settings,
para_settings,
split=None,
trainer=xval_merge.trainer)
xval_merge.add(1, data_pair, val_results)
xval_merge.finalize()
xval_merge.save()
def cmd_ci(args):
"""Local continuous integration check."""
from . import cmd_images
section = utils.load_config_file()
if not args.project:
set_args_project(args, section)
build = Build(args.project)
build.pickle['checkpatch'] = args.checkpatch
build.pickle['sparse'] = args.sparse
build.pickle['gerrit'] = args.gerrit
build.pickle['show_all'] = args.show_all
build.pickle["warnings"] = args.warnings
build.pickle["smatch"] = args.smatch
build.pickle["clang"] = args.clang
# FIXME: allow git revisions as input to --rev.
# But for now, let's give an option to provide
# commit.
build.pickle['rev'] = args.rev[0]
do_cmd = ["python3", "/plugins/do-ci.py"]
docker_exec(["run"] + build.run_ci_cmd(cmd_images.default_os) + do_cmd)
def watch() -> None:
"""
Start a developing server for your content
"""
# Get where the execution is being made
base_path = os.getcwd()
theme = load_config_file(base_path)['theme']
content_folder = os.path.join(base_path, 'content')
theme_folder = os.path.join(base_path, 'themes')
# Initialize the dev server
server = Server()
# Build content
content = build_content(base_path)
render_content(base_path, content, theme)
copy_static_assets(base_path, theme)
server.watch(content_folder, shell("arcade build", cwd=base_path))
server.watch(theme_folder, shell("arcade build", cwd=base_path))
server.serve(root="public")
def run_blast(out_dir, file_type, evalue):
configs = utils.load_config_file()
command = map(str, [
configs["makeblastdb"], "-in", out_dir + "/" + file_type + ".fasta",
"-dbtype", "prot"
])
command2 = map(str, [
configs["blastp"], "-db", out_dir + "/" + file_type + ".fasta",
"-query", out_dir + "/" + file_type + ".fasta", "-out", out_dir + "/" +
file_type + "_blast_results", "-outfmt", "6", "-evalue", evalue
])
attempt = 0
res = 1
while attempt < utils.MAX_ATTEMPTS and res != 0:
res = subprocess.call(command)
if res != 0:
attempt += 1
continue
res = subprocess.call(command2)
attempt += 1
if res != 0:
sys.exit("Error: Failed to run BLASTP for [" + file_type +
"]. Please check log files.")
def build() -> None:
base_path = os.getcwd()
theme = load_config_file(base_path)['theme']
content = build_content(base_path)
render_content(base_path, content, theme)
copy_static_assets(base_path, theme)
def cmd_run(args):
"""Run a system image container inside KVM"""
from . import cmd_images
section = utils.load_config_file()
docker_os = section.get('os', cmd_images.default_os)
# We have three possible options to execute:
# 1. "mkt run" without request to specific image. We will try to find
# default one
# 2. "mkt run --pci ..." or "mkt run --simx ...". We won't use default
# image but add supplied PCIs and SimX devices.
# 3. "mkt run image_name --pci ..." or "mkt run image_name --simx ...". We
# will add those PCIs and SimX devices to the container.
s = set()
if not args.pci and not args.simx:
if not args.image:
args.image = section.get('image', None)
if args.image:
pci = utils.get_images(args.image)['pci']
s = pci.split()
union = set(get_simx_rdma_devices()).union(
set(get_pci_rdma_devices().keys())).union(set(get_virt_rdma_devices()))
if not set(s).issubset(union):
# It is possible only for config files, because we sanitized
# input to ensure that valid data is supplied.
exit(
"There is an error in configuration file, PCI, SIMX or VIRT devices don't exists."
