def create_inference_model(checkpoint: str = None, model='resnet34', path='.'):
if model == 'resnet34':
model = resnet34
elif model == 'resnet18':
model = resnet18
elif model == 'mobilenet_v2':
model = mobilenet_v2
# Create an inference model instance and load the requested checkpoint
inf_db = DataBlock(blocks=[ImageBlock, CategoryBlock],
get_x=ItemGetter(0),
get_y=ItemGetter(1))
dummy_img = PILImage.create(np.zeros((415, 415, 3), dtype=np.uint8))
source = [(dummy_img, False), (dummy_img, True)]
inf_dls = inf_db.dataloaders(source)
if model == mobilenet_v2:
learner = cnn_learner(inf_dls,
model,
cut=-1,
splitter=_mobilenetv2_split,
pretrained=False)
else:
learner = cnn_learner(inf_dls, model, pretrained=False)
learner.path = Path(path)
if checkpoint is not None:
learner.load(checkpoint, with_opt=False, device='cpu')
return learner
def fake_dataloaders(a=2, b=3, bs=16, n=10):
def get_data(n):
x = torch.randn(bs * n, 1)
return torch.cat((x, a * x + b + 0.1 * torch.randn(bs * n, 1)), 1)
ds = get_data(n)
dblock = DataBlock()
return dblock.dataloaders(ds)
def pack_models(path: str) -> None:
model = LinearModel()
loss = Loss()
dblock = DataBlock(get_items=get_items, get_y=np.sum)
dls = dblock.datasets(None).dataloaders()
learner = Learner(dls, model, loss)
FastAIModel(learner).save(path)
def get_data(inputs,
df_all=None,
batch_tfms=None,
item_tfms=None,
verbose=False,
autoencoder=False):
if df_all is None:
df_all = get_dataframe(inputs, verbose)
if item_tfms is None:
tfms = [Resize(128, method="squish")]
else:
tfms = item_tfms
if autoencoder:
blocks = (ImageBlock, ImageBlock)
y_reader = ColReader("cam/image_array")
else:
blocks = (ImageBlock, RegressionBlock(n_out=2))
y_reader = ColReader(['user/angle', 'user/throttle'])
pascal = DataBlock(blocks=blocks,
splitter=RandomSplitter(),
get_x=ColReader("cam/image_array"),
get_y=y_reader,
item_tfms=tfms,
batch_tfms=batch_tfms,
n_inp=1)
dls = pascal.dataloaders(df_all)
if verbose:
dls.show_batch()
dls.one_batch()[0].shape
return dls
# %%
# %%
mnist_dls_rgb = ImageDataLoaders.from_folder(
mnist_dir,
train="training",
valid="testing",
device=device,
)
# %%
mnist_block = DataBlock(
blocks=(ImageBlock(cls=PILImageBW), CategoryBlock),
get_items=get_image_files,
splitter=GrandparentSplitter(train_name="training", valid_name="testing"),
get_y=parent_label,
# batch_tfms=aug_transforms(mult=1.2, do_flip=False)
)
# %%
mnist_dls = mnist_block.dataloaders(mnist_dir)
# %%
mnist_dls.train.one_batch()[0].shape
# %%
mnist_dls.show_batch()
# %% [markdown]
#
import torch
from fastai.data.block import DataBlock, CategoryBlock, get_image_files, GrandparentSplitter, parent_label
from fastai.metrics import error_rate
from fastai.vision.augment import Resize, RandomResizedCrop, aug_transforms
from fastai.vision.data import ImageBlock
from fastai.vision.all import cnn_learner, ClassificationInterpretation, load_learner, nn, partial, MixUp, xresnet50, \
accuracy, top_k_accuracy, Learner
from torchvision.models import resnet18
import matplotlib.pyplot as plt
from fastai.distributed import *
# In[]:
products = DataBlock(blocks=(ImageBlock, CategoryBlock),
get_items=get_image_files,
splitter=GrandparentSplitter(train_name="train",
valid_name="validation"),
get_y=parent_label,
item_tfms=Resize(192))
products = products.new(item_tfms=RandomResizedCrop(168, min_scale=0.8),
batch_tfms=aug_transforms())
project_path = Path("/home/yaro/Workspace/fastai/")
dataset_path = project_path.joinpath("for_test")
dls = products.dataloaders(dataset_path)
gpu = None
if torch.cuda.is_available():
if gpu is not None: torch.cuda.set_device(gpu)
n_gpu = torch.cuda.device_count()
else: