danielcwq

from fastai.vision.all import *
path = untar_data(URLs.PETS)/'images'

def is_cat(x): return x[0].isupper()
dls = ImageDataLoaders.from_name_func(
    path, get_image_files(path), valid_pct=0.2, seed=42,
    label_func=is_cat, item_tfms=Resize(224))

learn = cnn_learner(dls, resnet34, metrics=error_rate)
learn.fine_tune(1)

danielcwq

def init_params(size, std=1.0): 
  return (torch.randn(size)*std).requires_grad_()
weights = init_params((28*28,1))
bias = init_params(1)

danielcwq

def mnist_loss(predictions, targets):
    return torch.where(targets==1, 1-predictions, predictions).mean()
 

danielcwq

def sigmoid(x): 
  return 1/(1+torch.exp(-x))

danielcwq

def calc_grad(xb, yb, model):
    preds = model(xb)
    loss = mnist_loss(preds, yb)
    loss.backward()
weights.grad.zero_()
bias.grad.zero_();

danielcwq

#in pseudocode, this would be: 
w -= gradient(w) *lr 



#stepping learning function here
#converting this into a function: 
def train_epoch(model, lr, params):
    for xb,yb in dl:

danielcwq

def batch_accuracy(xb, yb):
    preds = xb.sigmoid()
    correct = (preds>0.5) == yb
    return correct.float().mean()
def validate_epoch(model):
    accs = [batch_accuracy(model(xb), yb) for xb,yb in valid_dl]
    return round(torch.stack(accs).mean().item(), 4)

danielcwq

res = res.max(tensor(0.0))

danielcwq

simple_net = nn.Sequential(
    nn.Linear(28*28,30),
    nn.ReLU(),
    nn.Linear(30,1)
)

danielcwq

valid_x = torch.cat([valid_2_tens, valid_9_tens]).view(-1, 28*28) 
valid_y = tensor([1]*len(valid_2_tens) +[0]*len(valid_9_tens)).unsqueeze(1) 
valid_dset = list(zip(valid_x,valid_y))