xdobro23/YOLOv8-16bit
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

ultralytics 8.0.49 task, exports and metadata updates (#1197)

Browse Source 
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Mehran Ghandehari <mehran.maps@gmail.com>
Co-authored-by: Paul Guerrie <97041392+paulguerrie@users.noreply.github.com>
This commit is contained in:
Glenn Jocher
2023-03-01 21:16:09 -08:00
committed by GitHub
parent 74e4c94806
commit 3861e6c82a
20 changed files with 111 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
ultralytics/nn/autobackend.py
View File

@ -75,7 +75,7 @@ class AutoBackend(nn.Module):
fp16 &= pt or jit or onnx or engine or nn_module # FP16
nhwc = coreml or saved_model or pb or tflite or edgetpu # BHWC formats (vs torch BCWH)
stride = 32 # default stride
model = None # TODO: resolves ONNX inference, verify effect on other backends
model, metadata = None, None
cuda = torch.cuda.is_available() and device.type != 'cpu' # use CUDA
if not (pt or triton or nn_module):
w = attempt_download_asset(w) # download if not local
@ -105,10 +105,7 @@ class AutoBackend(nn.Module):
model = torch.jit.load(w, _extra_files=extra_files, map_location=device)
model.half() if fp16 else model.float()
if extra_files['config.txt']: # load metadata dict
d = json.loads(extra_files['config.txt'],
object_hook=lambda d: {int(k) if k.isdigit() else k: v
for k, v in d.items()})
stride, names = int(d['stride']), d['names']
metadata = json.loads(extra_files['config.txt'], object_hook=lambda x: dict(x.items()))
elif dnn: # ONNX OpenCV DNN
LOGGER.info(f'Loading {w} for ONNX OpenCV DNN inference...')
check_requirements('opencv-python>=4.5.4')
@ -120,23 +117,23 @@ class AutoBackend(nn.Module):
providers = ['CUDAExecutionProvider', 'CPUExecutionProvider'] if cuda else ['CPUExecutionProvider']
session = onnxruntime.InferenceSession(w, providers=providers)
output_names = [x.name for x in session.get_outputs()]
meta = session.get_modelmeta().custom_metadata_map # metadata
if 'stride' in meta:
stride, names = int(meta['stride']), eval(meta['names'])
metadata = session.get_modelmeta().custom_metadata_map # metadata
elif xml: # OpenVINO
LOGGER.info(f'Loading {w} for OpenVINO inference...')
check_requirements('openvino') # requires openvino-dev: https://pypi.org/project/openvino-dev/
from openvino.runtime import Core, Layout, get_batch # noqa
ie = Core()
if not Path(w).is_file(): # if not *.xml
w = next(Path(w).glob('*.xml')) # get *.xml file from *_openvino_model dir
network = ie.read_model(model=w, weights=Path(w).with_suffix('.bin'))
w = Path(w)
if not w.is_file(): # if not *.xml
w = next(w.glob('*.xml')) # get *.xml file from *_openvino_model dir
network = ie.read_model(model=str(w), weights=w.with_suffix('.bin'))
if network.get_parameters()[0].get_layout().empty:
network.get_parameters()[0].set_layout(Layout('NCHW'))
batch_dim = get_batch(network)
if batch_dim.is_static:
batch_size = batch_dim.get_length()
executable_network = ie.compile_model(network, device_name='CPU') # device_name="MYRIAD" for NCS2
metadata = w.parent / 'metadata.yaml'
elif engine: # TensorRT
LOGGER.info(f'Loading {w} for TensorRT inference...')
import tensorrt as trt # https://developer.nvidia.com/nvidia-tensorrt-download
@ -148,7 +145,7 @@ class AutoBackend(nn.Module):
# Read file
with open(w, 'rb') as f, trt.Runtime(logger) as runtime:
meta_len = int.from_bytes(f.read(4), byteorder='little') # read metadata length
meta = json.loads(f.read(meta_len).decode('utf-8')) # read metadata
metadata = json.loads(f.read(meta_len).decode('utf-8')) # read metadata
model = runtime.deserialize_cuda_engine(f.read()) # read engine
context = model.create_execution_context()
bindings = OrderedDict()
@ -171,18 +168,17 @@ class AutoBackend(nn.Module):
bindings[name] = Binding(name, dtype, shape, im, int(im.data_ptr()))
binding_addrs = OrderedDict((n, d.ptr) for n, d in bindings.items())
batch_size = bindings['images'].shape[0] # if dynamic, this is instead max batch size
