def download(datapath):
opt = {'datapath': datapath}
opt['model'] = 'projects.personachat.kvmemnn.kvmemnn:Kvmemnn'
opt['model_file'] = 'models:convai2/kvmemnn/model'
opt['model_type'] = 'kvmemnn' # for builder
fnames = ['kvmemnn.tgz']
download_models(opt, fnames, 'convai2')
def download(datapath):
opt = {
'datapath': datapath,
'model_type': 'biranker_dialogue'
} # for builder
fnames = ['biranker_dialogue.tar.gz']
download_models(opt, fnames, 'light', version='v0.5', use_model_type=True)
def download(datapath):
"""
Download the model.
"""
opt = {'datapath': datapath, 'model_type': 'transresnet'}
fnames = ['transresnet.tgz']
download_models(opt, fnames, 'personality_captions')
def download(datapath):
opt = {
'datapath': datapath,
'model_type': 'seq2seq'
}
fnames = ['twitter_seq2seq_model.tgz']
download_models(opt, fnames, 'twitter', version='v1.0', use_model_type=True)
def download(datapath):
"""
Download the model.
"""
opt = {'datapath': datapath, 'model_type': 'transresnet_multimodal'}
fnames = ['transresnet_multimodal.tgz']
download_models(opt, fnames, 'image_chat')
def download(datapath):
opt = {'datapath': datapath}
fnames = ['hh131k_hb60k_fb60k_st1k_v1.tar.gz']
download_models(opt,
fnames,
'self_feeding',
version='v1.0',
use_model_type=False)
def download_with_model_type(datapath, model_type, version):
ddir = os.path.join(get_model_dir(datapath), 'light_whoami')
if not built(os.path.join(ddir, model_type), version):
opt = {'datapath': datapath, 'model_type': model_type}
fnames = ['model.tgz']
download_models(
opt, fnames, 'light_whoami', version=version, use_model_type=True
)
def download(datapath):
opt = {'datapath': datapath}
fnames = ['models_v1.tar.gz']
download_models(opt,
fnames,
'controllable_dialogue',
version='v1.0',
use_model_type=False)
def download(datapath):
ddir = os.path.join(get_model_dir(datapath), 'msc')
model_type = 'summsc_rag3B'
version = 'v0.1'
if not built(os.path.join(ddir, model_type), version):
opt = {'datapath': datapath, 'model_type': model_type}
fnames = [f'model_{version}.tar.gz']
download_models(opt, fnames, 'msc', version=version, use_model_type=True)
def download(datapath):
ddir = os.path.join(get_model_dir(datapath), 'sea')
model_type = 'bart_sq_gen'
version = 'v1.0'
if not built(os.path.join(ddir, model_type), version):
opt = {'datapath': datapath, 'model_type': model_type}
fnames = [f'model_{version}.tgz']
download_models(opt, fnames, 'sea', version=version, use_model_type=True)
def download(datapath):
opt = {'datapath': datapath}
fnames = ['glove.840B.300d.zip']
download_models(opt,
fnames,
'glove_vectors',
use_model_type=False,
path="http://nlp.stanford.edu/data")
def download(datapath):
model_name = 'pretrained_transformers'
mdir = os.path.join(get_model_dir(datapath), model_name)
version = 'v3.0'
if not built(mdir, version):
opt = {'datapath': datapath}
fnames = ['pretrained_transformers.tgz']
download_models(opt, fnames, model_name, version=version, use_model_type=False)
def download(datapath):
opt = {'datapath': datapath} # for builder
fnames = ['safety_models_v1.tgz']
download_models(opt,
fnames,
'dialogue_safety',
version='v0.5',
use_model_type=False)
def download(datapath):
opt = {'datapath': datapath}
fnames = ['end2end_generator_0.tar.gz']
download_models(opt,
fnames,
'wizard_of_wikipedia',
version='v0.5',
use_model_type=False)
def build(datapath, fname, model_type, version):
opt = {'datapath': datapath}
opt['model_type'] = model_type
dpath = os.path.join(datapath, 'models', 'dialogue_unlikelihood', model_type)
if not built(dpath, version):
download_models(
opt, [fname], 'dialogue_unlikelihood', version=version, use_model_type=False
)
def download(datapath):
opt = {'datapath': datapath}
model_filenames = [
'seq2seq.tar.gz',
'transformer_ranker.tar.gz',
'transformer_generator2.tar.gz',
'memnn.tar.gz',
]
download_models(opt, model_filenames, 'unittest', version='v3.0')
def download_unittest_models():
from parlai.core.params import ParlaiParser
from parlai.core.build_data import download_models
opt = ParlaiParser().parse_args(print_args=False)
model_filenames = [
