def run_training(head_host, head_port, debug_out=None):
"""Main worker training routine (creates the Seq2SeqTrainingService and connects it to the
head.
@param head_host: hostname of the head
@param head_port: head port number
@param debug_out: path to the debugging output file (debug output discarded if None)
"""
# setup debugging output, if applicable
if debug_out is not None:
set_debug_stream(file_stream(debug_out, mode='w'))
# start the server (in the background)
log_info('Creating training server...')
server = ThreadPoolServer(service=Seq2SeqTrainingService, nbThreads=1)
server_thread = Thread(target=server.start)
server_thread.start()
my_host = socket.getfqdn()
log_info('Worker server created at %s:%d. Connecting to head at %s:%d...' %
(my_host, server.port, head_host, head_port))
# notify main about this server
conn = connect(head_host, head_port, config={'allow_pickle': True})
conn.root.register_worker(my_host, server.port)
conn.close()
log_info('Worker is registered with the head.')
# now serve until we're killed (the server thread will continue to run)
server_thread.join()
def run_training(head_host, head_port, debug_out=None):
"""Main worker training routine (creates the Seq2SeqTrainingService and connects it to the
head.
@param head_host: hostname of the head
@param head_port: head port number
@param debug_out: path to the debugging output file (debug output discarded if None)
"""
# setup debugging output, if applicable
if debug_out is not None:
set_debug_stream(file_stream(debug_out, mode='w'))
# start the server (in the background)
log_info('Creating training server...')
server = ThreadPoolServer(service=Seq2SeqTrainingService, nbThreads=1)
server_thread = Thread(target=server.start)
server_thread.start()
my_host = socket.getfqdn()
log_info('Worker server created at %s:%d. Connecting to head at %s:%d...' %
(my_host, server.port, head_host, head_port))
# notify main about this server
conn = connect(head_host, head_port, config={'allow_pickle': True})
conn.root.register_worker(my_host, server.port)
conn.close()
log_info('Worker is registered with the head.')
# now serve until we're killed (the server thread will continue to run)
server_thread.join()
def load_from_file(model_fname):
"""Load the generator from a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph must be stored with a \
different extension
"""
log_info("Loading generator from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
data = pickle.load(fh)
ret = Seq2SeqGen(cfg=data['cfg'])
ret.load_all_settings(data)
if ret.classif_filter:
classif_filter_fname = re.sub(r'((.pickle)?(.gz)?)$', r'.tftreecl\1', model_fname)
if os.path.isfile(classif_filter_fname):
ret.classif_filter = RerankingClassifier.load_from_file(classif_filter_fname)
else:
log_warn("Classification filter data not found, ignoring.")
ret.classif_filter = False
# re-build TF graph and restore the TF session
tf_session_fname = re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname)
ret._init_neural_network()
ret.saver.restore(ret.session, tf_session_fname)
return ret
def load_from_file(fname):
log_info('Loading model from ' + fname)
with file_stream(fname, mode='rb', encoding=None) as fh:
candgen = pickle.load(fh)
# various backward compatibility tricks
if type(candgen) == dict:
child_type_counts = candgen
candgen = RandomCandidateGenerator({})
candgen.child_type_counts = child_type_counts
candgen.child_num_cdfs = pickle.load(fh)
candgen.max_children = pickle.load(fh)
if not hasattr(candgen, 'node_limits'):
candgen.node_limits = None
if not hasattr(candgen, 'child_type_counts'):
candgen.child_type_counts = candgen.form_counts
candgen.child_num_cdfs = candgen.child_cdfs
if not hasattr(candgen, 'exp_child_num'):
candgen.exp_child_num = candgen.exp_from_cdfs(
candgen.child_num_cdfs)
if not hasattr(candgen, 'compatible_dais'):
candgen.compatible_dais = None
candgen.compatible_dais_type = None
candgen.compatible_dais_limit = 1000
if not hasattr(candgen, 'compatible_slots'):
candgen.compatible_slots = False
if not hasattr(candgen, 'classif'):
candgen.classif = None
return candgen
def save_to_file(self, model_fname):
"""Save the generator to a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph will be stored with a \
different extension
"""
log_info("Saving generator to %s..." % model_fname)
if self.classif_filter:
classif_filter_fname = re.sub(r'((.pickle)?(.gz)?)$',
r'.tftreecl\1', model_fname)
self.classif_filter.save_to_file(classif_filter_fname)
if self.lexicalizer:
lexicalizer_fname = re.sub(r'((.pickle)?(.gz)?)$', r'.lexic\1',
model_fname)
self.lexicalizer.save_to_file(lexicalizer_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self.get_all_settings(),
fh,
protocol=pickle.HIGHEST_PROTOCOL)
tf_session_fname = re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname)
if hasattr(self, 'checkpoint_path') and self.checkpoint_path:
shutil.copyfile(self.checkpoint_path, tf_session_fname)
else:
self.saver.save(self.session, tf_session_fname)
def process_document(self, doc):
"Write a CoNLL-U file"
out = file_stream(self.get_output_file_name(doc),
'w',
encoding='UTF-8')
for bundle in doc.bundles:
zone = bundle.get_zone(self.language, self.selector)
nodes = zone.atree.get_descendants(ordered=1)
