def on_pubmsg(self, serv, event):
if event.arguments[0][0] == self.cmdprefix:
# it is a command
cmd, args = unpack(event.arguments[0][1:])
print(event.source.nick, 'launched the', cmd, 'command, with the following arguments:', args)
if cmd in fantasy.binding.keys():
print('command exists!')
# calling the function
ret = fantasy.binding[cmd](serv, self, event, args)
if ret is not None:
for i in ret.split('\n'):
for j in split_len(i.encode(), 470):
serv.privmsg(event.target, j.decode())
if event.arguments[0][0] == self.adminprefix:
# it is an admin command
cmd, args = unpack(event.arguments[0][1:])
print(event.source.nick, 'launched the', cmd, 'admin command, with the following arguments:', args)
if cmd in admin.binding.keys():
print('command exists!')
# calling the function
ret = admin.binding[cmd](serv, self, event, args)
if ret is not None:
for i in ret.split('\n'):
for j in split_len(i.encode(), 470):
serv.privmsg(event.target, j.decode())
def processRawData(path):
log.debug('processing raw data %s'%(path,))
xml_dir = os.path.join(util.getTempDir(), config.options['data']['raw-xml-dir'])
util.assureDirExist(xml_dir)
base = os.path.basename(path)
xml_path = os.path.join(xml_dir, base[:-3])
util.unpack(path, xml_path)
return load_xml(xml_path)
def read_cache(stream):
"""Read a cache file from the given stream
:return: tuple(version, entries_dict, extension_data, content_sha)
* version is the integer version number
* entries dict is a dictionary which maps IndexEntry instances to a path
at a stage
* extension_data is '' or 4 bytes of type + 4 bytes of size + size bytes
* content_sha is a 20 byte sha on all cache file contents"""
version, num_entries = read_header(stream)
count = 0
entries = dict()
read = stream.read
tell = stream.tell
while count < num_entries:
beginoffset = tell()
ctime = unpack(">8s", read(8))[0]
mtime = unpack(">8s", read(8))[0]
(dev, ino, mode, uid, gid, size, sha, flags) = \
unpack(">LLLLLL20sH", read(20 + 4 * 6 + 2))
path_size = flags & CE_NAMEMASK
path = read(path_size)
real_size = ((tell() - beginoffset + 8) & ~7)
data = read((beginoffset + real_size) - tell())
entry = IndexEntry(
(mode, sha, flags, path, ctime, mtime, dev, ino, uid, gid, size))
# entry_key would be the method to use, but we safe the effort
entries[(path, entry.stage)] = entry
count += 1
# END for each entry
# the footer contains extension data and a sha on the content so far
# Keep the extension footer,and verify we have a sha in the end
# Extension data format is:
# 4 bytes ID
# 4 bytes length of chunk
# repeated 0 - N times
extension_data = stream.read(~0)
assert len(
extension_data
) > 19, "Index Footer was not at least a sha on content as it was only %i bytes in size" % len(
extension_data)
content_sha = extension_data[-20:]
# truncate the sha in the end as we will dynamically create it anyway
extension_data = extension_data[:-20]
return (version, entries, extension_data, content_sha)
def read_entry(stream):
"""Return: One entry of the given stream"""
beginoffset = stream.tell()
read = stream.read
ctime = unpack(">8s", read(8))[0]
mtime = unpack(">8s", read(8))[0]
(dev, ino, mode, uid, gid, size, sha, flags) = \
unpack(">LLLLLL20sH", read(20 + 4 * 6 + 2))
path_size = flags & CE_NAMEMASK
path = read(path_size)
real_size = ((stream.tell() - beginoffset + 8) & ~7)
data = read((beginoffset + real_size) - stream.tell())
return IndexEntry((mode, sha, flags, path, ctime, mtime, dev, ino, uid, gid, size))
def read_entry(stream):
"""Return: One entry of the given stream"""
beginoffset = stream.tell()
read = stream.read
ctime = unpack(">8s", read(8))[0]
mtime = unpack(">8s", read(8))[0]
(dev, ino, mode, uid, gid, size, sha, flags) = \
unpack(">LLLLLL20sH", read(20 + 4 * 6 + 2))
path_size = flags & CE_NAMEMASK
path = read(path_size)
real_size = ((stream.tell() - beginoffset + 8) & ~7)
data = read((beginoffset + real_size) - stream.tell())
return IndexEntry(
(mode, sha, flags, path, ctime, mtime, dev, ino, uid, gid, size))
def read_cache(stream):
"""Read a cache file from the given stream
:return: tuple(version, entries_dict, extension_data, content_sha)
* version is the integer version number
* entries dict is a dictionary which maps IndexEntry instances to a path
at a stage
* extension_data is '' or 4 bytes of type + 4 bytes of size + size bytes
* content_sha is a 20 byte sha on all cache file contents"""
version, num_entries = read_header(stream)
count = 0
entries = dict()
read = stream.read
tell = stream.tell
while count < num_entries:
beginoffset = tell()
ctime = unpack(">8s", read(8))[0]
mtime = unpack(">8s", read(8))[0]
(dev, ino, mode, uid, gid, size, sha, flags) = \
unpack(">LLLLLL20sH", read(20 + 4 * 6 + 2))
path_size = flags & CE_NAMEMASK
path = read(path_size)
real_size = ((tell() - beginoffset + 8) & ~7)
data = read((beginoffset + real_size) - tell())
entry = IndexEntry((mode, sha, flags, path, ctime, mtime, dev, ino, uid, gid, size))
# entry_key would be the method to use, but we safe the effort
entries[(path, entry.stage)] = entry
count += 1
# END for each entry
# the footer contains extension data and a sha on the content so far
# Keep the extension footer,and verify we have a sha in the end
# Extension data format is:
# 4 bytes ID
# 4 bytes length of chunk
# repeated 0 - N times
extension_data = stream.read(~0)
assert len(
extension_data) > 19, "Index Footer was not at least a sha on content as it was only %i bytes in size" % len(
extension_data)
content_sha = extension_data[-20:]
# truncate the sha in the end as we will dynamically create it anyway
extension_data = extension_data[:-20]
return (version, entries, extension_data, content_sha)
def _learn(self, idx):
t_states, t_actions, t_qvals = unpack(self._memory,
self._model,
self._gamma,
self._unfolding_steps,
"cuda")
self._optimizer.zero_grad()
t_logits, t_values = self._model(t_states)
t_log_probs = t_logits.log_softmax(dim=1)
t_probs = t_logits.softmax(dim=1)
# Compute the value loss
value_loss = F.mse_loss(t_values.squeeze(-1), t_qvals)
# Compute the policy loss
t_advantages = t_qvals - t_values.detach()
policy_loss = -(t_advantages * t_log_probs[range(self._batch_size), t_actions]).mean()
# Compute the entropy and record the original probabilities for later
entropy = -(t_probs * t_log_probs).sum(dim=1).mean()
old_probs = t_probs
(policy_loss + value_loss - self._beta * entropy).backward()
nn_utils.clip_grad_norm_(self._model.parameters(), self._clip_grad)
self._optimizer.step()
# Compute KL divergence
new_probs = self._model(t_states)[0].softmax(dim=1)
kl_divergence = -((new_probs / old_probs).log() * old_probs).sum(dim=1).mean()
self._memory.clear()
# Plot
self._plotter.add_scalar("Combined Loss", (policy_loss + value_loss - self._beta * entropy).item(), idx)
self._plotter.add_scalar("Entropy", entropy.item(), idx)
self._plotter.add_scalar("KL Divergence", kl_divergence.item(), idx)
def print_line (self,start_time=None): t = (self.timestamp - start_time) if start_time and self.timestamp else self.timestamp if not self.removed: coords = unpack(self.coords) coords = special_coords[coords] if coords in special_coords else "(%s, %s)" % coords return "found %s at %s. (%s)" % (self.pickup().name, coords, t) else: return "lost %s! (%s)" % (self.pickup().name, t)
def run(self):
while self.flag and self._flag:
try:
data, _ = self.__socket.recvfrom(65565)
except Exception as err:
util.d(err)
else:
self.__catalogue.touch(util.unpack(data))
def unpack(self, mda_name):
if util.unpack(self.config, mda_name):
self.reload_config()
else:
return False
post_unpack_hook = self.config.get_post_unpack_hook()
if post_unpack_hook:
util.run_post_unpack_hook(post_unpack_hook)
def read_header(stream):
"""Return tuple(version_long, num_entries) from the given stream"""