)
args.pci += set(s).intersection(set(get_pci_rdma_devices().keys()))
args.virt += set(s).intersection(set(get_virt_rdma_devices()))
b = args.pci + args.virt
args.simx += [item for item in s if item not in b]
if len(args.simx) > 5:
exit("SimX doesn't support more than 5 devices")
if not args.kernel and not args.kernel_rpm:
exit(
"Must specify a linux kernel with --kernel, or a config file default"
)
# Invoke ourself as root to manipulate sysfs
if args.pci:
subprocess.check_call([
"sudo", sys.executable,
os.path.join(os.path.dirname(__file__), "../utils/vfio.py")
] + ["--pci=%s" % (I) for I in args.pci])
mapdirs = DirList()
if args.kernel_rpm is not None:
args.kernel_rpm = os.path.realpath(args.kernel_rpm)
if not os.path.isfile(args.kernel_rpm):
raise ValueError("Kernel RPM %r does not exist" %
(args.kernel_rpm))
mapdirs.add(os.path.dirname(args.kernel_rpm))
args.kernel = None
else:
args.kernel = os.path.realpath(args.kernel)
if not os.path.isdir(args.kernel):
raise ValueError(
"Kernel path %r is not a directory/does not exist" %
(args.kernel))
mapdirs.add(args.kernel)
if args.image:
try:
if utils.get_images(args.image)['custom_qemu'] != "true":
raise KeyError
args.custom_qemu = section.get('simx', None)
except KeyError:
args.custom_qemu = None
else:
args.custom_qemu = None
if args.custom_qemu:
args.custom_qemu = os.path.realpath(args.custom_qemu)
if not os.path.isdir(args.custom_qemu):
raise ValueError("SimX path %r is not a directory/does not exist" %
(args.custom_qemu))
mapdirs.add(args.custom_qemu)
usr = pwd.getpwuid(os.getuid())
args.dir.append(usr.pw_dir)
if 'dir' in section:
args.dir += section['dir'].split()
args.dir = list(set(args.dir))
for I in args.dir:
mapdirs.add(I)
args.boot_script = validate_and_set_boot(args)
if args.boot_script:
mapdirs.add(os.path.dirname(args.boot_script))
vm_addr = get_mac()
if args.run_shell:
do_kvm_args = ["/bin/bash"]
else:
do_kvm_args = ["python3", "/plugins/do-kvm.py"]
if vm_addr.ip:
# Open network for QEMU, relevant for bridged mode only
iprule = [
"FORWARD", "-m", "physdev", "--physdev-is-bridged", "-j",
"ACCEPT"
]
# First delete old rule
subprocess.call(["sudo", "iptables", "-D"] + iprule,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL)
subprocess.call(["sudo", "iptables", "-I"] + iprule,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL)
src_dir = os.path.dirname(
os.path.abspath(inspect.getfile(inspect.currentframe())))
ssh = False
if args.pci:
# chack if we have container running with bound PCI device to it
# sudo docker ps --filter "label=pci" --format "{{.Names}}"
# sudo docker inspect --format='{{.Config.Labels.pci}}' mkt_run_nps-server-14-015
cont = docker_get_containers(label="pci")
for c in cont:
c = c.decode()[1:-1]
cpci = docker_output(
["inspect", "--format", '"{{.Config.Labels.pci}}"', c])
cpci = cpci.decode()[2:-2].split(', ')
cpci = [x[1:-1] for x in cpci]
common = set(cpci).intersection(set(args.pci))
if common:
ssh = True
cname = c
if ssh:
subprocess.call(["ssh", "root@%s" % (get_host_name(cname))])
else:
cname = get_container_name(vm_addr)
docker_exec(["run"] + mapdirs.as_docker_bind() + [
"-v",
"%s:/plugins:ro" % (src_dir),
"--mount",
"type=bind,source=%s,destination=/logs" %
(utils.config.runtime_logs_dir),
"--rm",
"--net=host",
"--privileged",
"--name=%s" % (cname),
"--tty",
"-l",
"pci=%s" % (args.pci),
"--hostname",
vm_addr.hostname,
"-e",
"KVM_PICKLE=%s" % (get_pickle(args, vm_addr)),
"--interactive",
make_image_name("kvm", docker_os),
] + do_kvm_args)
choices=('annotation', 'train', 'evaluate',
'generate'),
default='annotation')
return parser.parse_args()
if __name__ == '__main__':
args = parse_args()
seed = 0
np.random.seed(seed)
mx.random.seed(seed)
cfg = load_config_file('config.yml')
root_dir = cfg['BASE_DIR']
gan = cfg['GAN']
gan_dir = cfg['GAN_DIR']
gan_gpu_ids = cfg['GAN_GPU_IDS']
gan_batch_size = cfg['GAN_BATCH_SIZE_PER_GPU']
solver_gpu_ids = cfg['SOLVER_GPU_IDS']
annotation = cfg['ANNOTATION']
no_gan = cfg.get('NO_GAN', False)
imgs_dir = cfg.get('IMGS_DIR', None)
n_generate = cfg.get('GENERATE_NUM', 10000)
if args.action == 'annotation':
import tkinter as tk
root = tk.Tk()
if annotation == 'segmentation':
def cmd_setup(args):
"""Setup environment."""