# Empty sentences are not allowed in CoNLL-U.
if len(nodes) == 0:
continue
comment = zone.wild['comment']
if comment:
out.write('#' + comment.rstrip('\r\n').replace('\n', '\n#') +
'\n')
index = 1
for node in nodes:
out.write('\t'.join(
'_' if value is None else value
for value in map((lambda x: str(x) if type(
x) == int else getattr(node, x, '_')), [
index, 'form', 'lemma', 'upos', 'xpos', 'feats',
node.parent.ord, 'deprel', 'deps', 'misc'
])) + '\n')
index += 1
out.write('\n')
def save_to_file(self, model_fname):
"""Save the whole ensemble into a file (get all settings and parameters, dump them in a
pickle)."""
log_info("Saving generator to %s..." % model_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self.__class__, fh, protocol=pickle.HIGHEST_PROTOCOL)
pickle.dump(self.cfg, fh, protocol=pickle.HIGHEST_PROTOCOL)
gens_dump = []
for gen in self.gens:
setting = gen.get_all_settings()
parset = gen.get_model_params()
setting['classif_filter'] = self.classif_filter is not None
gens_dump.append((setting, parset))
pickle.dump(gens_dump, fh, protocol=pickle.HIGHEST_PROTOCOL)
if self.classif_filter:
pickle.dump(self.classif_filter.get_all_settings(),
fh,
protocol=pickle.HIGHEST_PROTOCOL)
pickle.dump(self.classif_filter.get_model_params(),
fh,
protocol=pickle.HIGHEST_PROTOCOL)
def load_from_file(model_fname):
"""Load the generator from a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph must be stored with a \
different extension
"""
log_info("Loading generator from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
data = pickle.load(fh)
ret = Seq2SeqGen(cfg=data['cfg'])
ret.load_all_settings(data)
if ret.classif_filter:
classif_filter_fname = re.sub(r'((.pickle)?(.gz)?)$',
r'.tftreecl\1', model_fname)
if os.path.isfile(classif_filter_fname):
ret.classif_filter = RerankingClassifier.load_from_file(
classif_filter_fname)
else:
log_warn("Classification filter data not found, ignoring.")
ret.classif_filter = False
# re-build TF graph and restore the TF session
tf_session_fname = re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname)
ret._init_neural_network()
ret.saver.restore(ret.session, tf_session_fname)
return ret
def process_document(self, filename):
"Read a YAML file and return its contents as a Document object"
f = file_stream(filename, encoding=None)
data = yaml.load(f)
doc = Document(filename, data)
f.close()
return doc
def load_from_file(fname):
log_info('Loading model from ' + fname)
with file_stream(fname, mode='rb', encoding=None) as fh:
candgen = pickle.load(fh)
# various backward compatibility tricks
if type(candgen) == dict:
child_type_counts = candgen
candgen = RandomCandidateGenerator({})
candgen.child_type_counts = child_type_counts
candgen.child_num_cdfs = pickle.load(fh)
candgen.max_children = pickle.load(fh)
if not hasattr(candgen, 'node_limits'):
candgen.node_limits = None
if not hasattr(candgen, 'child_type_counts'):
candgen.child_type_counts = candgen.form_counts
candgen.child_num_cdfs = candgen.child_cdfs
if not hasattr(candgen, 'exp_child_num'):
candgen.exp_child_num = candgen.exp_from_cdfs(candgen.child_num_cdfs)
if not hasattr(candgen, 'compatible_dais'):
candgen.compatible_dais = None
candgen.compatible_dais_type = None
candgen.compatible_dais_limit = 1000
if not hasattr(candgen, 'compatible_slots'):
candgen.compatible_slots = False
if not hasattr(candgen, 'classif'):
candgen.classif = None
return candgen
def process_document(self, filename):
"Read a YAML file and return its contents as a Document object"
f = file_stream(filename, encoding=None)
data = yaml.load(f)
doc = Document(filename, data)
f.close()
return doc
def load_from_file(model_fname):
"""Detect correct model type (plain/ensemble) and start loading."""