type_id = stream.read(4)
if type_id != "DIRC":
raise AssertionError("Invalid index file header: %r" % type_id)
version, num_entries = unpack(">LL", stream.read(4 * 2))
# TODO: handle version 3: extended data, see read-cache.c
assert version in (1, 2)
return version, num_entries
def _learn(self, step_idx):
# Extract the states, actions and values as tensors from the memory
t_states, t_actions, t_qvals = unpack(self._memory,
self._model,
self._gamma,
self._unfolding_steps,
"cpu")
# Clear the old gradients
self._optimizer.zero_grad()
# Convert the network output into actual probabilities and their logs for later use
t_logits, t_values = self._model(t_states)
t_log_probs = t_logits.log_softmax(dim=1)
t_probs = t_logits.softmax(dim=1)
# Compute the value loss
value_loss = F.mse_loss(t_values.squeeze(-1), t_qvals)
# Compute the policy loss
t_advantages = t_qvals - t_values.detach()
policy_loss = -(t_advantages * t_log_probs[range(self._batch_size), t_actions]).mean()
# Compute the entropy and record the original probabilities for later
entropy = -(t_probs * t_log_probs).sum(dim=1).mean()
old_probs = t_probs
# Put the complete expression for the loss together and backpropagate the gradients
(policy_loss + value_loss - self._beta * entropy).backward()
# nn_utils.clip_grad_norm_(self._model.parameters(), self._clip_grad)
self._optimizer.step()
# Compute KL divergence
new_probs = self._model(t_states)[0].softmax(dim=1)
kl_divergence = -((new_probs / old_probs).log() * old_probs).sum(dim=1).mean()
# Empty the memory as the algorithm is on-policy and the policy has changed with the model
self._memory.clear()
# Plot
self._plotter.add_scalar("Combined Loss", (policy_loss + value_loss - self._beta * entropy).item(), step_idx)
self._plotter.add_scalar("Entropy", entropy.item(), step_idx)
self._plotter.add_scalar("KL Divergence", kl_divergence.item(), step_idx)
def mtime(self):
"""See ctime property, but returns modification time """
return unpack(">LL", self[5])
def ctime(self):
"""
:return:
Tuple(int_time_seconds_since_epoch, int_nano_seconds) of the
file's creation time"""
return unpack(">LL", self[4])
def unpack(self, mda_name):
util.unpack(self.config, mda_name)
self.reload_config()
post_unpack_hook = self.config.get_post_unpack_hook()
if post_unpack_hook:
util.run_post_unpack_hook(post_unpack_hook)
except IOError:
exit("Unable to open " + filename + ".")
if log_filename == "stdout":
log_file = sys.stdout
else:
try:
log_file = open(log_filename, 'w')
except IOError:
exit("Unable to open " + log_filename + ".")
next_acknum = 0
# Receive first packet
packet, addr = recv_sock.recvfrom(576)
source_port, dest_port, seqnum, acknum, header_length, \
ack, final, window_size, contents = util.unpack(packet)
checksum = util.get_checksum(packet)
packet_valid = checksum == 0 and next_acknum == acknum
if packet_valid:
recv_file.write(contents)
next_acknum += 1
log = str(datetime.datetime.now()) + " " + str(source_port) + " " + str(
dest_port) + " " + str(seqnum) + " " + str(acknum)
log_file.write(log + "\n")
# Establish ack socket connection
ack_sock.connect((sender_ip, sender_port))
out_port = ack_sock.getsockname()[1]
estimated_rtt = recv_time - send_time
dev_rtt = 0
recv_sock.settimeout(timeout_time)
except socket.timeout:
retransmitted += 1
continue
while True:
try:
ack = recv_sock.recv(20)
recv_time = time.time()
# Unpack packet information
ack_source_port, ack_dest_port, ack_seqnum,\
ack_acknum, ack_header_length, ack_valid,\
ack_final, ack_window_size, ack_contents = util.unpack(ack)
log = str(datetime.datetime.now()) + " " + \
str(ack_source_port) + " " + \
str(ack_dest_port) + " " + \
str(ack_seqnum) + " " + \
str(ack_acknum) + "\n"
# Log flags
if ack_valid:
log = log.strip("\n") + " ACK\n"
if ack_final:
log = log.strip("\n") + " FIN\n"
# If valid, here we go!
if ack_acknum == acknum and ack_valid:
