def main(ensemble, tta, output):
# Read in test data images from the 'data/test' folder
print("Loading test data.")
test_imgs = ImageList.from_folder(path=os.path.join(DATA_DIR,
TEST_FOLDER), )
# Get predictions
if ensemble:
# Load ensemble of learners
learners = []
learner_names = ['dpn92', 'inceptionv4', 'se_resnext101']
for name in learner_names:
print(f"Loading {name}")
learn = load_learner(SAVED_DIR, f'{name}.pkl', test=test_imgs)
learners.append(learn)
# Init ensemble
print("Initializing ensemble.")
ensemble = Ensemble(learners)
# Get predictions
print("Performing inference...")
preds = ensemble.predict(tta)
print("Predictions done.")
# Get classes list
classes = learners[0].data.classes
# Get image names list
img_names = [i.name for i in learners[0].data.test_ds.items]
else:
learner_name = 'se_resnext101'
# Initialize Learner
print(f"Loading {learner_name}")
learn = load_learner(SAVED_DIR, f'{learner_name}.pkl', test=test_imgs)
# Get predictions
print("Performing inference...")
if tta:
preds, _ = learn.TTA(ds_type=DatasetType.Test)
else:
preds, _ = learn.get_preds(ds_type=DatasetType.Test)
print("Predictions done.")
# Get classes list
classes = learn.data.classes
# Get image names list
img_names = [i.name for i in learn.data.test_ds.items]
# Initialize DataFrame with the predictions
df = pd.DataFrame(np.array(preds), columns=classes)
# Insert image names to DataFrame
df.insert(0, 'img_name', img_names)
# Save predictions as csv file
df.to_csv(output, index=False)
print(f"Predictions saved to {output}")
def __init__(self, root_model_path, fwd_model_file_name, bwd_model_file_name, threshold=0.5):
# root_model_path = "outputs/experiment_ckpts/ulmfit-dialog_babi_data_model"
self.fwd_model = load_learner(root_model_path, fwd_model_file_name)
self.bwd_model = load_learner(root_model_path, bwd_model_file_name)
self.ds = self.fwd_model.data.single_ds.y
self.trigger_functions = []
self.response_functions = [templates.functions.root_concern]
self.threshold = threshold
for func in function_groups:
self.trigger_functions.append(func[0])
self.response_functions.extend(func[1])
def __init__(self, root_model_path, fwd_model_file_name, bwd_model_file_name, threshold=0.5):
self.fwd_model_response = load_learner(root_model_path[0], fwd_model_file_name)
self.bwd_model_response = load_learner(root_model_path[0], bwd_model_file_name)
self.fwd_model_trigger = load_learner(root_model_path[1], fwd_model_file_name)
self.bwd_model_trigger = load_learner(root_model_path[1], bwd_model_file_name)
self.ds_response = self.fwd_model_response.data.single_ds.y
self.ds_trigger = self.fwd_model_trigger.data.single_ds.y
self.trigger_functions = []
self.response_functions = [templates.functions.root_concern]
self.threshold = threshold
for func in function_groups:
self.trigger_functions.append(func[0])
self.response_functions.extend(func[1])
def __init__(self, fastaimodelpkl, fp16=False, verbose=False):
"""
:param fastaimodelpkl - full path to the pkl file, e.g. "mystuff/fastaimodel.pkl"
:param fp16 - set it to use half precision
"""
dirname = ntpath.dirname(fastaimodelpkl)
filename = ntpath.basename(fastaimodelpkl)
if fp16:
self.learn = load_learner(dirname, file=filename).to_fp16()
else:
self.learn = load_learner(dirname, file=filename)
self.verbose=verbose
def load(self, path):
if load_learner is None:
raise ImportError("fastai package is required to use "
"bentoml.artifacts.FastaiModelArtifact")
model = load_learner(path, self._file_name)
return self.pack(model)
def setup_learner(url, file_name):
download_file(url, path / file_name)
try:
learner = load_learner(path, file_name)
return learner
except RuntimeError:
logging.exception('Error setting up learner')
def load_model(self, tmp_dir):
"""Load the model in preparation for one or more prediction calls."""