model_type = Seq2SeqGen # default to plain generator
with file_stream(model_fname, 'rb', encoding=None) as fh:
data = pickle.load(fh)
if isinstance(data, type):
model_type = data
return model_type.load_from_file(model_fname)
def process_document(self, doc):
"Write a YAML document"
data = []
for bundle in doc.bundles:
data.append(self.serialize_bundle(bundle))
out = file_stream(self.get_output_file_name(doc), 'w', encoding=None)
out.write(yaml.safe_dump(data, allow_unicode=True,
explicit_start=True))
out.close()
def convert_model(model_fname):
reset_default_graph()
param_fname = re.sub(r'((.pickle)?(.gz)?)$', r'.params.gz', model_fname)
log_info('Converting %s to %s...' % (model_fname, param_fname))
model = Seq2SeqBase.load_from_file(model_fname)
with file_stream(param_fname, 'wb', encoding=None) as fh:
pickle.dump(model.get_model_params(), fh, protocol=pickle.HIGHEST_PROTOCOL)
def process_document(self, doc):
"Write a YAML document"
data = []
for bundle in doc.bundles:
data.append(self.serialize_bundle(bundle))
out = file_stream(self.get_output_file_name(doc), 'w', encoding=None)
out.write(yaml.safe_dump(data, allow_unicode=True,
explicit_start=True))
out.close()
def save_to_file(self, model_file):
"""\
Save the model to a pickle file or stream (supports GZip compression).
"""
log_info('Saving model to file ' + str(model_file))
fh = file_stream(model_file, mode='wb', encoding=None)
pickle.Pickler(fh, pickle.HIGHEST_PROTOCOL).dump(self)
fh.close()
log_info('Model successfully saved.')
def write_data(dir, fname_base, fname_repl, data):
chunk_size = len(data[0])
for chunk_idx in xrange(chunk_size):
fname_suff = ".%d" % chunk_idx if chunk_size > 1 else ''
file_name = os.path.join(dir, re.sub(r'^[^-._]*', fname_repl + fname_suff, fname_base))
print 'WRITING ' + file_name
with file_stream(file_name, 'w') as fh:
for chunk in data:
print >> fh, chunk[chunk_idx],
def save_to_file(self, model_file):
"""\
Save the model to a pickle file or stream (supports GZip compression).
"""
log_info('Saving model to file ' + str(model_file))
fh = file_stream(model_file, mode='wb', encoding=None)
pickle.Pickler(fh, pickle.HIGHEST_PROTOCOL).dump(self)
fh.close()
log_info('Model successfully saved.')
def load_from_file(model_fname):
"""Detect correct model type (plain/ensemble) and start loading."""
model_type = Seq2SeqGen # default to plain generator
with file_stream(model_fname, 'rb', encoding=None) as fh:
data = pickle.load(fh)
if isinstance(data, type):
model_type = data
return model_type.load_from_file(model_fname)
def write_data(dir, fname_base, fname_repl, data):
chunk_size = len(data[0])
for chunk_idx in xrange(chunk_size):
fname_suff = ".%d" % chunk_idx if chunk_size > 1 else ''
file_name = os.path.join(
dir, re.sub(r'^[^-._]*', fname_repl + fname_suff, fname_base))
print 'WRITING ' + file_name
with file_stream(file_name, 'w') as fh:
for chunk in data:
print >> fh, chunk[chunk_idx],
def convert_model(model_fname):
reset_default_graph()
param_fname = re.sub(r'((.pickle)?(.gz)?)$', r'.params.gz', model_fname)
log_info('Converting %s to %s...' % (model_fname, param_fname))
model = Seq2SeqBase.load_from_file(model_fname)
with file_stream(param_fname, 'wb', encoding=None) as fh:
pickle.dump(model.get_model_params(),
fh,
protocol=pickle.HIGHEST_PROTOCOL)
def load_from_file(model_file):
"""\
Load the model from a pickle file or stream
(supports GZip compression).
