def main(args):
paddle.seed(12345)
# load config
config = load_yaml(args.config_yaml)
config["yaml_path"] = args.config_yaml
config["config_abs_dir"] = args.abs_dir
# load static model class
static_model_class = load_static_model_class(config)
input_data = static_model_class.create_feeds()
input_data_names = [data.name for data in input_data]
fetch_vars = static_model_class.net(input_data)
#infer_target_var = model.infer_target_var
logger.info("cpu_num: {}".format(os.getenv("CPU_NUM")))
static_model_class.create_optimizer()
use_gpu = config.get("runner.use_gpu", True)
use_auc = config.get("runner.use_auc", False)
auc_num = config.get("runner.auc_num", 1)
train_data_dir = config.get("runner.train_data_dir", None)
epochs = config.get("runner.epochs", None)
print_interval = config.get("runner.print_interval", None)
model_save_path = config.get("runner.model_save_path", "model_output")
model_init_path = config.get("runner.model_init_path", None)
batch_size = config.get("runner.train_batch_size", None)
reader_type = config.get("runner.reader_type", "DataLoader")
os.environ["CPU_NUM"] = str(config.get("runner.thread_num", 1))
logger.info("**************common.configs**********")
logger.info(
"use_gpu: {}, train_data_dir: {}, epochs: {}, print_interval: {}, model_save_path: {}"
.format(use_gpu, train_data_dir, epochs, print_interval,
model_save_path))
logger.info("**************common.configs**********")
place = paddle.set_device('gpu' if use_gpu else 'cpu')
exe = paddle.static.Executor(place)
# initialize
exe.run(paddle.static.default_startup_program())
last_epoch_id = config.get("last_epoch", -1)
if reader_type == 'QueueDataset':
dataset, file_list = get_reader(input_data, config)
elif reader_type == 'DataLoader':
train_dataloader = create_data_loader(config=config, place=place)
for epoch_id in range(last_epoch_id + 1, epochs):
epoch_begin = time.time()
if use_auc:
reset_auc(auc_num)
if reader_type == 'DataLoader':
fetch_batch_var = dataloader_train(epoch_id, train_dataloader,
input_data_names, fetch_vars,
exe, config)
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name,
fetch_batch_var[var_idx])
logger.info("epoch: {} done, ".format(epoch_id) + metric_str +
"epoch time: {:.2f} s".format(time.time() -
epoch_begin))
elif reader_type == 'QueueDataset':
fetch_batch_var = dataset_train(epoch_id, dataset, fetch_vars, exe,
config)
logger.info("epoch: {} done, ".format(epoch_id) +
"epoch time: {:.2f} s".format(time.time() -
epoch_begin))
else:
logger.info("reader type wrong")
save_static_model(paddle.static.default_main_program(),
model_save_path,
epoch_id,
prefix='rec_static')
def main(args):
paddle.seed(12345)
# load config
config = load_yaml(args.config_yaml)
config["yaml_path"] = args.config_yaml
config["config_abs_dir"] = args.abs_dir
# modify config from command
if args.opt:
for parameter in args.opt:
parameter = parameter.strip()
key, value = parameter.split("=")
config[key] = value
# load static model class
static_model_class = load_static_model_class(config)
input_data = static_model_class.create_feeds()
input_data_names = [data.name for data in input_data]
fetch_vars = static_model_class.net(input_data)
#infer_target_var = model.infer_target_var
logger.info("cpu_num: {}".format(os.getenv("CPU_NUM")))
static_model_class.create_optimizer()
use_gpu = config.get("runner.use_gpu", True)
use_auc = config.get("runner.use_auc", False)
use_visual = config.get("runner.use_visual", False)
use_inference = config.get("runner.use_inference", False)
auc_num = config.get("runner.auc_num", 1)
train_data_dir = config.get("runner.train_data_dir", None)
epochs = config.get("runner.epochs", None)
print_interval = config.get("runner.print_interval", None)
model_save_path = config.get("runner.model_save_path", "model_output")
model_init_path = config.get("runner.model_init_path", None)
batch_size = config.get("runner.train_batch_size", None)
reader_type = config.get("runner.reader_type", "DataLoader")
os.environ["CPU_NUM"] = str(config.get("runner.thread_num", 1))
logger.info("**************common.configs**********")
logger.info(
"use_gpu: {}, use_visual: {}, train_batch_size: {}, train_data_dir: {}, epochs: {}, print_interval: {}, model_save_path: {}"
