AntoineH/smart_augmentation
1
0
Fork 0
mirror of https://github.com/AntoineHX/smart_augmentation.git synced 2025-05-04 12:10:45 +02:00
Code Issues Projects Releases Packages Wiki Activity

F1 par classes + plot OK

Browse source
This commit is contained in:
Harle, Antoine (Contracteur) 2020-02-03 11:21:54 -05:00
parent 3ccacd0366
commit fcd0217d54
5 changed files with 57 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

29
higher/smart_aug/compare_res.py
View file

@ -4,7 +4,7 @@ if __name__ == "__main__":
#'''
files=[
"res/log/Aug_mod(Data_augV5(Mix0.8-23TFx4-Mag)-LeNet)-100 epochs (dataug:0)- 1 in_it.json",
"../res/log/Aug_mod(Data_augV5(Mix0.8-3TFx2-MagFx)-resnet18)-2 epochs (dataug:0)- 1 in_it.json",
#"res/brutus-tests/log/Aug_mod(Data_augV5(Uniform-14TFx3-MagFxSh)-LeNet)-150epochs(dataug:0)-10in_it-0.json",
#"res/brutus-tests/log/Aug_mod(Data_augV5(Uniform-14TFx3-MagFxSh)-LeNet)-150epochs(dataug:0)-10in_it-1.json",
#"res/brutus-tests/log/Aug_mod(Data_augV5(Uniform-14TFx3-MagFxSh)-LeNet)-150epochs(dataug:0)-10in_it-2.json",
@ -15,7 +15,7 @@ if __name__ == "__main__":
#legend+=str(idx)+'-'+file+'\n'
with open(file) as json_file:
data = json.load(json_file)
plot_resV2(data['Log'], fig_name=file.replace('.json','').replace('log/',''), param_names=data['Param_names'])
plot_resV2(data['Log'], fig_name=file.replace("/log","").replace(".json",""), param_names=data['Param_names'])
#plot_TF_influence(data['Log'], param_names=data['Param_names'])
#'''
## Loss , Acc, Proba = f(epoch) ##
@ -92,4 +92,29 @@ if __name__ == "__main__":
print(idx, data['Accuracy'])
print(files[0], np.mean(accs), np.std(accs), np.mean(times))
'''
'''
inner_its = [1]
dist_mix = [0]#[0.5, 0.8, 1.0] #Uniform
N_seq_TF= [4, 3, 2]
nb_run= 3
for n_inner_iter in inner_its:
for n_tf in N_seq_TF:
for dist in dist_mix:
#files = ["../res/brutus-tests2/log/Aug_mod(Data_augV5(Mix%.1f-14TFx%d-MagFxSh)-ResNet18)-150 epochs (dataug:0)- 1 in_it-%s.json"%(dist, n_tf, str(run)) for run in range(nb_run)]
files = ["../res/brutus-tests2/log/Aug_mod(Data_augV5(Uniform-14TFx%d-MagFxSh)-ResNet18)-150 epochs (dataug:0)- 1 in_it-%s.json"%(n_tf, str(run)) for run in range(nb_run)]
accs = []
times = []
for idx, file in enumerate(files):
#legend+=str(idx)+'-'+file+'\n'
with open(file) as json_file:
data = json.load(json_file)
accs.append(data['Accuracy'])
times.append(data['Time'][0])
print(idx, data['Accuracy'])
print(files[0], 'acc', np.mean(accs), '+-',np.std(accs), ',t', np.mean(times))
'''

6
higher/smart_aug/datasets.py
View file

@ -13,7 +13,7 @@ TEST_SIZE = BATCH_SIZE
#TEST_SIZE = 10000 #legerement +Rapide / + Consomation memoire !
#Wether to download data.
download_data=True
download_data=False
#Number of worker to use.
num_workers=2 #4
#Pin GPU memory
@ -46,7 +46,7 @@ transform_train = torchvision.transforms.Compose([
#data_test = torchvision.datasets.MNIST("../data", train=False, download=True, transform=transform)
#CIFAR
data_train = torchvision.datasets.CIFAR10("../data", train=True, download=download_data, transform=transform_train)
data_val = torchvision.datasets.CIFAR10("../data", train=True, download=download_data, transform=transform)
#data_val = torchvision.datasets.CIFAR10("../data", train=True, download=download_data, transform=transform)
data_test = torchvision.datasets.CIFAR10("../data", train=False, download=download_data, transform=transform)
train_subset_indices=range(int(len(data_train)/2))
@ -55,5 +55,5 @@ val_subset_indices=range(int(len(data_train)/2),len(data_train))
#val_subset_indices=range(BATCH_SIZE*10, BATCH_SIZE*20)
dl_train = torch.utils.data.DataLoader(data_train, batch_size=BATCH_SIZE, shuffle=False, sampler=SubsetRandomSampler(train_subset_indices), num_workers=num_workers, pin_memory=pin_memory)
dl_val = torch.utils.data.DataLoader(data_val, batch_size=BATCH_SIZE, shuffle=False, sampler=SubsetRandomSampler(val_subset_indices), num_workers=num_workers, pin_memory=pin_memory)
dl_val = torch.utils.data.DataLoader(data_train, batch_size=BATCH_SIZE, shuffle=False, sampler=SubsetRandomSampler(val_subset_indices), num_workers=num_workers, pin_memory=pin_memory)
dl_test = torch.utils.data.DataLoader(data_test, batch_size=TEST_SIZE, shuffle=False, num_workers=num_workers, pin_memory=pin_memory)

