def load_mode(args):
core = get_plugin(args.device, args.l, args.config)
log.info('IR for {} : {}'.format(args.device, args.model))
log.info('Loading blob from {}'.format(args.load))
net = get_net(model=args.model, core=core)
net_layers, net_inputs, net_outputs = get_model_info(net)
out_layers = get_layers_list(net_layers, net_inputs, net_outputs, args.layers)
print_input_layers(net_inputs)
print_output_layers(out_layers)
layers_map = manage_user_outputs_with_mapping(mapping=args.mapping, reference_mapping=args.reference_mapping,
user_layers=out_layers)
inputs = input_processing(args.model, net_inputs, args.input, layers_map)
global_accuracy = []
loaded = load_dump(args.load)
for out_layer in layers_map:
ref_out_layer = layers_map[out_layer]
if out_layer == ref_out_layer:
log.info('Layer {} statistics'.format(out_layer))
else:
log.info('Statistics \'{}\' vs \'{}\''.format(out_layer, ref_out_layer))
net_copy = get_net_copy_with_output(model=args.model, output=out_layer, core=core)
results = infer(net=net_copy, core=core, device=args.device, inputs=inputs, output=[out_layer])
if out_layer not in results:
continue
out_blob, pc = results[out_layer]
if ref_out_layer not in loaded:
continue
ref_out_blob = loaded[ref_out_layer]['blob']
a_m = accuracy_metrics(out_blob=out_blob, ref_out_blob=ref_out_blob)
if 'pc' in loaded[ref_out_layer]:
ref_pc = loaded[ref_out_layer]['pc']
performance_metrics(pc=pc, ref_pc=ref_pc)
blob_counters(out_blob=out_blob, ref_out_blob=ref_out_blob)
global_accuracy = update_global_accuracy_matrics(global_accuracy=global_accuracy, current_accuracy=a_m)
print_all_over_the_net_metrics(global_accuracy=global_accuracy)
def eff_dump_csv_upload(request):
""" Allows to import the logs contained in an Eff formatted file.
The file must be structured as one generated by L{EffCsvWriter<sitio.eff.utils.EffCsvWriter>}.
If there are any external id's for the logs contained in the file being uploaded not found
on the database, then a form that allows to associate these external id's with
existing user profiles will be presented to the user
(see L{eff_admin_users_association<eff_admin_users_association>}).
@param request: the request object
@type request: django.core.handlers.wsgi.WSGIRequest
@return: The django.http.HttpResponse object for the view
"""
context = {'title' : 'CSV Dump Upload'}
if request.method == 'POST':
form = DumpUploadForm(request.POST, request.FILES)
if form.is_valid():
n_rows, n_users, n_projects, n_project_assocs, temp_file = load_dump(request.FILES['csv_file'].file)
if temp_file:
request.session['log_entries_file'] = temp_file
request.session['n_users'] = n_users
return HttpResponseRedirect('/efi/administration/users_association/')
context['notices'] = ['File Uploaded Sucessfully!']
else:
context['errors'] = ['Invalid Form']
else:
form = DumpUploadForm()
context['form'] = form
return render_to_response('admin_dump_csv_upload.html', context, context_instance=RequestContext(request))
def main():
parser = ArgumentParser()
parser.add_argument('--config', default='./configs/lgb_template.yaml')
parser.add_argument('--create-features', action='store_true')
options = parser.parse_args()
config = yaml.safe_load(open(options.config))
kfold = get_kfold(config)
if options.create_features:
train_path = get_dataset_filename(config, 'train')
test_path = get_dataset_filename(config, 'test')
with timer('Load train/test dump files'):
train_df = load_dump(train_path)
test_df = load_dump(test_path)
create_features(config, train_df, test_df, kfold)
del train_df, test_df
gc.collect()
target_col = config['target']
target_path = get_dataset_filename(config, 'target')
use_features = extract_use_features(config)
x_train = load_train_dataset(use_features)
y_train = load_dump(target_path)[target_col]
output_dir = config['dataset']['output_directory']
basename = Path(options.config).stem
logger = setup_logger(output_dir, basename)
clfs, importances = train_model(x_train, y_train, kfold, config, logger)
save_feature_importances(importances, basename, output_dir)
del x_train, y_train, importances
gc.collect()
pred = predict_model(clfs, use_features, config)
print('Creating a submission csv file...')
submission_path = get_dataset_filename(config, 'sample_submit')
submission = pd.read_csv(submission_path)
submission[target_col] = pred
submission.to_csv(f'{output_dir}/submit_{basename}.csv.gz', index=False)
print('Done.')
def randomsample(args):
"""
Produce a dataset by sampling.