.format(use_gpu, use_visual, batch_size, train_data_dir, epochs,
print_interval, model_save_path))
logger.info("**************common.configs**********")
place = paddle.set_device('gpu' if use_gpu else 'cpu')
exe = paddle.static.Executor(place)
# initialize
exe.run(paddle.static.default_startup_program())
last_epoch_id = config.get("last_epoch", -1)
# Create a log_visual object and store the data in the path
if use_visual:
from visualdl import LogWriter
log_visual = LogWriter(args.abs_dir + "/visualDL_log/train")
else:
log_visual = None
step_num = 0
if reader_type == 'QueueDataset':
dataset, file_list = get_reader(input_data, config)
elif reader_type == 'DataLoader':
train_dataloader = create_data_loader(config=config, place=place)
for epoch_id in range(last_epoch_id + 1, epochs):
epoch_begin = time.time()
if use_auc:
reset_auc(auc_num)
if reader_type == 'DataLoader':
fetch_batch_var, step_num = dataloader_train(
epoch_id, train_dataloader, input_data_names, fetch_vars, exe,
config, use_visual, log_visual, step_num)
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name,
fetch_batch_var[var_idx])
logger.info("epoch: {} done, ".format(epoch_id) + metric_str +
"epoch time: {:.2f} s".format(time.time() -
epoch_begin))
elif reader_type == 'QueueDataset':
fetch_batch_var = dataset_train(epoch_id, dataset, fetch_vars, exe,
config)
logger.info("epoch: {} done, ".format(epoch_id) +
"epoch time: {:.2f} s".format(time.time() -
epoch_begin))
else:
logger.info("reader type wrong")
save_static_model(paddle.static.default_main_program(),
model_save_path,
epoch_id,
prefix='rec_static')
if use_inference:
feed_var_names = config.get("runner.save_inference_feed_varnames",
[])
feedvars = []
fetch_var_names = config.get(
"runner.save_inference_fetch_varnames", [])
fetchvars = []
for var_name in feed_var_names:
if var_name not in paddle.static.default_main_program(
).global_block().vars:
raise ValueError(
"Feed variable: {} not in default_main_program, global block has follow vars: {}"
.format(
var_name,
paddle.static.default_main_program().global_block(
).vars.keys()))
else:
feedvars.append(paddle.static.default_main_program().
global_block().vars[var_name])
for var_name in fetch_var_names:
if var_name not in paddle.static.default_main_program(
).global_block().vars:
raise ValueError(
"Fetch variable: {} not in default_main_program, global block has follow vars: {}"
.format(
var_name,
paddle.static.default_main_program().global_block(
).vars.keys()))
else:
fetchvars.append(paddle.static.default_main_program().
global_block().vars[var_name])
save_inference_model(model_save_path, epoch_id, feedvars,
fetchvars, exe)
def main(args):
paddle.seed(12345)
# load config
config = load_yaml(args.config_yaml)
config["config_abs_dir"] = args.abs_dir
# load static model class
static_model_class = load_static_model_class(config)
input_data = static_model_class.create_feeds(is_infer=True)
input_data_names = [data.name for data in input_data]
fetch_vars = static_model_class.infer_net(input_data)
logger.info("cpu_num: {}".format(os.getenv("CPU_NUM")))
use_gpu = config.get("runner.use_gpu", True)
use_auc = config.get("runner.use_auc", False)
test_data_dir = config.get("runner.test_data_dir", None)
print_interval = config.get("runner.print_interval", None)
model_load_path = config.get("runner.infer_load_path", "model_output")
start_epoch = config.get("runner.infer_start_epoch", 0)
end_epoch = config.get("runner.infer_end_epoch", 10)
batch_size = config.get("runner.infer_batch_size", None)
os.environ["CPU_NUM"] = str(config.get("runner.thread_num", 1))
logger.info("**************common.configs**********")
logger.info(
"use_gpu: {}, test_data_dir: {}, start_epoch: {}, end_epoch: {}, print_interval: {}, model_load_path: {}"
.format(use_gpu, test_data_dir, start_epoch, end_epoch, print_interval,
model_load_path))
logger.info("**************common.configs**********")
place = paddle.set_device('gpu' if use_gpu else 'cpu')
exe = paddle.static.Executor(place)
# initialize
exe.run(paddle.static.default_startup_program())
test_dataloader = create_data_loader(config=config,
place=place,
mode="test")
for epoch_id in range(start_epoch, end_epoch):
logger.info("load model epoch {}".format(epoch_id))