stride, names = int(meta['stride']), meta['names']
elif coreml: # CoreML
LOGGER.info(f'Loading {w} for CoreML inference...')
import coremltools as ct
model = ct.models.MLModel(w)
names, stride, task = (model.user_defined_metadata.get(k) for k in ('names', 'stride', 'task'))
names, stride = eval(names), int(stride)
metadata = model.user_defined_metadata
elif saved_model: # TF SavedModel
LOGGER.info(f'Loading {w} for TensorFlow SavedModel inference...')
import tensorflow as tf
keras = False # assume TF1 saved_model
model = tf.keras.models.load_model(w) if keras else tf.saved_model.load(w)
metadata = Path(w) / 'metadata.yaml'
elif pb: # GraphDef https://www.tensorflow.org/guide/migrate#a_graphpb_or_graphpbtxt
LOGGER.info(f'Loading {w} for TensorFlow GraphDef inference...')
import tensorflow as tf
@ -221,23 +217,23 @@ class AutoBackend(nn.Module):
with contextlib.suppress(zipfile.BadZipFile):
with zipfile.ZipFile(w, 'r') as model:
meta_file = model.namelist()[0]
meta = ast.literal_eval(model.read(meta_file).decode('utf-8'))
stride, names = int(meta['stride']), meta['names']
metadata = ast.literal_eval(model.read(meta_file).decode('utf-8'))
elif tfjs: # TF.js
raise NotImplementedError('YOLOv8 TF.js inference is not supported')
elif paddle: # PaddlePaddle
LOGGER.info(f'Loading {w} for PaddlePaddle inference...')
check_requirements('paddlepaddle-gpu' if cuda else 'paddlepaddle')
import paddle.inference as pdi
if not Path(w).is_file(): # if not *.pdmodel
w = next(Path(w).rglob('*.pdmodel')) # get *.pdmodel file from *_paddle_model dir
weights = Path(w).with_suffix('.pdiparams')
config = pdi.Config(str(w), str(weights))
w = Path(w)
if not w.is_file(): # if not *.pdmodel
w = next(w.rglob('*.pdmodel')) # get *.pdmodel file from *_paddle_model dir
config = pdi.Config(str(w), str(w.with_suffix('.pdiparams')))
if cuda:
config.enable_use_gpu(memory_pool_init_size_mb=2048, device_id=0)
predictor = pdi.create_predictor(config)
input_handle = predictor.get_input_handle(predictor.get_input_names()[0])
output_names = predictor.get_output_names()
metadata = w.parents[1] / 'metadata.yaml'
elif triton: # NVIDIA Triton Inference Server
LOGGER.info('Triton Inference Server not supported...')
'''
@ -254,14 +250,16 @@ class AutoBackend(nn.Module):
f'\n\n{EXPORT_FORMATS_TABLE}')
# Load external metadata YAML
w = Path(w)
if xml or saved_model or paddle:
metadata = (w if saved_model else w.parents[1] if paddle else w.parent) / 'metadata.yaml'
if metadata.exists():
metadata = yaml_load(metadata)
stride, names = int(metadata['stride']), metadata['names'] # load metadata
else:
LOGGER.warning(f"WARNING ⚠️ Metadata not found at '{metadata}'")
if isinstance(metadata, (str, Path)) and Path(metadata).exists():
metadata = yaml_load(metadata)
if metadata:
stride = int(metadata['stride'])
task = metadata['task']
batch = int(metadata['batch'])
imgsz = eval(metadata['imgsz']) if isinstance(metadata['imgsz'], str) else metadata['imgsz']
names = eval(metadata['names']) if isinstance(metadata['names'], str) else metadata['names']
elif not (pt or triton or nn_module):
LOGGER.warning(f"WARNING ⚠️ Metadata not found for 'model={weights}'")
# Check names
if 'names' not in locals(): # names missing

4
ultralytics/nn/tasks.py
View File

@ -257,7 +257,7 @@ class ClassificationModel(BaseModel):
cfg=None,
model=None,
ch=3,
nc=1000,
nc=None,
cutoff=10,
verbose=True): # yaml, model, channels, number of classes, cutoff index, verbose flag
super().__init__()
@ -286,6 +286,8 @@ class ClassificationModel(BaseModel):
if nc and nc != self.yaml['nc']:
LOGGER.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}")
self.yaml['nc'] = nc # override yaml value
elif not nc and not self.yaml.get('nc', None):
raise ValueError('nc not specified. Must specify nc in model.yaml or function arguments.')
self.model, self.save = parse_model(deepcopy(self.yaml), ch=ch, verbose=verbose) # model, savelist
self.stride = torch.Tensor([1]) # no stride constraints
self.names = {i: f'{i}' for i in range(self.yaml['nc'])} # default names dict