def create_model(session, batch_size):
"""
Create model and initialize it or load its parameters in a session
Args
session: tensorflow session
batch_size: integer. Number of examples in each batch
Returns
model: The created (or loaded) model
Raises
ValueError if asked to load a model, but the checkpoint specified by
FLAGS.load cannot be found.
"""
model = linear_model.LinearModel(FLAGS.linear_size,
FLAGS.num_layers,
FLAGS.residual,
FLAGS.batch_norm,
FLAGS.max_norm,
batch_size,
FLAGS.learning_rate,
summaries_dir,
dtype=tf.float32)
print("\n\nCreating model with fresh parameters.")
session.run(tf.global_variables_initializer())
return model
def load_model(session, batch_size):
model = linear_model.LinearModel(
FLAGS.linear_size,
FLAGS.num_layers,
FLAGS.residual,
FLAGS.batch_norm,
FLAGS.max_norm,
batch_size,
FLAGS.learning_rate,
summaries_dir,
FLAGS.predict_14,
dtype=tf.float16 if FLAGS.use_fp16 else tf.float32)
# Load a previously saved model
ckpt = tf.train.get_checkpoint_state( train_dir, latest_filename="checkpoint")
print( "train_dir", train_dir )
if ckpt and ckpt.model_checkpoint_path:
# Check if the specific checkpoint exists
if os.path.isfile(os.path.join(train_dir,"checkpoint-{0}.index".format(FLAGS.load))):
ckpt_name = os.path.join( os.path.join(train_dir,"checkpoint-{0}".format(FLAGS.load)) )
else:
raise ValueError("Asked to load checkpoint {0}, but it does not seem to exist".format(FLAGS.load))
print("Loading model {0}".format( ckpt_name ))
model.saver.restore( session, ckpt.model_checkpoint_path )
return model
else:
print("Could not find checkpoint. Aborting.")
raise( ValueError, "Checkpoint {0} does not seem to exist".format( ckpt.model_checkpoint_path ) )
return model
def create_model(session, batch_size):
"""
Loads model in a session
Args
session: tensorflow session
batch_size: integer. Number of examples in each batch
Returns
model: The loaded model
"""
model = linear_model.LinearModel(FLAGS.linear_size,
FLAGS.num_layers,
FLAGS.residual,
FLAGS.batch_norm,
FLAGS.max_norm,
batch_size,
dtype=tf.float32)
to_restore = tf.train.latest_checkpoint(summaries_dir)
print("\nRestoring model from {}\n".format(to_restore))
model.saver.restore(session,
to_restore) # restore model from last checkpoint
return model
def create_model( session, batch_size ):
"""
Create model and initialize it or load its parameters in a session
Args
session: tensorflow session
actions: list of string. Actions to train/test on
batch_size: integer. Number of examples in each batch
Returns
model: The created (or loaded) model
Raises
ValueError if asked to load a model, but the checkpoint specified by
FLAGS.load cannot be found.
"""
model = linear_model.LinearModel(
FLAGS.linear_size,
FLAGS.num_layers,
FLAGS.residual,
FLAGS.batch_norm,
FLAGS.max_norm,
batch_size,
FLAGS.learning_rate,
summaries_dir,
FLAGS.predict_14,
dtype=tf.float16 if FLAGS.use_fp16 else tf.float32)
if FLAGS.load <= 0:
# Create a new model from scratch
print("Creating model with fresh parameters.")
session.run( tf.global_variables_initializer() )
return model
# Load a previously saved model
ckpt = tf.train.get_checkpoint_state( train_dir, latest_filename="checkpoint")
print( "train_dir", train_dir )
if ckpt and ckpt.model_checkpoint_path:
# Check if the specific checkpoint exists
if FLAGS.load > 0:
if os.path.isfile(os.path.join(train_dir,"checkpoint-{0}.index".format(FLAGS.load))):
ckpt_name = os.path.join( os.path.join(train_dir,"checkpoint-{0}".format(FLAGS.load)) )
else:
raise ValueError("Asked to load checkpoint {0}, but it does not seem to exist".format(FLAGS.load))
else:
ckpt_name = os.path.basename( ckpt.model_checkpoint_path )
print("Loading model {0}".format( ckpt_name ))
model.saver.restore( session, ckpt.model_checkpoint_path )
return model
else:
print("Could not find checkpoint. Aborting.")
raise( ValueError, "Checkpoint {0} does not seem to exist".format( ckpt.model_checkpoint_path ) )
return model
def create_model(session, actions, batch_size):
model = linear_model.LinearModel(
FLAGS.linear_size,
FLAGS.num_layers,
FLAGS.residual,
FLAGS.batch_norm,
FLAGS.max_norm,
batch_size,
FLAGS.learning_rate,
summaries_dir,
FLAGS.predict_14,
dtype=tf.float16 if FLAGS.use_fp16 else tf.float32)
if FLAGS.load <= 0:
# Create a new model from scratch
print("Creating model with fresh parameters.")