model_path = os.path.join(model_load_path, str(epoch_id))
load_static_model(paddle.static.default_main_program(),
model_path,
prefix='rec_static')
runner_results = []
epoch_begin = time.time()
interval_begin = time.time()
if use_auc:
reset_auc()
for batch_id, batch_data in enumerate(test_dataloader()):
batch_runner_result = {}
fetch_batch_var = exe.run(
program=paddle.static.default_main_program(),
feed=dict(zip(input_data_names, batch_data)),
fetch_list=[var for _, var in fetch_vars.items()])
for var_idx, var_name in enumerate(fetch_vars):
batch_runner_result[var_name] = np.array(
fetch_batch_var[var_idx]).tolist()
runner_results.append(batch_runner_result)
if batch_id % print_interval == 0:
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(
var_name, fetch_batch_var[var_idx][0])
logger.info(
"epoch: {}, batch_id: {}, ".format(epoch_id, batch_id) +
metric_str +
"speed: {:.2f} ins/s".format(print_interval * batch_size /
(time.time() -
interval_begin)))
interval_begin = time.time()
reader_start = time.time()
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name,
fetch_batch_var[var_idx][0])
logger.info("epoch: {} done, ".format(epoch_id) + metric_str +
"epoch time: {:.2f} s".format(time.time() - epoch_begin))
runner_result_save_path = config.get("runner.runner_result_dump_path",
None)
if runner_result_save_path:
logging.info(
"Dump runner result in {}".format(runner_result_save_path))
with open(runner_result_save_path, 'w+') as fout:
json.dump(runner_results, fout)
def main(args):
paddle.seed(12345)
# load config
config = load_yaml(args.config_yaml)
config["config_abs_dir"] = args.abs_dir
# load static model class
static_model_class = load_static_model_class(config)
input_data = static_model_class.create_feeds(is_infer=True)
input_data_names = [data.name for data in input_data]
fetch_vars = static_model_class.infer_net(input_data)
logger.info("cpu_num: {}".format(os.getenv("CPU_NUM")))
use_gpu = config.get("runner.use_gpu", True)
use_auc = config.get("runner.use_auc", False)
use_visual = config.get("runner.use_visual", False)
auc_num = config.get("runner.auc_num", 1)
test_data_dir = config.get("runner.test_data_dir", None)
print_interval = config.get("runner.print_interval", None)
model_load_path = config.get("runner.infer_load_path", "model_output")
start_epoch = config.get("runner.infer_start_epoch", 0)
end_epoch = config.get("runner.infer_end_epoch", 10)
batch_size = config.get("runner.infer_batch_size", None)
os.environ["CPU_NUM"] = str(config.get("runner.thread_num", 1))
logger.info("**************common.configs**********")
logger.info(
"use_gpu: {}, use_visual: {}, test_data_dir: {}, start_epoch: {}, end_epoch: {}, print_interval: {}, model_load_path: {}".
format(use_gpu, use_visual, test_data_dir, start_epoch, end_epoch,
print_interval, model_load_path))
logger.info("**************common.configs**********")
place = paddle.set_device('gpu' if use_gpu else 'cpu')
exe = paddle.static.Executor(place)
# initialize
exe.run(paddle.static.default_startup_program())
test_dataloader = create_data_loader(
config=config, place=place, mode="test")
# Create a log_visual object and store the data in the path
if use_visual:
from visualdl import LogWriter
log_visual = LogWriter(args.abs_dir + "/visualDL_log/infer")
step_num = 0
for epoch_id in range(start_epoch, end_epoch):
logger.info("load model epoch {}".format(epoch_id))
model_path = os.path.join(model_load_path, str(epoch_id))
load_static_model(
paddle.static.default_main_program(),
model_path,
prefix='rec_static')
epoch_begin = time.time()
interval_begin = time.time()
if use_auc:
reset_auc(auc_num)
for batch_id, batch_data in enumerate(test_dataloader()):
fetch_batch_var = exe.run(
program=paddle.static.default_main_program(),
feed=dict(zip(input_data_names, batch_data)),
fetch_list=[var for _, var in fetch_vars.items()])
if batch_id % print_interval == 0:
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(
var_name, fetch_batch_var[var_idx][0])
if use_visual:
log_visual.add_scalar(
tag="infer/" + var_name,
step=step_num,
value=fetch_batch_var[var_idx][0])
logger.info("epoch: {}, batch_id: {}, ".format(
epoch_id, batch_id) + metric_str + "speed: {:.2f} ins/s".
