def main():
str1 = input("Please input path:")
template = train.learn(r'C:\Users\hp\Desktop\code\code\easy\train')
with open('result.csv', 'w', newline='') as f:
writer = csv.writer(f)
writer.writerow(('name', 'code'))
for home, dir, files in os.walk(str1):
for filename in files:
numList = pictureProcess.imageProcess(
os.path.join(home, filename))
result = ''
for i in numList:
distance = []
for j in template:
distance.append(calDistance(i, j))
result += str(distance.index(min(distance)))
writer.writerow((filename, result))
def main():
all=learn(r'C:\Users\hp\Desktop\code\code\hard\train')
str1=input("Please input path:")
with open('result.csv','w',newline='') as f:
writer = csv.writer(f)
writer.writerow(('name','code'))
for home,dir,files in os.walk(str1):
for filename in files:
info = preProcess(os.path.join(home, filename))
answer=''
for i in info:
vote=[]
#print(len(i))
distance=[]
for j in all:
distance.append(calDistance(j[0],i))
index=distance.index(min(distance))
vote.append([distance[index],all[index][1]])
distance[index]=1000
index=distance.index(min(distance))
vote.append([distance[index],all[index][1]])
distance[index]=1000
index=distance.index(min(distance))
vote.append([distance[index],all[index][1]])
distance[index]=1000
if(vote[0][1]==vote[1][1]):
answer+=str(vote[0][1])
continue
elif (vote[0][1]==vote[2][1]):
answer+=str(vote[0][1])
continue
elif (vote[1][1]==vote[2][1]):
answer+=str(vote[1][1])
continue
else:
answer+=str(vote[0][1])
writer.writerow((filename,answer))
'''
def match(query):
res = matchQuerySkipgram(bootstrapQuery(query),learn())
return res
except:
signals, labels = get_ecg(PATH, length=LENGTH)
segments = np.zeros((245990, 1001))
k = 0
for i, record in enumerate(signals):
rp = qrs_detection(record, sample_rate=FS)
seg = get_segments(record, rp, labels[i])
if seg is not None:
segments[k:k + seg.shape[0], :] = seg
k += seg.shape[0]
del signals, labels
np.save('./data/segment.npy', segments)
X, y = segments[:, :-1], segments[:, -1][:, np.newaxis]
del segments
train, test = build_dataloader(X, y, resamp=RESAMP, batch_size=BATCH_SIZE)
del X, y
net = cnn_feed_lstm()
try:
params = torch.load("../params/net_0.81.pkl")
net.load_state_dict(params["model_state_dict"])
except:
pass
loss, val_score = learn(net, train, test, lr=LR, epoch=EPOCH)
plot(loss, val_score)
"optimizer": optimizer_name,
"loss": loss_name,
"scheduler": scheduler_name,
"tolerance_es": tolerance,
"delta_es": delta,
"gamma_scheduler": gamma_scheduler,
"top_k": top_k,
"num_warmup": num_warmup_steps,
"lr_layer_decay": lr_layerdecay,
"freeze": freeze,
"alpha_link": alpha_link,
"comment": comment,
}
# Module used in the model
modules = {
"model": model,
"optimizer": optimizer,
"scheduler": scheduler,
"loss_function": loss_function,
"device": device,
"early_stopper": early_stopper,
}
# Train the model
learn(
train_dataloader,
val_dataloader,
modules,
hyperparameters,
experiment_name="seed",
)
attention = Attention(attention_dim, attention_dim, attention_dim)
model = Classifier(embedding, encoder, attention, attention_dim, nlabels)
model.to(device)
criterion = nn.MultiLabelSoftMarginLoss()
optimizer = torch.optim.Adam(model.parameters(),
model_argument['lr'],
amsgrad=True)
R = []
P = []
try:
best_valid_loss = None
for epoch in range(1, model_argument['epochs'] + 1):
loss_train = learn(model, train_iter, optimizer, criterion)
print("[{} loss]: {:.5f}".format('Train', loss_train))
loss, r, p = evaluate(model, test_iter, criterion)
R.append(r)
P.append(p)
if not best_valid_loss or loss < best_valid_loss:
best_valid_loss = loss
np.save('recall.npy', R)
np.save('precison.npy', P)
except KeyboardInterrupt:
print("[Ctrl+C] Training stopped!")
torch.save(model, 'training_history/mlc_20180903.pt')
def match(query):
res = matchQuerySkipgram(bootstrapQuery(query), learn())
return res
# for _ in range(n_drones)
# ]
# optimizers = [optim.AdamW(nets[i].parameters(), lr=args.lr) for i in range(n_drones)]
# train(
# args=args,
# nets=nets,
# optimizers=optimizers,
# env=env,
# obs_size=state_size,
# n_drones=n_drones,
# )
actor_critics = [
Policy((state_size + 2,), (action_size,), base_kwargs={"recurrent": True})
for _ in range(n_drones)
]
optimizers = [
optim.RMSprop(actor_critics[i].parameters(), lr=args.lr) for i in range(n_drones)
]
learn(
args=args,
actor_critics=actor_critics,
optimizers=optimizers,
env=env,
obs_size=state_size,
n_drones=n_drones,
)
[nn.Softplus()] * nmlp + [None])
for _ in range(nlayers)
], [
utils.SimpleMLPreshape(
[num] + [hidden] * nmlp + [num],
[nn.Softplus()] * nmlp + [utils.ScalableTanh(num)])
for _ in range(nlayers)
])
return fl
from utils import flowBuilder
f = flowBuilder(n,
numFlow,
innerBuilder,
1,
relax=args.relax,
shift=args.shift).to(device)
if not args.double:
f = f.to(torch.float32)
LOSS = train.learn(target,
f,
batchSize,
epochs,
lr,
saveSteps=lossPlotStep,
savePath=rootFolder)