"""
log_info('Loading model from file ' + str(model_file))
fh = file_stream(model_file, mode='rb', encoding=None)
unpickler = pickle.Unpickler(fh)
model = unpickler.load()
fh.close()
log_info('Model loaded successfully.')
return model
def load_from_file(model_file):
"""\
Load the model from a pickle file or stream
(supports GZip compression).
"""
log_info('Loading model from file ' + str(model_file))
fh = file_stream(model_file, mode='rb', encoding=None)
unpickler = pickle.Unpickler(fh)
model = unpickler.load()
fh.close()
log_info('Model loaded successfully.')
return model
def save_to_file(self, model_fname):
"""Save the generator to a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph will be stored with a \
different extension
"""
log_info("Saving classifier to %s..." % model_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self.get_all_settings(), fh, protocol=pickle.HIGHEST_PROTOCOL)
tf_session_fname = re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname)
if self.checkpoint_path:
shutil.copyfile(self.checkpoint_path, tf_session_fname)
else:
self.saver.save(self.session, tf_session_fname)
def process_document(self, filename):
"""\
Read a Tecto-Template file and return its contents as
a Document object.
"""
fh = file_stream(filename, encoding=self.encoding)
doc = Document(filename)
for line in fh:
bundle = doc.create_bundle()
zone = bundle.create_zone(self.language, self.selector)
ttree = zone.create_ttree()
self.parse_line(line, ttree)
log_info('Parsed a tree with %d nodes.' %
len(ttree.get_descendants()))
fh.close()
return doc
def save_to_file(self, model_fname):
"""Save the classifier to a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph will be stored with a \
different extension
"""
log_info("Saving classifier to %s..." % model_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self.get_all_settings(),
fh,
protocol=pickle.HIGHEST_PROTOCOL)
tf_session_fname = re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname)
if hasattr(self, 'checkpoint_path') and self.checkpoint_path:
self.restore_checkpoint()
shutil.rmtree(os.path.dirname(self.checkpoint_path))
self.saver.save(self.session, tf_session_fname)
def load_from_file(model_fname):
"""Load the reranker from a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph must be stored with a \
different extension
"""
log_info("Loading reranker from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
data = pickle.load(fh)
ret = RerankingClassifier(cfg=data['cfg'])
ret.load_all_settings(data)
# re-build TF graph and restore the TF session
tf_session_fname = re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname)
ret._init_neural_network()
ret.saver.restore(ret.session, tf_session_fname)
return ret
def save_to_arff(self, filename, encoding='UTF-8'):
"""
Save the data set to an ARFF file
"""
# open the file
fh = file_stream(filename, 'w', encoding)
# print the relation name
print >> fh, '@relation ' + (self.relation_name
if self.relation_name is not None
else '<noname>')
# print the list of attributes
for attrib in self.attribs:
print >> fh, '@attribute ' + attrib.name + ' ' + \
attrib.get_arff_type()
# print instances
print >> fh, '@data'
for inst, weight in zip(self.data, self.inst_weights):
print >> fh, self.__get_arff_line(inst, weight)
def load_from_file(model_fname):
"""Load the reranker from a file (actually two files, one for configuration and one
for the TensorFlow graph, which must be stored separately).
@param model_fname: file name (for the configuration file); TF graph must be stored with a \
different extension
"""
log_info("Loading reranker from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
data = pickle.load(fh)
ret = RerankingClassifier(cfg=data['cfg'])
ret.load_all_settings(data)
# re-build TF graph and restore the TF session
tf_session_fname = os.path.abspath(re.sub(r'(.pickle)?(.gz)?$', '.tfsess', model_fname))
ret._init_neural_network()
ret.saver.restore(ret.session, tf_session_fname)
return ret
def create_training_job(config,
work_dir,
train_file,
name=None,
memory=8,
encoding='UTF-8'):
"""\
Submit a training process on the cluster which will save the
model to a pickle. Return the submitted job and the future location of
the model pickle.
train_file cannot be a stream, it must be an actual file.