format(print_interval * batch_size / (time.time(
) - interval_begin)))
interval_begin = time.time()
reader_start = time.time()
step_num = step_num + 1
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name,
fetch_batch_var[var_idx][0])
logger.info("epoch: {} done, ".format(epoch_id) + metric_str +
"epoch time: {:.2f} s".format(time.time() - epoch_begin))
def main(args):
paddle.seed(12345)
# load config
config = load_yaml(args.config_yaml)
config["config_abs_dir"] = args.abs_dir
# modify config from command
if args.opt:
for parameter in args.opt:
parameter = parameter.strip()
key, value = parameter.split("=")
if type(config.get(key)) is int:
value = int(value)
if type(config.get(key)) is float:
value = float(value)
if type(config.get(key)) is bool:
value = (True if value.lower() == "true" else False)
config[key] = value
# load static model class
static_model_class = load_static_model_class(config)
input_data = static_model_class.create_feeds(is_infer=True)
input_data_names = [data.name for data in input_data]
fetch_vars = static_model_class.infer_net(input_data)
logger.info("cpu_num: {}".format(os.getenv("CPU_NUM")))
use_gpu = config.get("runner.use_gpu", True)
use_xpu = config.get("runner.use_xpu", False)
use_auc = config.get("runner.use_auc", False)
use_visual = config.get("runner.use_visual", False)
auc_num = config.get("runner.auc_num", 1)
test_data_dir = config.get("runner.test_data_dir", None)
print_interval = config.get("runner.print_interval", None)
model_load_path = config.get("runner.infer_load_path", "model_output")
start_epoch = config.get("runner.infer_start_epoch", 0)
end_epoch = config.get("runner.infer_end_epoch", 10)
batch_size = config.get("runner.infer_batch_size", None)
use_save_data = config.get("runner.use_save_data", False)
reader_type = config.get("runner.reader_type", "DataLoader")
use_fleet = config.get("runner.use_fleet", False)
os.environ["CPU_NUM"] = str(config.get("runner.thread_num", 1))
logger.info("**************common.configs**********")
logger.info(
"use_gpu: {}, use_xpu: {}, use_visual: {}, infer_batch_size: {}, test_data_dir: {}, start_epoch: {}, end_epoch: {}, print_interval: {}, model_load_path: {}"
.format(use_gpu, use_xpu, use_visual, batch_size, test_data_dir,
start_epoch, end_epoch, print_interval, model_load_path))
logger.info("**************common.configs**********")
if use_xpu:
xpu_device = 'xpu:{0}'.format(os.getenv('FLAGS_selected_xpus', 0))
place = paddle.set_device(xpu_device)
else:
place = paddle.set_device('gpu' if use_gpu else 'cpu')
exe = paddle.static.Executor(place)
# initialize
exe.run(paddle.static.default_startup_program())
if reader_type == 'DataLoader':
test_dataloader = create_data_loader(config=config,
place=place,
mode="test")
elif reader_type == "CustomizeDataLoader":
test_dataloader = static_model_class.create_data_loader()
# Create a log_visual object and store the data in the path
if use_visual:
from visualdl import LogWriter
log_visual = LogWriter(args.abs_dir + "/visualDL_log/infer")
step_num = 0
for epoch_id in range(start_epoch, end_epoch):
logger.info("load model epoch {}".format(epoch_id))
model_path = os.path.join(model_load_path, str(epoch_id))
load_static_model(paddle.static.default_main_program(),
model_path,
prefix='rec_static')
epoch_begin = time.time()
interval_begin = time.time()
infer_reader_cost = 0.0
infer_run_cost = 0.0
reader_start = time.time()
if use_auc:
reset_auc(use_fleet, auc_num)
#we will drop the last incomplete batch when dataset size is not divisible by the batch size
assert any(
test_dataloader()
), "test_dataloader's size is null, please ensure batch size < dataset size!"
