def test_snapshot_download_no_user_or_org(self):
# This is the smallest snapshot I could find that is not associated with a user / org.
model_name = "distilbert-base-german-cased"
path = cached_path(f"hf://{model_name}")
assert os.path.isdir(path)
assert os.path.isfile(path + ".json")
meta = _Meta.from_path(path + ".json")
assert meta.resource == f"hf://{model_name}"
def test_cached_download_no_user_or_org(self):
path = cached_path("hf://t5-small/config.json",
cache_dir=self.TEST_DIR)
assert os.path.isfile(path)
assert pathlib.Path(os.path.dirname(path)) == self.TEST_DIR
assert os.path.isfile(path + ".json")
meta = _Meta.from_path(path + ".json")
assert meta.etag is not None
assert meta.resource == "hf://t5-small/config.json"
def test_meta_backwards_compatible(self):
url = "http://fake.datastore.com/glove.txt.gz"
etag = "some-fake-etag"
filename = os.path.join(self.TEST_DIR,
_resource_to_filename(url, etag))
with open(filename, "wb") as f:
f.write(self.glove_bytes)
with open(filename + ".json", "w") as meta_file:
json.dump({"url": url, "etag": etag}, meta_file)
meta = _Meta.from_path(filename + ".json")
assert meta.resource == url
assert meta.etag == etag
assert meta.creation_time is not None
assert meta.size == len(self.glove_bytes)
def test_get_from_cache(self):
url = "http://fake.datastore.com/glove.txt.gz"
set_up_glove(url, self.glove_bytes, change_etag_every=2)
filename = get_from_cache(url, cache_dir=self.TEST_DIR)
assert filename == os.path.join(self.TEST_DIR,
_resource_to_filename(url, etag="0"))
assert os.path.exists(filename + ".json")
meta = _Meta.from_path(filename + ".json")
assert meta.resource == url
# We should have made one HEAD request and one GET request.
method_counts = Counter(call.request.method
for call in responses.calls)
assert len(method_counts) == 2
assert method_counts["HEAD"] == 1
assert method_counts["GET"] == 1
# And the cached file should have the correct contents
with open(filename, "rb") as cached_file:
assert cached_file.read() == self.glove_bytes
# A second call to `get_from_cache` should make another HEAD call
# but not another GET call.
filename2 = get_from_cache(url, cache_dir=self.TEST_DIR)
assert filename2 == filename
method_counts = Counter(call.request.method
for call in responses.calls)
assert len(method_counts) == 2
assert method_counts["HEAD"] == 2
assert method_counts["GET"] == 1
with open(filename2, "rb") as cached_file:
assert cached_file.read() == self.glove_bytes
# A third call should have a different ETag and should force a new download,
# which means another HEAD call and another GET call.
filename3 = get_from_cache(url, cache_dir=self.TEST_DIR)
assert filename3 == os.path.join(self.TEST_DIR,
_resource_to_filename(url, etag="1"))
method_counts = Counter(call.request.method
for call in responses.calls)
assert len(method_counts) == 2
assert method_counts["HEAD"] == 3
assert method_counts["GET"] == 2
with open(filename3, "rb") as cached_file:
assert cached_file.read() == self.glove_bytes
def test_snapshot_download_no_user_or_org(self):
path = cached_path("hf://t5-small")
assert os.path.isdir(path)
assert os.path.isfile(path + ".json")
meta = _Meta.from_path(path + ".json")
assert meta.resource == "hf://t5-small"