def on_received_observations(self, context: Context,
data: DB20Observations):
logger.info("received", data=data)
camera: JPGImage = data.camera
odometry = data.odometry
print(odometry)
_rgb = jpg2rgb(camera.jpg_data)
def check_tensorflow_gpu(self):
import tensorflow as tf
req = os.environ.get('AIDO_REQUIRE_GPU', None)
name = tf.test.gpu_device_name()
logger.info(f'gpu_device_name: {name!r} AIDO_REQUIRE_GPU = {req!r}')
if req is not None:
if not name: # None or ''
msg = 'Could not find gpu device.'
logger.error(msg)
def check_tensorflow_gpu(self):
import tensorflow as tf
req = os.environ.get('AIDO_REQUIRE_GPU', None)
force_cpu = os.environ.get('FORCE_CPU_INFERENCE', None)
if force_cpu:
return
name = tf.test.gpu_device_name()
logger.info(f'gpu_device_name: {name!r} AIDO_REQUIRE_GPU = {req!r}')
if req is not None:
if not name: # None or ''
msg = 'Could not find gpu device.'
logger.error(msg)
raise RuntimeError(msg)
def limit_gpu_memory(memory_limit=1024):
""" Restricts TensorFlow to only allocated 1GB of memory on the first GPU"""
import tensorflow as tf
physical_gpus = tf.config.experimental.list_physical_devices('GPU')
if physical_gpus:
try:
c = [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=memory_limit)]
tf.config.experimental.set_virtual_device_configuration(physical_gpus[0], c)
logical_gpus = tf.config.experimental.list_logical_devices('GPU')
logger.info(num_physical_gpus=len(physical_gpus), num_logical_gpus=len(logical_gpus))
except RuntimeError as e:
# Virtual devices must be set before GPUs have been initialized
logger.error(e)
def __init__(self, load_model=False, model_path=None):
logger.info('PytorchAgent init')
self.preprocessor = DTPytorchWrapper()
self.model = Model()
self.current_image = np.zeros((640, 480, 3))
self.steering_to_wheel_wrapper = SteeringToWheelVelWrapper()
self.controller = Controller()
self.dt = None
self.last_t = None
self.old_obs = None
logger.info('PytorchAgent init complete')
def __init__(self, load_model=False, model_path=None):
logger.info('PytorchRLTemplateAgent init')
self.preprocessor = DTPytorchWrapper()
self.image_size = (120,160, 3)
self.wrapper = SteeringToWheelVelWrapper()
self.model = Dronet()
self._device = self.model._device
self.model.to(self._device)
self.current_image = np.zeros((3,self.image_size[0],self.image_size[1]))
if load_model:
logger.info('PytorchRLTemplateAgent loading models')
fp = model_path if model_path else "model.pt"
self.model.load(fp, "models", for_inference=True)
logger.info('PytorchRLTemplateAgent model loaded')
logger.info('PytorchRLTemplateAgent init complete')
def __init__(self, load_model=False, model_path=None):
logger.info('PytorchRLTemplateAgent init')
self.preprocessor = DTPytorchWrapper()
self.model = DDPG(state_dim=self.preprocessor.shape,
action_dim=2,
max_action=1,
net_type="cnn")
self.current_image = np.zeros((640, 480, 3))
if load_model:
logger.info('PytorchRLTemplateAgent loading models')
fp = model_path if model_path else "model"
self.model.load(fp, "models", for_inference=True)
logger.info('PytorchRLTemplateAgent init complete')
def init(self, context: Context):
self.check_gpu_available(context)
logger.info("PytorchRLTemplateAgent init")
from model import DDPG
self.preprocessor = DTPytorchWrapper()
self.model = DDPG(state_dim=self.preprocessor.shape,
action_dim=2,
max_action=1,
net_type="cnn")
self.current_image = np.zeros((640, 480, 3))
if self.load_model:
logger.info("Pytorch Template Agent loading models")
fp = self.model_path if self.model_path else "model"
self.model.load(fp, "models", for_inference=True)
logger.info("PytorchRLTemplateAgent init complete")
def on_received_episode_start(self, context: Context, data: EpisodeStart):
context.info(f'Starting episode "{data.episode_name}"')
logger.info(data=data)
def on_received_observations(self, context: Context, data: DB20Observations):
logger.info("received", data=data)
camera: JPGImage = data.camera
self.rgb = dcu.bgr_from_rgb(dcu.bgr_from_jpg(camera.jpg_data))