if self.inf_learner is None:
model_uri = self.config['model_uri']
model_path = download_if_needed(model_uri, tmp_dir)
self.inf_learner = load_learner(dirname(model_path),
basename(model_path))
def _get_learner_object(data, layers, emb_szs, ps, emb_drop, pretrained_path):
if pretrained_path:
learn = load_learner(
os.path.dirname(pretrained_path),
os.path.basename(pretrained_path).split('.')[0] + "_exported.pth")
if not data._is_empty:
learn.data = data._databunch
else:
databunch = data._databunch
if not emb_szs or isinstance(emb_szs, dict):
emb_szs = databunch.get_emb_szs({} if not emb_szs else emb_szs)
model = TabularModel(emb_szs,
len(databunch.cont_names),
out_sz=databunch.c,
layers=layers,
ps=ps,
emb_drop=emb_drop,
y_range=None,
use_bn=False)
learn = Learner(databunch,
model,
model_dir=tempfile.TemporaryDirectory().name)
return learn
def __init__(self,
model_path='./models',
od_model='faster-rcnn.pt',
class_model='class_resnet.pkl',
ls_model='ls_resnet.pkl',
gam_model='gam_resnet.pkl',
cutoffs=[1.5, 2.5]):
model_path = Path(model_path)
device = torch.device(
'cuda') if torch.cuda.is_available() else torch.device('cpu')
self.od_model = torch.load(str(model_path / od_model), device)
self.od_model.eval()
self.class_model = load_learner(path=model_path, file=class_model)
self.ls_model = load_learner(path=model_path, file=ls_model)
self.gam_model = load_learner(path=model_path, file=gam_model)
self.cutoffs = cutoffs
def __init__(self,
model_path:PathOrStr,
model_file_name:PathOrStr):
self.learn = load_learner(path=model_path, file=model_file_name)
self.learn.model.eval() # turn off dropout, etc. only need to do this after loading model.
self.encoder = self.learn.model[0]
def upload_file():
"""
retrieve the image uploaded and make sure it is an image file
"""
file = request.files['file']
image_extensions = ['jpg', 'jpeg', 'png']
if file.filename.split('.')[1] not in image_extensions:
return jsonify('Please upload an appropriate image file')
"""
Load the trained model in export.pkl
"""
learn = load_learner(path=".")
"""
Perform prediction
"""
#image_bytes = file.read()
#img = Image.open(io.BytesIO(image_bytes))
img = open_image(file)
pred_class, pred_idx, outputs = learn.predict(img)
i = pred_idx.item()
classes = [
'Domestic Medium Hair', 'Persian', 'Ragdoll', 'Siamese', 'Snowshoe'
]
prediction = classes[i]
return jsonify(f'Your cat is a {prediction}')
def get_pred_new_data_old_model(
valid_df: pd.DataFrame,
path: Path = MODELS_PATH) -> Tuple[Learner, float]:
"""Get a RSMPE score for predictions from the existing best model, with
new data.
Input: a pd.DataFrame for the validation data and the path for the model.
Output: the model ready to save, and the root mean squared percentage error
for the predicted sales. (If this model is second-best, we'll still want
to save it to a different file for record-keeping purposes.)
"""
valid_df = preprocess.preprocess(valid_df)
# Get the right model to load
models = [
file for file in os.listdir(path) if file.startswith('current_best')
]
best_model = sorted(models, reverse=True)[0]
learn = load_learner(path=path,
fname=best_model,
test=TabularList.from_df(valid_df, path=path))
# get log predictions and compare to actual values
log_preds, _ = learn.get_preds(ds_type=DatasetType.Test)
valid_preds = np.exp(np.array(log_preds.flatten()))
valid_reals = valid_df.loc[valid_df.sales != 0, 'sales'].values
new_rmspe = rmspe(valid_preds, valid_reals)
return (learn, new_rmspe)
def get_pred_single_val(data: pd.Series, path: Path) -> float:
"""Get a prediction for a single row of data.
Input: a pd.Series for the data and the path for the model.
Output: the predicted sales for that row of data.
"""
# Get the right model to load
models = [
file for file in os.listdir(path) if file.startswith('current_best')
]
best_model = sorted(models, reverse=True)[0]
# Load the model and get the prediction
#learn = load_learner(path/best_model)
learn = load_learner(path=path, fname=best_model)
log_pred_tens, _, _ = learn.predict(data)
# The model returns a tensor (Float [x]) so we need to get x
log_pred = log_pred_tens.data[0]
# Also it gives log predictions, so we need to exp it
prediction = math.exp(log_pred)
return prediction
def get_prediction_data(image_filepath, model_path):
learn = basic_train.load_learner(model_path)
image_for_fastai = vision.image.open_image(image_filepath)
prediction = learn.predict(image_for_fastai)
predicted_class = str(prediction[0]).replace('_', ' ')
predicted_class_index = prediction[1].item()
confidence = int(100*prediction[2][predicted_class_index].item())
return predicted_class, confidence
def get_classified_prob(img):
learn: Learner = load_learner(path=os.path.join(
settings.BASE_DIR, 'vandydj/ai_models/'),
file='model_v1.pkl')
prediction = learn.predict(img)
label = prediction[0]
prob = max(prediction[2])
return prob, label
def load_model(self, tmp_dir):
"""Load the model in preparation for one or more prediction calls."""