if not args.dirs:
args.kernel = False
args.rdma_core = False
args.iproute = False
if args.installs:
print(""" This setup script will update your hypervisor to latest
distribution packages and install docker. Please restart
the hypervisor to complete the installation. """)
if args.yes is False and utils.query_yes_no("Do you want to proceed?",
'no') is False:
exit("Exiting ...")
supported_os = {
'fedora': '26',
'ubuntu': '16',
'rhel': '8',
'redhat': '8',
}
# Python API stability is dissaster
# module platform was deprecated https://docs.python.org/3/library/platform.html
# Luckily enough, they added distro module before removing platform
try:
import distro
distro_id = distro.id()
distro_v = distro.major_version()
except ModuleNotFoundError:
import platform
distro_id = platform.dist()[0].lower()
distro_v = platform.dist()[1].split('.')[0]
if distro_id not in supported_os.keys(
) or distro_v < supported_os[distro_id]:
exit(""" Your hypervisor is not supported. Exiting ...""")
if args.installs:
setuphv = utils.get_internal_fn('scripts/')
if distro_id == 'redhat':
distro_id = 'rhel'
setuphv += 'setup-hv.' + distro_id
subprocess.check_call(setuphv)
utils.init_config_file()
section = utils.load_config_file()
if args.dirs:
for key, value in section.items():
if args.force:
subprocess.call(["sudo", "rm", "-rf", value])
if os.path.exists(value):
exit("Please remove " + value + " Exiting ...")
if key == "kernel" and not args.kernel:
continue
if key == "rdma-core" and not args.rdma_core:
continue
if key == "iproute2" and not args.iproute:
continue
if key == 'os':
continue
print("Prepare " + key)
subprocess.call(["sudo", "mkdir", "-p", value])
subprocess.call([
"sudo", "chown", "-R",
utils.username() + ":" + utils.group(), value
])
if key == "src" or key == "logs" or key == "ccache":
continue
if key == "kernel":
key = "linux"
p = subprocess.Popen([
"git", "clone", "ssh://" + utils.username() +
"@l-gerrit.mtl.labs.mlnx:29418/upstream/" + key, "."
],
cwd=value)
p.wait()
p = subprocess.Popen([
"scp", "-p", "-P", "29418",
utils.username() + "@l-gerrit.mtl.labs.mlnx:hooks/commit-msg",
".git/hooks/"
],
cwd=value)
p.wait()
if key == "linux":
shutil.copy(
os.path.join(os.path.dirname(__file__),
"../configs/kconfig-kvm"), value + "/.config")
print("Completed, PLEASE RESTART server")
def main(args):
print("Loading config file: ", args.config)
params = utils.load_config_file(args.config)
params["test_dataset_paths"] = utils.format_dataset_path(
params["test_dataset_paths"])
if args.existing_experiment:
experiment = ExistingExperiment(
api_key="jBFVYFo9VUsy0kb0lioKXfTmM",
previous_experiment=args.existing_experiment)
else:
experiment = Experiment(api_key="jBFVYFo9VUsy0kb0lioKXfTmM",
project_name="fastdepth")
# Data loading code
print("Creating data loaders...")