session.run(tf.global_variables_initializer())
return model
# Load a previously saved model
ckpt = tf.train.get_checkpoint_state(train_dir,
latest_filename="checkpoint")
print("train_dir", train_dir)
if ckpt and ckpt.model_checkpoint_path:
# Check if the specific checkpoint exists
if FLAGS.load > 0:
if os.path.isfile(
os.path.join(train_dir,
"checkpoint-{0}.index".format(FLAGS.load))):
ckpt_name = os.path.join(
os.path.join(train_dir,
"checkpoint-{0}".format(FLAGS.load)))
else:
raise ValueError(
"Asked to load checkpoint {0}, but it does not seem to exist"
.format(FLAGS.load))
else:
ckpt_name = os.path.basename(ckpt.model_checkpoint_path)
print("Loading model {0}".format(ckpt_name))
init_g = tf.global_variables_initializer()
init_l = tf.local_variables_initializer()
with tf.Session() as sess:
session.run(init_g)
session.run(init_l)
model.saver.restore(session, ckpt.model_checkpoint_path)
return model
else:
print("Could not find checkpoint. Aborting.")
raise (ValueError, "Checkpoint {0} does not seem to exist".format(
ckpt.model_checkpoint_path))
return model
ax1 = fig.add_subplot(2, 2, 1, projection='3d')
ax1.set_title("System Transition Function")
viz_1_1d.plot_1_1d_system(system, ax=ax1)
# Collect data by applying random actions in the system
states, actions, rewards = system.sim_traj(random_control)
states_prime = states[1:]
states = states[:-1]
# Matching expected format for model.fit()
data = [(np.array([states[i]]), np.array([actions[i]]), rewards[i],
np.array([states_prime[i]]), False) for i in range(len(states))]
print("Fitting on {} datapoints".format(len(data)))
# Fit linear model to data
model = linear_model.LinearModel(1, 1)
model.fit(data)
model.limits_low = np.array([system.state_limits[0]])
model.limits_high = np.array([system.state_limits[1]])
print("Model is A={}, B={} (not fitting reward function now)\n".format(
model.A, model.B))
ax2 = fig.add_subplot(2, 2, 2, projection='3d')
ax2.set_title("Linear Model of Transition Function")
viz_1_1d.plot_1_1d_model(model,
ax=ax2,
action_limit_low=system.action_limits[0],
action_limit_high=system.action_limits[1])
# Compute optimal controller for fit model
system.reset()
def create_model(session, actions,
batch_size): #主要分为两部分：要么生成模型，要么加载之前的checkpoint
"""
Create model and initialize it or load its parameters in a session
Args
session: tensorflow session
actions: list of string. Actions to train/test on
batch_size: integer. Number of examples in each batch
Returns
model: The created (or loaded) model
Raises
ValueError if asked to load a model, but the checkpoint specified by
FLAGS.load cannot be found.
"""
model = linear_model.LinearModel(
FLAGS.linear_size,
FLAGS.num_layers,
FLAGS.residual,
FLAGS.batch_norm,
FLAGS.max_norm,
batch_size,
FLAGS.learning_rate,
summaries_dir,
FLAGS.predict_14,
dtype=tf.float16 if FLAGS.use_fp16 else tf.float32)
if FLAGS.load <= 0:
# Create a new model from scratch
print("Creating model with fresh parameters.")
session.run(
tf.global_variables_initializer()
) #session.run使用已定义好的会话来计算关心的结果 在TensorFlow中一个变量的值在被明确调用前，这个变量的初始化过程需要被明确地调用。
return model
# Load a previously saved model
ckpt = tf.train.get_checkpoint_state(
train_dir, latest_filename="checkpoint") #experiment中的checkpoint
print("train_dir", train_dir)
##########################下面都是用来checkpoint的#############################
if ckpt and ckpt.model_checkpoint_path:
# Check if the specific checkpoint exists
if FLAGS.load > 0:
if os.path.isfile(
os.path.join(train_dir,
"checkpoint-{0}.index".format(FLAGS.load))
): #train_dir+'checkpoint-1.index' 即experiment.checkpoint-24371.index是否为文件
ckpt_name = os.path.join(
os.path.join(train_dir, "checkpoint-{0}".format(
FLAGS.load))) # experiments\checkpoint-24371
else:
raise ValueError(
"Asked to load checkpoint {0}, but it does not seem to exist"
.format(FLAGS.load))
else:
ckpt_name = os.path.basename(
ckpt.model_checkpoint_path) #返回path最后的文件名
print("Loading model {0}".format(ckpt_name))
model.saver.restore(session,
ckpt.model_checkpoint_path) #之前保存的结果会被加载到当前的对话中
return model #model是个类
else:
print("Could not find checkpoint. Aborting.")
raise (ValueError, "Checkpoint {0} does not seem to exist".format(
ckpt.model_checkpoint_path))
##############################################################################
return model