'seq2seq.tar.gz', 'transformer_ranker.tar.gz',
'transformer_generator2.tar.gz'
]
with capture_output() as _:
download_models(opt, model_filenames, 'unittest', version='v2.0')
def build(datapath, fname, model_type, version):
opt = {'datapath': datapath}
opt['model_type'] = model_type
dpath = os.path.join(datapath, 'models', 'blender', model_type)
if not built(dpath, version):
print_blender()
download_models(opt, [fname],
'blender',
version=version,
use_model_type=False)
def download(datapath):
ddir = os.path.join(get_model_dir(datapath), 'blenderbot2')
model_type = 'memory_decoder'
version = 'v1.0'
if not built(os.path.join(ddir, model_type), version):
opt = {'datapath': datapath, 'model_type': model_type}
fnames = ['model.tgz']
download_models(
opt, fnames, 'blenderbot2', version=version, use_model_type=True
)
def download(datapath):
ddir = os.path.join(get_model_dir(datapath), 'tod_base_no_api')
model_type = 'tod_base_no_api'
version = 'v1.0'
if not built(os.path.join(ddir, model_type), version):
opt = {'datapath': datapath, 'model_type': model_type}
fnames = ['model.tar.gz']
download_models(
opt, fnames, 'tod', version=version, path='aws', use_model_type=True
)
def download(datapath):
version = 'v1.0'
fnames = ['md_gender_classifier.tgz']
download_models(
opt={'datapath': datapath},
fnames=fnames,
model_folder='md_gender',
version=version,
use_model_type=False,
)
def download(datapath):
model_name = 'tutorial_transformer_generator'
mdir = os.path.join(get_model_dir(datapath), model_name)
version = 'v1'
if not built(mdir, version):
opt = {'datapath': datapath}
fnames = ['tutorial_transformer_generator_v1.tar.gz']
download_models(
opt, fnames, model_name, version=version, use_model_type=False,
)
def download(datapath):
ddir = os.path.join(get_model_dir(datapath), 'hallucination')
model_type = 'bart_rag_dpr_poly'
version = 'v1.0'
if not built(os.path.join(ddir, model_type), version):
opt = {'datapath': datapath, 'model_type': model_type}
fnames = ['model.tgz']
download_models(
opt, fnames, 'hallucination', version=version, use_model_type=True
)
def download(datapath):
ddir = os.path.join(get_model_dir(datapath), 'hallucination')
version = 'v1.0'
if not built(ddir, version):
opt = {'datapath': datapath, 'model_type': 'wow_passages'}
fnames = ['wow_articles.paragraphs.tgz', 'exact.tgz', 'compressed.tgz']
download_models(opt,
fnames,
'hallucination',
version=version,
use_model_type=True)
def download(datapath):
opt = {'datapath': datapath}
# download all relevant wizard of wikipedia models
generator_download(datapath)
retrieval_download(datapath)
# now download knowledge retriever
fnames = ['knowledge_retriever.tgz']
opt['model_type'] = 'knowledge_retriever'
download_models(opt, fnames, 'wizard_of_wikipedia', version='v3.0')
def download(datapath):
opt = {'datapath': datapath}
version = 'v0.1'
fnames = [f'models_{version}.tar.gz']
download_models(
opt,
fnames,
model_folder='saferdialogues',
version=version,
use_model_type=False,
)
def download(datapath):
version = 'v1'
model_type = 'multi_turn'
opt = {'datapath': datapath, 'model_type': model_type}
fnames = [f'models_{version}.tar.gz']
download_models(
opt=opt,
fnames=fnames,
model_folder='bot_adversarial_dialogue',
version=version,
use_model_type=True,
)
def download(datapath):
model_type = 'gender_ethnicity__name_scrambling'
version = 'v1.0'
opt = {'datapath': datapath, 'model_type': model_type}
fnames = [f'{version}.tar.gz']
download_models(
opt=opt,
fnames=fnames,
model_folder='dialogue_bias',
version=version,
use_model_type=True,
)
def download(datapath):
model_type = 'gender__unlikelihood_sequence_level'
version = 'v1.0'
opt = {'datapath': datapath, 'model_type': model_type}
fnames = [f'{version}.tar.gz']
download_models(
opt=opt,
fnames=fnames,
model_folder='dialogue_bias',
version=version,
use_model_type=True,
)
def download(datapath):
model_type = 'convai2_single_task'
version = 'v1.0'
opt = {'datapath': datapath, 'model_type': model_type}
fnames = [f'{version}.tar.gz']
download_models(
opt=opt,
fnames=fnames,
model_folder='blended_skill_talk',
version=version,
use_model_type=True,
)