if args.nyu:
from dataloaders.nyu import NYUDataset
val_dataset = NYUDataset(params["test_dataset_paths"], split='val')
else:
val_dataset = Datasets.FastDepthDataset(params["test_dataset_paths"],
split='val',
depth_min=params["depth_min"],
depth_max=params["depth_max"],
input_shape_model=(224, 224))
# set batch size to be 1 for validation
val_loader = torch.utils.data.DataLoader(val_dataset,
batch_size=1,
shuffle=True,
num_workers=params["num_workers"],
pin_memory=True)
# Set GPU
params["device"] = torch.device(
"cuda:{}".format(params["device"])
if params["device"] >= 0 and torch.cuda.is_available() else "cpu")
print("Using device", params["device"])
print("Loading model '{}'".format(args.model))
if not args.nyu:
model, _ = utils.load_model(params, args.model, params["device"])
else:
# Maintain compatibility for fastdepth NYU model format
state_dict = torch.load(args.model, map_location=params["device"])
model = models.MobileNetSkipAdd(output_size=(224, 224),
pretrained=True)
model.load_state_dict(state_dict)
params["start_epoch"] = 0
model.to(params["device"])
# Create output directory
output_directory = os.path.join(os.path.dirname(args.model), "images")
if not os.path.exists(output_directory):
os.makedirs(output_directory)
params["experiment_dir"] = output_directory
print("Saving results to " + output_directory)
evaluate(params, val_loader, model, experiment)
def set_up(self):
spec = load_config_file(self.args.yaml) # spec is a dict of dicts of dicts
# Import the correct model
self.params_dict = apply_defaults(spec["params"])
# time some things, like epoch time
start_time = time.time()
# ---------------------------------------- #
# DEFINE XVAL DATASETS #
# ---------------------------------------- #
# Create self.dataset_pair: DatasetPair containing train and val Datasets.
self._prepare_data(spec["data"])
# Number of instances to put in a training batch.
self.n_batch = min(self.params_dict['n_batch'], self.dataset_pair.n_train)
# This is already a model object because of the use of "!!python/object:... in the yaml file.
model = self.params_dict["model"]
# Set various attributes of the model
model.init_with_params(self.params_dict, self.procdata.relevance_vectors)
# Import priors from YAML
parameters = Parameters()
parameters.load(self.params_dict)
print("----------------------------------------------")
if self.args.verbose:
print("parameters:")
parameters.pretty_print()
n_vals = LocalAndGlobal.from_list(parameters.get_parameter_counts())
self.n_theta = n_vals.sum()
# TENSORFLOW PARTS #
self.placeholders = Placeholders(self.dataset_pair, n_vals)
# feed_dicts are used to supply placeholders, these are for the entire train/val dataset, there is a batch one below.
self._create_feed_dicts()
# time-series of species differences: x_delta_obs is BATCH x (nTimes-1) x nSpecies
x_delta_obs = self.placeholders.x_obs[:, 1:, :] - self.placeholders.x_obs[:, :-1, :]
# DEFINE THE ENCODER NN: for LOCAL PARAMETERS
print("Set up encoder")
self.encoder = Encoder(self.args.verbose, parameters, self.placeholders, x_delta_obs)
# DEFINE THE DECODER NN
print("Set up decoder")
self.decoder = Decoder(self.args.verbose, self.params_dict, self.placeholders, self.dataset_pair.times, self.encoder)
# DEFINE THE OBJECTIVE and GRADIENTS
# likelihood p (x | theta)
print("Set up objective")
self.objective = Objective(self.encoder, self.decoder, model, self.placeholders)
# SET-UP tensorflow LEARNING/OPTIMIZER
self.training_stepper = TrainingStepper(self.args.dreg, self.encoder, self.objective, self.params_dict)
time_interval = time.time() - start_time
print("Time before sess: %g" % time_interval)
# TENSORBOARD VISUALIZATION #
ts_to_vis = 1
self.encoder.q.attach_summaries() # global and local parameters of q distribution
unnormed_iw = self.objective.log_unnormalized_iws[ts_to_vis, :]
self_normed_iw = self.objective.normalized_iws[ts_to_vis, :] # not in log space
