def prepare_cifar_10_data(use_norm_shift=False, use_norm_scale=True):
validation_data_size = 5000 # Size of the validation set.
train_data = []
train_label = []
args = parse_args()
for id in range(1, 6):
train_filename = os.path.join(args.data_dir, "data_batch_%d" % id)
train_data_batch, train_label_batch = load_CIFAR_batch(train_filename)
train_data.append(train_data_batch)
train_label.append(train_label_batch)
train_data = np.concatenate(train_data) #50000 * 3072(32 * 32 * 3)
train_label = np.concatenate(train_label).reshape(-1, 1) #50000 * 1
test_filename = os.path.join(args.data_dir, "test_batch")
test_data, test_label = load_CIFAR_batch(test_filename)
test_label = test_label.reshape(-1, 1)
# Generate a validation set.
validation_data = train_data[:validation_data_size, :]
validation_labels = train_label[:validation_data_size, :]
train_data = train_data[validation_data_size:, :]
train_label = train_label[validation_data_size:, :]
#data normalization
train_data = data_normalize(train_data, use_norm_shift, use_norm_scale)
test_data = data_normalize(test_data, use_norm_shift, use_norm_scale)
validation_data = data_normalize(validation_data, use_norm_shift,
use_norm_scale)
return train_data, train_label, validation_data, validation_labels, test_data, test_label
def get_args(output_dir):
return run.parse_args(
[output_dir],
{
k: int(v.value) if "boolean" in v.tags else v.value
for k, v in run_models.items() if not v.disabled
},
)
def __init__(self):
self.resource_dir = "../resource/"
self.instance_dir = "../instances/"
args = parse_args()
args.optimizer = 'adam'
args.loss = 'binary_crossentropy'
args.need_char_level = True
args.need_word_level = True
args.word_trainable = False
args.char_trainable = False
args.lr = 0.001
args.save_dir = "../saved_models/"
args.word_emb_dir = "../instances/word_embed.txt"
args.char_emb_dir = "../instances/char_embed.txt"
args.r_dir = "../resource/"
self.name = "Bi-GRU2"
super(LiuModel1, self).__init__(args)
# print(param.sim_times[0:step].shape)
# exit()
result = np.hstack((param.sim_times[0:step].reshape(-1,1), states[0:step]))
# store in binary format
basename = os.path.splitext(os.path.basename(instance))[0]
folder_name = "../results/doubleintegrator/{}".format(name)
if not os.path.exists(folder_name):
os.mkdir(folder_name)
output_file = "{}/{}.npy".format(folder_name, basename)
with open(output_file, "wb") as f:
np.save(f, result.astype(np.float32), allow_pickle=False)
if __name__ == '__main__':
args = parse_args()
param = DoubleIntegratorParam()
env = DoubleIntegrator(param)
if args.il:
run(param, env, None, None, args)
exit()
controllers = {
# 'emptywapf': Empty_Net_wAPF(param,env,torch.load('../results/doubleintegrator/exp1Empty_0/il_current.pt')),
# 'e2e':torch.load('../results/doubleintegrator/exp1Barrier_0/il_current.pt'),
# 'empty':torch.load('../results/doubleintegrator/exp1Empty_0/il_current.pt'),
# 'current':torch.load(param.il_train_model_fn),
# 'current_wapf': Empty_Net_wAPF(param,env,torch.load(param.il_train_model_fn)),
# 'gg': GoToGoalPolicy(param,env),
def test_argument_parser_accept_one_bug(self) -> None:
"""
Tests that it is possible to give one bug and one plugin
"""
args = run.parse_args(["success", "pbzip-2094"])
self.assertEqual(len(args["bugs"]), 1)
def test_argument_parser_accept_one_bug(self) -> None:
"""
Tests that it is possible to give one bug and one plugin
"""
args = run.parse_args(["success", "pbzip-2094"])
self.assertEqual(len(args["bugs"]), 1)
"""
测试MULT中使用的ATTENTION机制的特点，是否是强变强、其它变弱？
"""
import torch
from config.config_run import Config
from data.load_data import MMDataLoader
from models.AMIO import AMIO
from run import parse_args
from trains.ATIO import ATIO
model_path = "/home/zhuchuanbo/paper_code/results/model_saves/mult-sims-M.pth"
# 进行参数配置
configs = Config(parse_args()).get_config()
device = torch.device('cuda:%d' % configs.gpu_ids[0])
configs.device = device
# 定义并且加载模型
dataloader = MMDataLoader(configs)
model = AMIO(configs).to(device)
model.load_state_dict(torch.load(model_path))
model.eval()
atio = ATIO().get_train(configs)
results = atio.do_test(model, dataloader['test'], mode="TEST")
import torchvision.models as models
def test_run():
# Test running the training for the bug model.
run.main(run.parse_args(["--train", "--goal", "defect"]))
# Test loading the trained model.
run.main(run.parse_args(["--goal", "defect"]))
# --viz-action custom ^
# --viz-camera 0 ^
# --viz-video "video_input/%myvideo%" ^
# --viz-output "%output%".mp4 ^
# --viz-size 6 ^
# --output_json "%output%".json
#
#
import inference.infer_video_d2 as step2
import data.prepare_data_2d_custom as step4
import run as step5
if __name__ == "__main__":
step2.setup_logger()
# we parse the args only once so we can catch all of them.
args = step5.parse_args()
steps: str = args.steps
if '2' in steps:
step2.main(args)
if '4' in steps:
step4.the_main_thing(args)
if '5' in steps:
step5.the_main_kaboose(args)