for batch_id, batch_data in enumerate(test_dataloader()):
infer_reader_cost += time.time() - reader_start
infer_start = time.time()
fetch_batch_var = exe.run(
program=paddle.static.default_main_program(),
feed=dict(zip(input_data_names, batch_data)),
fetch_list=[var for _, var in fetch_vars.items()])
infer_run_cost += time.time() - infer_start
if batch_id % print_interval == 0:
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(
var_name, fetch_batch_var[var_idx][0])
if use_visual:
log_visual.add_scalar(
tag="infer/" + var_name,
step=step_num,
value=fetch_batch_var[var_idx][0])
logger.info(
"epoch: {}, batch_id: {}, ".format(epoch_id, batch_id) +
metric_str +
"avg_reader_cost: {:.5f} sec, avg_batch_cost: {:.5f} sec, avg_samples: {:.5f}, ips: {:.2f} ins/s"
.format(
infer_reader_cost /
print_interval, (infer_reader_cost + infer_run_cost) /
print_interval, batch_size, print_interval *
batch_size / (time.time() - interval_begin)))
interval_begin = time.time()
infer_reader_cost = 0.0
infer_run_cost = 0.0
reader_start = time.time()
step_num = step_num + 1
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name,
fetch_batch_var[var_idx][0])
logger.info("epoch: {} done, ".format(epoch_id) + metric_str +
"epoch time: {:.2f} s".format(time.time() - epoch_begin))
if use_save_data:
save_data(fetch_batch_var, model_load_path)
def main(args):
paddle.seed(12345)
# load config
config = load_yaml(args.config_yaml)
config["config_abs_dir"] = args.abs_dir
# load static model class
static_model_class = load_static_model_class(config)
input_data = static_model_class.create_feeds()
input_data_names = [data.name for data in input_data]
fetch_vars = static_model_class.net(input_data)
#infer_target_var = model.infer_target_var
logger.info("cpu_num: {}".format(os.getenv("CPU_NUM")))
static_model_class.create_optimizer()
use_gpu = config.get("runner.use_gpu", True)
use_auc = config.get("runner.use_auc", False)
train_data_dir = config.get("runner.train_data_dir", None)
epochs = config.get("runner.epochs", None)
print_interval = config.get("runner.print_interval", None)
model_save_path = config.get("runner.model_save_path", "model_output")
model_init_path = config.get("runner.model_init_path", None)
batch_size = config.get("runner.train_batch_size", None)
os.environ["CPU_NUM"] = str(config.get("runner.thread_num", 1))
logger.info("**************common.configs**********")
logger.info(
"use_gpu: {}, train_data_dir: {}, epochs: {}, print_interval: {}, model_save_path: {}".
format(use_gpu, train_data_dir, epochs, print_interval,
model_save_path))
logger.info("**************common.configs**********")
place = paddle.set_device('gpu' if use_gpu else 'cpu')
exe = paddle.static.Executor(place)
# initialize
exe.run(paddle.static.default_startup_program())
last_epoch_id = config.get("last_epoch", -1)
train_dataloader = create_data_loader(config=config, place=place)
for epoch_id in range(last_epoch_id + 1, epochs):
epoch_begin = time.time()
interval_begin = time.time()
train_reader_cost = 0.0
train_run_cost = 0.0
total_samples = 0
reader_start = time.time()
if use_auc:
reset_auc()
for batch_id, batch_data in enumerate(train_dataloader()):
train_reader_cost += time.time() - reader_start
train_start = time.time()
fetch_batch_var = exe.run(
program=paddle.static.default_main_program(),
feed=dict(zip(input_data_names, batch_data)),
fetch_list=[var for _, var in fetch_vars.items()])
train_run_cost += time.time() - train_start
total_samples += batch_size
if batch_id % print_interval == 0:
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name,
fetch_batch_var[var_idx])
logger.info(
"epoch: {}, batch_id: {}, ".format(epoch_id,
batch_id) + metric_str +
"avg_reader_cost: {:.5f} sec, avg_batch_cost: {:.5f} sec, avg_samples: {:.5f}, ips: {:.5f} images/sec".
format(train_reader_cost / print_interval, (
train_reader_cost + train_run_cost) / print_interval,
total_samples / print_interval, total_samples / (
train_reader_cost + train_run_cost)))
train_reader_cost = 0.0
train_run_cost = 0.0
total_samples = 0
reader_start = time.time()
metric_str = ""
for var_idx, var_name in enumerate(fetch_vars):
metric_str += "{}: {}, ".format(var_name, fetch_batch_var[var_idx])
logger.info("epoch: {} done, ".format(epoch_id) + metric_str +
"epoch time: {:.2f} s".format(time.time() - epoch_begin))
save_static_model(
paddle.static.default_main_program(),
model_save_path,
epoch_id,
prefix='rec_static')