"""
# purge name
if name is None:
name = 'TR-' + re.sub(r'[^A-Za-z0-9_]', '_', train_file)
else:
name = re.sub(r'[^A-Za-z0-9_]', '_', name)
# create working directory, if not existing
if not os.path.isdir(work_dir):
os.mkdir(work_dir)
train_file = os.path.abspath(train_file)
# generate model file name
model_file = os.path.abspath(
os.path.join(work_dir, name + '-model.pickle.gz'))
config_pickle = os.path.abspath(
os.path.join(work_dir, name + '-cfg.pickle.gz'))
# create the configuration pickle
fh = file_stream(config_pickle, mode='wb', encoding=None)
pickle.Pickler(fh, pickle.HIGHEST_PROTOCOL).dump(config)
fh.close()
# create the job
job = Job(name=name, work_dir=work_dir)
job.code = "fh = file_stream('" + config_pickle + \
"', mode='rb', encoding=None)\n" + \
"cfg = pickle.Unpickler(fh).load()\n" + \
"fh.close()\n" + \
"model = Model(cfg)\n" + \
"model.train('" + train_file + "', encoding='" + \
encoding + "')\n" \
"model.save_to_file('" + model_file + "')\n"
job.header += "from pytreex.tool.ml.model import Model\n" + \
"import pickle\n" + \
"from pytreex.core.util import file_stream\n"
return job, model_file
def run_worker(head_host, head_port, debug_out=None):
# setup debugging output, if applicable
if debug_out is not None:
set_debug_stream(file_stream(debug_out, mode='w'))
# start the server (in the background)
log_info('Creating worker server...')
server = ThreadPoolServer(service=RankerTrainingService, nbThreads=1)
server_thread = Thread(target=server.start)
server_thread.start()
my_host = socket.getfqdn()
log_info('Worker server created at %s:%d. Connecting to head at %s:%d...' %
(my_host, server.port, head_host, head_port))
# notify main about this server
conn = connect(head_host, head_port, config={'allow_pickle': True})
conn.root.register_worker(my_host, server.port)
conn.close()
log_info('Worker is registered with the head.')
# now serve until we're killed (the server thread will continue to run)
server_thread.join()
def run_worker(head_host, head_port, debug_out=None):
# setup debugging output, if applicable
if debug_out is not None:
set_debug_stream(file_stream(debug_out, mode='w'))
# start the server (in the background)
log_info('Creating worker server...')
server = ThreadPoolServer(service=RankerTrainingService, nbThreads=1)
server_thread = Thread(target=server.start)
server_thread.start()
my_host = socket.getfqdn()
log_info('Worker server created at %s:%d. Connecting to head at %s:%d...' %
(my_host, server.port, head_host, head_port))
# notify main about this server
conn = connect(head_host, head_port, config={'allow_pickle': True})
conn.root.register_worker(my_host, server.port)
conn.close()
log_info('Worker is registered with the head.')
# now serve until we're killed (the server thread will continue to run)
server_thread.join()
def load_from_arff(self, filename, encoding='UTF-8'):
"""
Load an ARFF file/stream, filling the data structures.
"""
# initialize
if not self.is_empty:
raise IOError('Cannot store second data set into the same object.')
status = 'header' # we first assume to read the header
line_num = 1 # line counter
instances = []
weights = []
# open the file
fh = file_stream(filename, encoding=encoding)
# parse the file
for line in fh:
line = line.strip()
# skip comments
if line.startswith('%'):
continue
# relation name
elif line.lower().startswith('@relation'):
self.relation_name = line.split(None, 1)[1]
# attribute definition
elif line.lower().startswith('@attribute'):
attr_name, attr_type = line.split(None, 2)[1:]
self.attribs.append(Attribute(attr_name, attr_type))
# data section start
elif line.lower().startswith('@data'):
status = 'data'
# data lines
elif status == 'data' and line != '':
inst, weight = self.__parse_line(line, line_num)
instances.append(inst)
weights.append(weight)
line_num += 1
fh.close()
# store the resulting matrix
self.data = instances
self.inst_weights = weights
# remember attribute names
self.attribs_by_name = {attr.name: idx
for idx, attr in enumerate(self.attribs)}
def process_document(self, doc):
"Write a CoNLL-U file"
out = file_stream(self.get_output_file_name(doc), 'w', encoding='UTF-8')
for bundle in doc.bundles:
zone = bundle.get_zone(self.language, self.selector)
nodes = zone.atree.get_descendants(ordered=1)