24
higher/smart_aug/test_dataug.py
View file

@ -79,7 +79,7 @@ if __name__ == "__main__":
}
#Parameters
n_inner_iter = 1
epochs = 2
epochs = 10
dataug_epoch_start=0
optim_param={
'Meta':{
@ -108,7 +108,7 @@ if __name__ == "__main__":
print("{} on {} for {} epochs".format(model_name, device_name, epochs))
log= train_classic(model=model, opt_param=optim_param, epochs=epochs, print_freq=5)
log= train_classic(model=model, opt_param=optim_param, epochs=epochs, print_freq=1)
#log= train_classic_higher(model=model, epochs=epochs)
exec_time=time.process_time() - t0
@ -116,13 +116,21 @@ if __name__ == "__main__":
print('-'*9)
times = [x["time"] for x in log]
out = {"Accuracy": max([x["acc"] for x in log]), "Time": (np.mean(times),np.std(times), exec_time), 'Optimizer': optim_param['Inner'], "Device": device_name, "Log": log}
print(str(model),": acc", out["Accuracy"], "in:", out["Time"][0], "+/-", out["Time"][1])
print(model_name,": acc", out["Accuracy"], "in:", out["Time"][0], "+/-", out["Time"][1])
filename = "{}-{} epochs".format(model_name,epochs)
with open("../res/log/%s.json" % filename, "w+") as f:
json.dump(out, f, indent=True)
print('Log :\"',f.name, '\" saved !')
try:
json.dump(out, f, indent=True)
print('Log :\"',f.name, '\" saved !')
except:
print("Failed to save logs :",f.name)
print(sys.exc_info()[1])
#plot_res(log, fig_name="../res/"+filename)
try:
plot_resV2(log, fig_name="../res/"+filename)
except:
print("Failed to plot res")
print(sys.exc_info()[1])
print('Execution Time : %.00f '%(exec_time))
print('-'*9)
@ -160,12 +168,12 @@ if __name__ == "__main__":
print('Log :\"',f.name, '\" saved !')
except:
print("Failed to save logs :",f.name)
print(sys.exc_info()[0])
print(sys.exc_info()[1])
try:
plot_resV2(log, fig_name="../res/"+filename, param_names=aug_model.TF_names())
except:
print("Failed to plot res")
print(sys.exc_info()[0])
print(sys.exc_info()[1])
print('Execution Time : %.00f '%(exec_time))
print('-'*9)

6
higher/smart_aug/train_utils.py
View file

@ -52,7 +52,7 @@ def test(model):
#from sklearn.metrics import f1_score
#f1 = f1_score(labels.data.to('cpu'), predicted.data.to('cpu'), average="macro")
return accuracy, confmat.f1_metric(average="macro")
return accuracy, confmat.f1_metric(average=None)
def compute_vaLoss(model, dl_it, dl):
"""Evaluate a model on a batch of data.
@ -167,7 +167,7 @@ def train_classic(model, opt_param, epochs=1, print_freq=1):
tf = time.process_time()
val_loss = compute_vaLoss(model=model, dl_it=dl_val_it, dl=dl_val)
accuracy, _ =test(model)
accuracy, f1 =test(model)
model.train()
#### Print ####
@ -177,6 +177,7 @@ def train_classic(model, opt_param, epochs=1, print_freq=1):
print('Time : %.00f'%(tf - t0))
print('Train loss :',loss.item(), '/ val loss', val_loss.item())
print('Accuracy :', accuracy)
print('F1 :', f1.data)
#### Log ####
data={
@ -184,6 +185,7 @@ def train_classic(model, opt_param, epochs=1, print_freq=1):
"train_loss": loss.item(),
"val_loss": val_loss.item(),
"acc": accuracy,
"f1": f1.cpu().numpy().tolist(),
"time": tf - t0,
"param": None,

14
higher/smart_aug/utils.py
View file

@ -115,15 +115,15 @@ def plot_resV2(log, fig_name='res', param_names=None):
ax[1, 0].plot(epochs,[x["acc"] for x in log], label='Acc')
if "f1" in log[0].keys():
ax[1, 0].plot(epochs,[x["f1"]*100 for x in log], label='F1')
'''
#ax[1, 0].plot(epochs,[x["f1"]*100 for x in log], label='F1')
#'''
#print(log[0]["f1"])
if len(log[0]["f1"])==1:
ax[1, 0].plot(epochs,[x["f1"]*100 for x in log], label='F1')
else:
if isinstance(log[0]["f1"], list):
for c in range(len(log[0]["f1"])):
ax[1, 0].plot(epochs,[x["f1"][c]*100 for x in log], label='F1-'+str(c))
'''
ax[1, 0].plot(epochs,[x["f1"][c]*100 for x in log], label='F1-'+str(c), ls='--')
else:
ax[1, 0].plot(epochs,[x["f1"]*100 for x in log], label='F1', ls='--')
#'''
ax[1, 0].legend()