with tf.name_scope('IWS'):
variable_summaries(unnormed_iw, 'iws_unn_log')
variable_summaries(self_normed_iw, 'iws_normed')
tf.summary.scalar('nonzeros', tf.count_nonzero(self_normed_iw))
#print(tf.shape(log_p_observations))
with tf.name_scope('ELBO'):
tf.summary.scalar('log_p', tf.reduce_mean(self.training_stepper.logsumexp_log_p)) # [batch, 1]
tf.summary.scalar('log_prior', tf.reduce_mean(self.training_stepper.logsumexp_log_p_theta))
tf.summary.scalar('loq_q', tf.reduce_mean(self.training_stepper.logsumexp_log_q_theta))
tf.summary.scalar('elbo', self.objective.elbo)
def get_result():
test_subset_name = 'test'
all_params = load_config_file(tal_config_file)
locals().update(all_params)
# cas_file allows user to select and upload a single CAS file (i.e. Explosion008_x264.npz file)
# Doesn't matter from which CAS modalities(dirs), as all the modalities detection will be displayed
cas_file = request.files['cas_file']
videoname = cas_file.filename[:-4]
class_name = ''
num_class = len(ucf_crime_old_cls_names.keys())
for clss in range(num_class):
class_name = ucf_crime_old_cls_names[clss+1]
if videoname.startswith(class_name):
class_name = class_name
break
global video_path
video_path = os.path.join(ucf_video_dir, '{}/{}.mp4'.format(class_name, videoname))
video_path = video_path.replace('\\', '/')
list_modality = os.listdir(cas_dir)
talImageB64String, predicted_class_dict = {}, {}
merge_localized_actions = []
#Retrieve data from TAL
for mod in list_modality:
modality = mod.split('-')[-1] #Only get the modality name
pth_to_modality = os.path.join(cas_dir, mod)
pth_to_modality = os.path.join(pth_to_modality, '{}.npz'.format(videoname))
pth_to_modality = pth_to_modality.replace('\\', '/')
_, frame_cnt, _, detected_list = metric_scores(pth_to_modality, **all_params)
talImageB64String[modality], predicted_class, start_end_conf = tal.plot_tal_charts(
videoname, gt_file_pth, detected_list,
frame_cnt, modality
)
predicted_class_dict[modality] = predicted_class
start_end_conf['modality'] = modality
start_end_conf.start = start_end_conf.start * fps
start_end_conf.end = start_end_conf.end * fps
merge_localized_actions.append(start_end_conf)
# predict_df_display contains [start,end,confidence_score,modality] of every localized actions
predict_df_display = pd.concat(merge_localized_actions)
# For each modalities, sort [start & end] values
predict_df_display = predict_df_display.groupby(['modality']).apply(
lambda x: x.sort_values(['start', 'end'])
)
gt_talImageB64String, gt_start_end = tal.plot_gt_charts(
videoname, gt_file_pth, detected_list,
frame_cnt, _
)
# gt_df_display contains [start,end]
gt_df_display = pd.DataFrame(gt_start_end, columns=['start', 'end'])
gt_df_display = gt_df_display.sort_values(by=['start', 'end'])
#Retrieve data from MIL
x_axis_frame, y_axis_score, gt_bar = mil.visualize_mil(
videoname,
frame_cnt,
path_to_MIL_test_output_dir,
gt_file_pth
)
milImageB64String = mil.plot_mil_charts(
x_axis_frame, y_axis_score, gt_bar)
api_response = {
'modality': modality,
'video_path': video_path,
'fig_mil': milImageB64String,
'tal_gt': gt_talImageB64String,
'fig_tal_both': talImageB64String['both'],
'class_both': predicted_class_dict['both'],
'fig_tal_rgb': talImageB64String['rgb'],
'class_rgb': predicted_class_dict['rgb'],
'fig_tal_flow': talImageB64String['flow'],
'class_flow': predicted_class_dict['flow'],
'fig_tal_late_fusion': talImageB64String['fusion'],
'class_late_fusion': predicted_class_dict['fusion'],
'frame_cnt': frame_cnt,
'predict_df_display': [predict_df_display.to_html(index=False)],
'gt_df_display': [gt_df_display.to_html(index=False)]
}
return render_template(
"result.html",
**api_response
)
import numpy as np
import os
import pickle
from utils import load_config_file
import sys
if __name__ == '__main__':
config_file_name = sys.argv[1]