# Empty sentences are not allowed in CoNLL-U.
if len(nodes)==0:
continue
comment = zone.wild['comment']
if comment:
out.write('#' + comment.rstrip('\r\n').replace('\n','\n#') + '\n')
index = 1
for node in nodes:
out.write('\t'.join(
'_' if value is None else value for value in
map((lambda x: str(x) if type(x)==int else getattr(node, x, '_')),
[index, 'form', 'lemma', 'upos', 'xpos', 'feats', node.parent.ord, 'deprel', 'deps', 'misc'])
) + '\n')
index += 1
out.write('\n')
def create_training_job(config, work_dir, train_file,
name=None, memory=8, encoding='UTF-8'):
"""\
Submit a training process on the cluster which will save the
model to a pickle. Return the submitted job and the future location of
the model pickle.
train_file cannot be a stream, it must be an actual file.
"""
# purge name
if name is None:
name = 'TR-' + re.sub(r'[^A-Za-z0-9_]', '_', train_file)
else:
name = re.sub(r'[^A-Za-z0-9_]', '_', name)
# create working directory, if not existing
if not os.path.isdir(work_dir):
os.mkdir(work_dir)
train_file = os.path.abspath(train_file)
# generate model file name
model_file = os.path.abspath(os.path.join(work_dir,
name + '-model.pickle.gz'))
config_pickle = os.path.abspath(os.path.join(work_dir,
name + '-cfg.pickle.gz'))
# create the configuration pickle
fh = file_stream(config_pickle, mode='wb', encoding=None)
pickle.Pickler(fh, pickle.HIGHEST_PROTOCOL).dump(config)
fh.close()
# create the job
job = Job(name=name, work_dir=work_dir)
job.code = "fh = file_stream('" + config_pickle + \
"', mode='rb', encoding=None)\n" + \
"cfg = pickle.Unpickler(fh).load()\n" + \
"fh.close()\n" + \
"model = Model(cfg)\n" + \
"model.train('" + train_file + "', encoding='" + \
encoding + "')\n" \
"model.save_to_file('" + model_file + "')\n"
job.header += "from pytreex.tool.ml.model import Model\n" + \
"import pickle\n" + \
"from pytreex.core.util import file_stream\n"
return job, model_file
def load_from_file(model_fname):
"""Load the whole ensemble from a file (load settings and model parameters, then build the
ensemble network)."""
log_info("Loading ensemble generator from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
typeid = pickle.load(fh)
if typeid != Seq2SeqEnsemble:
raise ValueError('Wrong type identifier in file %s' % model_fname)
cfg = pickle.load(fh)
ret = Seq2SeqEnsemble(cfg)
gens_dump = pickle.load(fh)
if 'classif_filter' in cfg:
rerank_settings = pickle.load(fh)
rerank_params = pickle.load(fh)
else:
rerank_settings = None
rerank_params = None
ret.build_ensemble(gens_dump, rerank_settings, rerank_params)
return ret
def save_to_file(self, model_fname):
"""Save the whole ensemble into a file (get all settings and parameters, dump them in a
pickle)."""
log_info("Saving generator to %s..." % model_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self.__class__, fh, protocol=pickle.HIGHEST_PROTOCOL)
pickle.dump(self.cfg, fh, protocol=pickle.HIGHEST_PROTOCOL)
gens_dump = []
for gen in self.gens:
setting = gen.get_all_settings()
parset = gen.get_model_params()
setting['classif_filter'] = self.classif_filter is not None
gens_dump.append((setting, parset))
pickle.dump(gens_dump, fh, protocol=pickle.HIGHEST_PROTOCOL)
if self.classif_filter:
pickle.dump(self.classif_filter.get_all_settings(), fh,
protocol=pickle.HIGHEST_PROTOCOL)
pickle.dump(self.classif_filter.get_model_params(), fh,
protocol=pickle.HIGHEST_PROTOCOL)
def load_from_file(model_fname):
"""Load the whole ensemble from a file (load settings and model parameters, then build the
ensemble network)."""