if self.inf_learner is None:
self.log_options()
model_uri = self.backend_opts.model_uri
model_path = download_if_needed(model_uri, tmp_dir)
self.inf_learner = load_learner(
dirname(model_path), basename(model_path))
def load_model() -> None:
# Define model
global model
try:
model = load_learner(path='../../models/', file='wm_remove.pkl')
model.model.eval()
except:
print('Error Loading model, check model')
def __init__(self, model_folder_name:
'(str) locate to the folder dont specify in any extension',
model_name: '(str) the model name in .pkl extension',
min_char_len:
'(int) ignore input if its length is less than this number'):
self.learn = load_learner(model_folder_name, model_name)
self.classes = self.learn.data.train_ds.classes
self.min_char_len = min_char_len
def __init__(self, bbz_path, model_name="bnet20191028"):
self._bbz_path = bbz_path
model_path = bbz_path / "models" / model_name
from fastai.basic_train import load_learner
# see https://docs.fast.ai/tutorial.inference.html
self._learner = load_learner(model_path)
def load(cls, path: str, reset_paths=True, verbose=True):
from fastai.basic_train import load_learner
model = super().load(path, reset_paths=reset_paths, verbose=verbose)
model.model = load_pkl.load_with_fn(
f'{model.path}{model.model_internals_file_name}',
lambda p: load_learner(model.path, p),
verbose=verbose)
return model
def load_fastai():
modelpath = Path(__file__).parent
modelpath = modelpath / 'trained'
logging.info('load model path {0}'.format(os.getcwd()))
defaults.device = device('cpu')
learn = load_learner(modelpath)
learn.load('stage-2')
return learn
def load_model(self, tmp_dir):
"""Load the model in preparation for one or more prediction calls."""
if self.inf_learner is None:
self.print_options()
model_uri = self.backend_opts.model_uri
model_path = download_if_needed(model_uri, tmp_dir)
self.inf_learner = load_learner(
dirname(model_path), basename(model_path))
self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
def load(self, path):
try:
# We need matplotlib and torch for fastai. Make sure we install those.
from fastai.basic_train import load_learner
except ImportError:
raise ImportError(
'fastai package is required to use FastaiModelArtifact')
model = load_learner(path, self._file_name)
return self.pack(model)
def test_save_models(self, mock_Learner_export, mock_Learner_save):
"""save_models should save the second-best model to an appropriate
file, and the best model to the current best model file.
"""
# TODO: this test was edited due to some changes - this test can and
# should be rewritten to be more specific (ie useful).
# fake winner/loser models and the names they should be saved with
winner = load_learner(self.model_path, fname='current_best')
# fname below could be replaced by any other not-as-good model
# Even better would be to replace this with a pre-generated model as
# a pytest funcarg
loser = load_learner(path=self.model_path, fname='current_best')
# Call the function
train_model.save_models(winner, loser)
# Assertions
mock_Learner_export.assert_called()
def _load_learner(export_file_name, label):
try:
from fastai.basic_train import load_learner
learn = load_learner('notebooks', export_file_name)
print(f'INFO: running with {label}')
return learn
except ModuleNotFoundError as ex:
print(f'INFO: running without {label} as fastai not found')
except FileNotFoundError as ex:
print(f'INFO: running without {label} as {export_file_name} not found')
def load(cls, path: str, file_prefix='', reset_paths=False, verbose=True):
from fastai.basic_train import load_learner
obj = super().load(path,
file_prefix=file_prefix,
reset_paths=reset_paths,
verbose=verbose)
obj.model = load_pkl.load_with_fn(
f'{obj.path}{obj.model_internals_file_name}',
lambda p: load_learner(obj.path, p),
verbose=verbose)
return obj
def __init__(self, model_path: PathOrStr, model_file_name: PathOrStr):
"""Load the Learner object from model_path/model_file_name.
Args:
model_path: The path (directory) of the Learner object.
e.g., ./model_files
model_file_name: The file name of the Learner object.
e.g., model.pkl
"""
self.learn = load_learner(path=model_path, file=model_file_name)
self.learn.model.eval(
) # turn off dropout, etc. only need to do this after loading model.
self.encoder = self.learn.model[0]
async def setup_learner(url, dest):
await download_file(url, path / dest)
try:
learn = load_learner(path, dest)
return learn
except RuntimeError as e:
if len(e.args) > 0 and 'CPU-only machine' in e.args[0]:
print(e)
message = "\n\nThis model was trained with an old version of fastai and will not work in a CPU environment.\n\nPlease update the fastai library in your training environment and export your model again.\n\nSee instructions for 'Returning to work' at https://course.fast.ai."
raise RuntimeError(message)
else:
raise
def __init__(self, tokenizer="fastai"):
"""Initalize the ULMFiT model for script generation
Keyword Arguments:
tokenizer {str} -- Text tokenizer to use for script generation (default: {"fastai"})
1. fastai
2. sentencepiece
"""
self.tokenizer = tokenizer
self.checkModelFile()
logging.info("Model file present!")
self.predictor = load_learner(self.modelFolder, self.modelFile)
async def setup_learner():
await download_file(export_file_url, path / export_file_name)
try:
learn = load_learner(path, export_file_name)
return learn
except RuntimeError as e:
if len(e.args) > 0 and 'CPU-only machine' in e.args[0]:
print(e)
message = "ERROR: This model was trained with an old version of fastai, UPDATE IT!"
raise RuntimeError(message)
else:
raise