config = load_config_file(config_file_name)
# create output folder if it doesn't exist
output_folder = config["Output Folder"]
if not os.path.isdir(output_folder):
os.mkdir(output_folder)
# load test set
input_feature_path_testing = config["Input Feature Path Testing"]
X_testing = np.load(input_feature_path_testing)
input_age_path_testing = config["Input Age Path Testing"]
Y_testing = np.load(input_age_path_testing)
input_subjects_path_testing = config["Input Subject IDs Path Testing"]
IDs_testing = np.load(input_subjects_path_testing)
# extract PCs of test set
pretrained_folder = config["Pretrained Folder"]
with open(os.path.join(pretrained_folder, 'pca_training_set.pickle'), 'rb') as input_file:
pca = pickle.load(input_file)
X_testing = pca.transform(X_testing)
age_mean = np.load(os.path.join(pretrained_folder, 'mean_age_training_set.npy'))
Y_training_orig = np.load(os.path.join(pretrained_folder, 'y_training_set.npy'))
"""Build sources to remove dependencies from host
"""
import os
import utils
from utils.docker import *
import inspect
import pickle
import base64
from utils.config import username, group
section = utils.load_config_file()
class Build(object):
def __init__(self, project):
if project == 'custom':
self.src = section.get('src', None)
else:
self.src = section.get(project, None)
if self.src is None:
exit("Please configure source directory in MKT config.")
self.project = project
self.pickle = dict()
def _get_pickle(self):
self.pickle["project"] = self.project
self.pickle["src"] = self.src
self.pickle["checkpatch_root_dir"] = section.get('kernel', None)
if self.project == 'custom':
row_values[1],
"errand_status":
row_values[2],
"reporter":
row_values[3].title(),
"apartment":
row_values[4].upper(),
"errand_type":
row_values[8],
"errand_details":
row_values[9].replace("\n", " ").replace("\r", "").strip(),
}
transformed_errands.append(single_errand_data)
logging.info(
f"Raw data transformed, found {len(transformed_errands)} errands")
transformed_errands.reverse(
) # Reverse to insert oldest errand first in database
return transformed_errands
if __name__ == "__main__":
setup_logging()
configs = load_config_file()
browser = setup_browser(configs, runs_local=True)
page_soup = download_data(browser, configs)
all_table_rows = get_table_rows_from_soup(page_soup)
all_errands = transform_table_rows(all_table_rows)
print(all_errands)
parser.add_argument('--include-train',
dest='include_train',
action='store_true')
parser.add_argument('--no-include-train',
dest='include_train',
action='store_false')
parser.set_defaults(include_train=True)
args = parser.parse_args()
print(args.config_file)
print(args.train_subset_name)
print(args.test_subset_name)
print(args.include_train)
all_params = load_config_file(args.config_file)
locals().update(all_params)
if args.include_train:
train_dataset_dict = get_dataset(
dataset_name=dataset_name,
subset=args.train_subset_name,
file_paths=file_paths,
sample_rate=sample_rate,
base_sample_rate=base_sample_rate,
action_class_num=action_class_num,
modality='both',
feature_type=feature_type,
feature_oversample=False,
temporal_aug=False,
)
with open(
os.path.join(config["MODEL_OUTPUT_RESULTS_PATH"], "accuracy.csv"),
"w") as f_:
f_.write("class_number,accuracy\n")
for i in range(accuracy_per_class.shape[0]):
f_.write("{},{:.2f}\n".format(class_labels[i],
accuracy_per_class[i]))
return
if __name__ == "__main__":
print("Loading Config")
config_file = "./default-config,yaml"
config = load_config_file(config_file)
# dataloaders
print("Getting Dataloader")
test_loader, total_classes = load_dataset(config)
# creating model
print("Loading Model")
model = get_model(num_classes=total_classes)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model.to(device)
# loading model
if not os.path.isfile(config["MODEL_TEST_PATH"]):
print("Model checkpoint weights does no exist, exiting....")
import sys