# TODO support for lexicalizer
log_info("Loading ensemble generator from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
typeid = pickle.load(fh)
if typeid != Seq2SeqEnsemble:
raise ValueError('Wrong type identifier in file %s' % model_fname)
cfg = pickle.load(fh)
ret = Seq2SeqEnsemble(cfg)
gens_dump = pickle.load(fh)
if 'classif_filter' in cfg:
rerank_settings = pickle.load(fh)
rerank_params = pickle.load(fh)
else:
rerank_settings = None
rerank_params = None
ret.build_ensemble(gens_dump, rerank_settings, rerank_params)
return ret
def save_to_file(self, fname):
log_info('Saving model to ' + fname)
with file_stream(fname, mode='wb', encoding=None) as fh:
pickle.dump(self, fh, pickle.HIGHEST_PROTOCOL)
def load_from_file(fname):
log_info('Loading model from ' + fname)
with file_stream(fname, mode='rb', encoding=None) as fh:
classif = pickle.load(fh)
return classif
def process_document(self, filename):
"""\
Read a CoNLL-U file and return its contents as a Document object.
"""
fh = file_stream(filename, encoding=self.encoding)
doc = Document(filename)
bundle = doc.create_bundle()
zone = bundle.create_zone(self.language, self.selector)
root = zone.create_atree()
last_node = root
nodes = [root]
parents = [0]
comment = ''
for line in fh:
# Strip newline character (\n or \r\n)
line = line.rstrip('\r\n')
# Empty line as a end of sentence
if not line:
# Ignore (multiple) empty lines before start of sentence (invalid CoNLL-U)
if len(nodes)==1:
continue
# Rehang to correct parents and save nonempty comment to root
for i in xrange(1,len(nodes)):
nodes[i].parent = nodes[parents[i]]
if len(comment):
zone.wild['comment'] = comment
# Prepare a new bundle
bundle = doc.create_bundle()
zone = bundle.create_zone(self.language, self.selector)
root = zone.create_atree()
last_node = root
nodes = [root]
parents = [0]
comment = ''
# Comment
elif line[0] == '#':
comment += line[1:] + "\n"
# A normal line with one token
else:
columns = line.split('\t')
# TODO: multi-word tokens
if '-' in columns[0]:
continue
# Create new node
new_node = root.create_child(data = dict(
(key, value) for key, value in
zip(['form', 'lemma', 'upos', 'xpos', 'feats', 'deprel', 'deps', 'misc'],
columns[1:6] + columns[7:10] )
if value is not None and value != '_'
) )
nodes.append(new_node)
try:
parent_index = int(columns[6])
except (ValueError, TypeError):
# TODO: warning?
parent_index = 0
parents.append(parent_index)
# Word order TODO is this needed?
new_node.shift_after_subtree(last_node)