"""
if args.language:
langs = args.language
elif args.langs:
langs = map(str.strip, open(args.langs))
tempfile.tempdir = args.temp
now = time.time()
handle, path = tempfile.mkstemp(suffix='.tar', prefix='sample-{0}-'.format(args.number))
os.close(handle)
#path = os.path.join(args.temp, 'sample-{0}.tar'.format(args.number))
build_index = not(args.skip_index)
with tarfile.open(path, 'w') as tar:
for lang in langs:
try:
dump = load_dump(lang, build_index=build_index, unpack=True)
except KeyError:
logger.error("do not have a dump for %s, skipping", lang)
continue
chosen = set() #keeps track of ids that have been chosen
used = set() #keeps track of ids that have been examined
# Adapted from wikicontent
while len(chosen) < args.number:
logger.debug("chose {0}/{1} so far".format(len(chosen), args.number))
try:
id = random.choice(xrange(dump.metadata['size']))
if id in used:
if len(used) >= dump.metadata['size']:
# We have run out of documents to consider. Bail out.
logger.warning("ran out of documents for %s", lang)
break
raise ReTry("already considered {0}".format(id))
used.add(id)
logger.debug("processing {0}".format(id))
page = dump.get_page_by_index(id)
if args.clean:
# apply mediawiki removal
text = remove_mediawiki_syntax(page.text)
para = paragraphs(text)
content = []
for p in para:
p = re.sub('\s',' ', p)
if regexps.langref.match(p): continue
if regexps.assoc.match(p): continue
if regexps.syntax.search(p): continue
p = regexps.tripquote.sub('\g<name>',p)
p = regexps.doubquote.sub('\g<name>',p)
u = p.decode('utf8').strip()
if regexps.category_name.search(u): continue
content.append(u)
if not content: raise ReTry, "No usable content"
document = '\n\n'.join(u.encode('utf8') for u in content) + '\n'
else:
# output raw mediawiki
document = page.text
if len(document) < args.minlen:
raise ReTry, "Too short ({0}<{1})".format(len(document), args.minlen)
logger.debug("adding {0}".format(id))
info = tarfile.TarInfo('{0}/{1}'.format(lang, id))
info.size = len(document)
info.mtime = now
chosen.add(id)
tar.addfile(info, StringIO(document))
logger.debug("added {0}".format(id))
except ReTry, e:
logger.debug("Reject: %s", e.args[0])
continue
logger.info("chose %d documents for %s", len(chosen), lang)
def randomsample(args):
"""
Produce a dataset by sampling.
"""
if args.language:
langs = args.language
elif args.langs:
langs = map(str.strip, open(args.langs))
tempfile.tempdir = args.temp
now = time.time()
path = os.path.join(config.get("paths", "scratch"), "sample-{0}.tar".format(args.number))
build_index = not (args.skip_index)
with tarfile.open(path, "w") as tar:
for lang in langs:
try:
dump = load_dump(lang, build_index=build_index, unpack=True)
except KeyError:
logger.error("do not have a dump for %s, skipping", lang)
continue
chosen = set() # keeps track of ids that have been chosen
used = set() # keeps track of ids that have been examined
# Adapted from wikicontent
while len(chosen) < args.number:
logger.debug("chose {0}/{1} so far".format(len(chosen), args.number))
try:
id = random.choice(xrange(dump.metadata["size"]))
if id in used:
if len(used) >= dump.metadata["size"]:
# We have run out of documents to consider. Bail out.
logger.warning("ran out of documents for %s", lang)
break
raise ReTry("already considered {0}".format(id))
used.add(id)
logger.debug("processing {0}".format(id))
page = dump.get_page_by_index(id)
if args.clean:
# apply mediawiki removal
text = remove_mediawiki_syntax(page.text)
para = paragraphs(text)
content = []
for p in para:
p = re.sub("\s", " ", p)
if regexps.langref.match(p):
continue
if regexps.assoc.match(p):
continue
if regexps.syntax.search(p):
continue
p = regexps.tripquote.sub("\g<name>", p)
p = regexps.doubquote.sub("\g<name>", p)
u = p.decode("utf8").strip()
if regexps.category_name.search(u):
continue
content.append(u)
if not content:
raise ReTry, "No usable content"
document = "\n\n".join(u.encode("utf8") for u in content) + "\n"
else:
# output raw mediawiki
document = page.text
if len(document) < args.minlen:
raise ReTry, "Too short ({0}<{1})".format(len(text), args.minlen)
logger.debug("adding {0}".format(id))
info = tarfile.TarInfo("{0}/{1}".format(lang, id))
info.size = len(document)
info.mtime = now
chosen.add(id)
tar.addfile(info, StringIO(document))
logger.debug("added {0}".format(id))
except ReTry, e:
logger.debug("Reject: %s", e.args[0])
continue
logger.info("chose %d documents for %s", len(chosen), lang)
