def main(args=None) -> None:
"""
CLI function to manually download all the dependencies for a pre-trained model.
Example of usage:
.. code-block:: sh
download_model fasttext
"""
if args is None:
args = sys.argv[1:]
parsed_args = get_args(args)
model_type = parsed_args.model_type
if "fasttext" in model_type and "fasttext-light" not in model_type:
download_fasttext_embeddings(saving_dir=CACHE_PATH)
elif model_type == "fasttext-light":
download_fasttext_magnitude_embeddings(saving_dir=CACHE_PATH)
elif "bpemb" in model_type:
BPEmb(
lang="multi", vs=100000, dim=300
) # The class manage the download of the pre-trained words embedding
model_path = os.path.join(CACHE_PATH, f"{model_type}.ckpt")
version_path = os.path.join(CACHE_PATH, f"{model_type}.version")
if not os.path.isfile(model_path) or not os.path.isfile(version_path):
download_weights(model_type, CACHE_PATH)
elif not latest_version(model_type, cache_path=CACHE_PATH):
print(
"A new version of the pre-trained model is available. The newest model will be downloaded."
)
download_weights(model_type, CACHE_PATH)
def test_givenModelBPEmbWeightsToDownloadVerbose_whenDownloadOk_thenVerbose(
self):
self._capture_output()
with patch("deepparse.tools.download_from_url"):
download_weights(model="bpemb", saving_dir="./", verbose=True)
actual = self.test_out.getvalue().strip()
expected = "Downloading the weights for the network bpemb."
self.assertEqual(actual, expected)
def test_givenModelWeightsToDownload_whenDownloadOk_thenWeightsAreDownloaded(self):
with patch("deepparse.tools.download_from_url") as downloader:
download_weights(model="fasttext", saving_dir="./", verbose=self.verbose)
downloader.assert_any_call("fasttext", "./", "ckpt")
downloader.assert_any_call("fasttext", "./", "version")
with patch("deepparse.tools.download_from_url") as downloader:
download_weights(model="bpemb", saving_dir="./", verbose=self.verbose)
downloader.assert_any_call("bpemb", "./", "ckpt")
downloader.assert_any_call("bpemb", "./", "version")
def main(args: argparse.Namespace) -> None:
"""
Script to manually download all the dependancies for a pre-trained model.
"""
model = args.model
root_path = os.path.join(os.path.expanduser('~'), ".cache", "deepparse")
os.makedirs(root_path, exist_ok=True)
if model == 'fasttext':
download_fasttext_embeddings("fr", saving_dir=root_path)
if model == 'bpemb':
BPEmb(
lang="multi", vs=100000, dim=300
) # The class manage the download of the pre-trained words embedding
model_path = os.path.join(root_path, f"{model}.ckpt")
version_path = os.path.join(root_path, f"{model}.version")
if not os.path.isfile(model_path) or not os.path.isfile(version_path):
download_weights(model, root_path)
elif verify_latest_version(model, root_path):
warnings.warn(
"A new version of the pre-trained model is available. The newest model will be downloaded."
)
download_weights(model, root_path)