last_node = new_node
# The last bundle should be empty (if the file ended with an empty line),
# so we need to remove it. But let's check it.
if len(nodes)==1:
doc.bundles.pop()
else:
for i in xrange(1,len(nodes)):
nodes[i].parent = nodes[parents[i]]
if len(comment):
zone.wild['comment'] = comment
fh.close()
return doc
def main(argv):
opts, files = getopt(argv, 'f:c:d:')
folds = 10
chunk_size = 2
dir_prefix = 'cv'
for opt, arg in opts:
if opt == '-f':
folds = int(arg)
elif opt == '-c':
chunk_size = int(arg)
elif opt == '-d':
dir_prefix = arg
if not files:
sys.exit(__doc__)
random.seed(1206)
ordering = None
for file in files:
# read all data
data = []
with file_stream(file) as fh:
chunk = []
for line in fh:
chunk.append(line)
if len(chunk) == chunk_size:
data.append(chunk)
chunk = []
if chunk:
log_warn('Incomplete chunk at end of file %s, size %d' % (file, len(chunk)))
if ordering is None:
# create ordering
ordering = range(len(data))
random.shuffle(ordering)
# create directories
for fold_no in xrange(folds):
os.mkdir(dir_prefix + "%02d" % fold_no)
# output as train and test into all CV portions
fold_size, bigger_folds = divmod(len(data), folds)
for fold_no in xrange(folds):
# compute test data bounds
if fold_no < bigger_folds:
test_lo = (fold_size + 1) * fold_no
test_hi = (fold_size + 1) * (fold_no + 1)
else:
test_lo = fold_size * fold_no + bigger_folds
test_hi = fold_size * (fold_no + 1) + bigger_folds
# select train and test data instances
train_data = [data[idx] for ord, idx in enumerate(ordering)
if ord < test_lo or ord >= test_hi]
test_data = [data[idx] for ord, idx in enumerate(ordering)
if ord >= test_lo and ord < test_hi]
# write them out to a file (replace `all' in name with train/test)
fname_base = os.path.basename(file)
write_data(dir_prefix + "%02d" % fold_no, fname_base, 'train', train_data)
write_data(dir_prefix + "%02d" % fold_no, fname_base, 'test', test_data)
def save_to_file(self, model_fname):
"""Save the model to a file."""
log_info("Saving ranker to %s..." % model_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self, fh, protocol=pickle.HIGHEST_PROTOCOL)
def load_from_file(fname):
log_info('Loading model from ' + fname)
with file_stream(fname, mode='rb', encoding=None) as fh:
classif = pickle.load(fh)
return classif
def save_to_file(self, model_fname):
"""Save the model to a file."""
log_info("Saving ranker to %s..." % model_fname)
with file_stream(model_fname, 'wb', encoding=None) as fh:
pickle.dump(self, fh, protocol=pickle.HIGHEST_PROTOCOL)
def load_from_file(model_fname):
"""Load a pre-trained model from a file."""
log_info("Loading ranker from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
return pickle.load(fh)
def load_from_file(model_fname):
"""Load a pre-trained model from a file."""
log_info("Loading ranker from %s..." % model_fname)
with file_stream(model_fname, 'rb', encoding=None) as fh:
return pickle.load(fh)
def save_to_file(self, fname):
log_info('Saving model to ' + fname)
with file_stream(fname, mode='wb', encoding=None) as fh:
pickle.dump(self, fh, pickle.HIGHEST_PROTOCOL)
def main(argv):
opts, files = getopt(argv, 'f:c:d:')
folds = 10
chunk_size = 2
dir_prefix = 'cv'
for opt, arg in opts:
if opt == '-f':
folds = int(arg)
elif opt == '-c':
chunk_size = int(arg)
elif opt == '-d':
dir_prefix = arg
if not files:
sys.exit(__doc__)
random.seed(1206)
ordering = None
for file in files:
# read all data
data = []
with file_stream(file) as fh:
chunk = []
for line in fh:
chunk.append(line)
if len(chunk) == chunk_size:
data.append(chunk)
chunk = []
if chunk:
log_warn('Incomplete chunk at end of file %s, size %d' %
(file, len(chunk)))
if ordering is None:
# create ordering
ordering = range(len(data))
random.shuffle(ordering)
# create directories
for fold_no in xrange(folds):
os.mkdir(dir_prefix + "%02d" % fold_no)
# output as train and test into all CV portions
fold_size, bigger_folds = divmod(len(data), folds)
for fold_no in xrange(folds):
# compute test data bounds
if fold_no < bigger_folds:
test_lo = (fold_size + 1) * fold_no
test_hi = (fold_size + 1) * (fold_no + 1)
else:
test_lo = fold_size * fold_no + bigger_folds
test_hi = fold_size * (fold_no + 1) + bigger_folds
# select train and test data instances
train_data = [
data[idx] for ord, idx in enumerate(ordering)
if ord < test_lo or ord >= test_hi
]
test_data = [
data[idx] for ord, idx in enumerate(ordering)
if ord >= test_lo and ord < test_hi
]
# write them out to a file (replace `all' in name with train/test)
fname_base = os.path.basename(file)
write_data(dir_prefix + "%02d" % fold_no, fname_base, 'train',
train_data)
write_data(dir_prefix + "%02d" % fold_no, fname_base, 'test',
test_data)
