def test_add_reference_local_file(runner):
with runner.isolated_filesystem():
open("file1.txt", "w").write("hello")
artifact = wandb.Artifact(type="dataset", name="my-arty")
artifact.add_reference("file://file1.txt")
assert artifact.digest == "a00c2239f036fb656c1dcbf9a32d89b4"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["file1.txt"] == {
"digest": "XUFAKrxLKna5cZ2REBfFkg==",
"ref": "file://file1.txt",
"size": 5,
}
def test_add_reference_s3_no_checksum(runner):
with runner.isolated_filesystem():
open("file1.txt", "w").write("hello")
artifact = wandb.Artifact(type="dataset", name="my-arty")
# TODO: Should we require name in this case?
artifact.add_reference("s3://my_bucket/file1.txt", checksum=False)
assert artifact.digest == "52631787ed3579325f985dc0f2374040"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["file1.txt"] == {
"digest": "s3://my_bucket/file1.txt",
"ref": "s3://my_bucket/file1.txt",
}
def log_model(self, path, opt, epoch, fitness_score, best_model=False):
model_artifact = wandb.Artifact('run_' + wandb.run.id + '_model', type='model', metadata={
'original_url': str(path),
'epochs_trained': epoch + 1,
'save period': opt.save_period,
'project': opt.project,
'total_epochs': opt.epochs,
'fitness_score': fitness_score
})
model_artifact.add_file(str(path / 'last.pt'), name='last.pt')
wandb.log_artifact(model_artifact,
aliases=['latest', 'epoch ' + str(self.current_epoch), 'best' if best_model else ''])
print("Saving model artifact on epoch ", epoch + 1)
def test_add_named_dir(runner):
with runner.isolated_filesystem():
open("file1.txt", "w").write("hello")
artifact = wandb.Artifact(type="dataset", name="my-arty")
artifact.add_dir(".", name="subdir")
assert artifact.digest == "a757208d042e8627b2970d72a71bed5b"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["subdir/file1.txt"] == {
"digest": "XUFAKrxLKna5cZ2REBfFkg==",
"size": 5,
}
def test_table_slice_reference_artifact():
with wandb.init() as run:
artifact = wandb.Artifact("table_data", "data")
table = _make_wandb_table()
artifact.add(table, "table")
run.log_artifact(artifact)
with wandb.init() as run:
artifact_1 = run.use_artifact("table_data:latest")
t1 = artifact_1.get("table")
artifact = wandb.Artifact("intermediate_data", "data")
i1 = wandb.Table(t1.columns, t1.data[:1])
i2 = wandb.Table(t1.columns, t1.data[1:])
artifact.add(i1, "table1")
artifact.add(i2, "table2")
run.log_artifact(artifact)
with wandb.init() as run:
artifact_2 = run.use_artifact("intermediate_data:latest")
i1 = artifact_2.get("table1")
i2 = artifact_2.get("table2")
artifact = wandb.Artifact("reference_data", "data")
table1 = wandb.Table(t1.columns, i1.data)
table2 = wandb.Table(t1.columns, i2.data)
artifact.add(table1, "table1")
artifact.add(table2, "table2")
run.log_artifact(artifact)
_cleanup()
with wandb.init() as run:
artifact_3 = run.use_artifact("reference_data:latest")
table1 = artifact_3.get("table1")
table2 = artifact_3.get("table2")
assert not os.path.isdir(os.path.join(artifact_2._default_root()))
# assert os.path.islink(os.path.join(artifact_3._default_root(), "media", "images", "test.png"))
# assert os.path.islink(os.path.join(artifact_3._default_root(), "media", "images", "test2.png"))
assert t1.data[:1] == table1.data
assert t1.data[1:] == table2.data
def test_artifact_finish_distributed_id(runner, live_mock_server,
test_settings):
# NOTE: these tests are against a mock server so they are testing the internal flows, but
# not the actual data transfer.
artifact_name = "distributed_artifact_{}".format(round(time.time()))
group_name = "test_group_{}".format(round(np.random.rand()))
artifact_type = "dataset"
# Finish with a distributed_id should succeed
run = wandb.init(settings=test_settings)
artifact = wandb.Artifact(artifact_name, type=artifact_type)
run.finish_artifact(artifact, distributed_id=group_name)
run.finish()
def save_model_to_artifact(model, path, name, artifact_path="final_model"):
"""During a wandb.Run, save a model to path and as a wandb.Artifact
and returns the resulting Artifact's complete identifier.
See PyTorch documentation for details on saving and loading models:
https://pytorch.org/tutorials/beginner/saving_loading_models.html
"""
model_artifact = wandb.Artifact(name=name, type="trained-model")
torch.save(model.state_dict(), path)
model_artifact.add_file(path, artifact_path)
wandb.run.log_artifact(model_artifact)
return "/".join([wandb.run.entity, wandb.run.project, model_artifact.name])
def on_train_end(self, trainer, pl_module):
logger = get_wandb_logger(trainer=trainer)
experiment = logger.experiment
ckpts = wandb.Artifact("experiment-ckpts", type="checkpoints")
if self.upload_best_only:
ckpts.add_file(trainer.checkpoint_callback.best_model_path)
else:
for path in glob.glob(os.path.join(self.ckpt_dir, "**/*.ckpt"), recursive=True):
ckpts.add_file(path)
experiment.use_artifact(ckpts)
def store_model_artifact(path: str, name: str):
"""
Store a tf model directory as a WandB artifact
Args:
path: Path to tensorflow saved model (e.g., /path/to/model.tf/)
name: name for the WandB artifact. If it already exists a new
version is stored
"""
model_artifact = wandb.Artifact(name, type="model")
model_artifact.add_dir(path)
wandb.log_artifact(model_artifact)
def test_add_reference_named_local_file(runner):
with runner.isolated_filesystem():
open("file1.txt", "w").write("hello")
artifact = wandb.Artifact(type="dataset", name="my-arty")
artifact.add_reference("file://file1.txt", name="great-file.txt")
assert artifact.digest == "585b9ada17797e37c9cbab391e69b8c5"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["great-file.txt"] == {
"digest": "XUFAKrxLKna5cZ2REBfFkg==",
"ref": "file://file1.txt",
"size": 5,
}
def test_audio_refs():
audioObj = wandb.Audio(
"https://wandb-artifacts-refs-public-test.s3-us-west-2.amazonaws.com/StarWars3.wav"
)
art = wandb.Artifact("audio_ref_test", "dataset")
art.add(audioObj, "audio_ref")
audio_expected = {
"_type": "audio-file",
"caption": None,
}
assert utils.subdict(audioObj.to_json(art),
audio_expected) == audio_expected
def setup_training(self, opt):
"""
Setup the necessary processes for training YOLO models:
- Attempt to download model checkpoint and dataset artifacts if opt.resume stats with WANDB_ARTIFACT_PREFIX
- Update data_dict, to contain info of previous run if resumed and the paths of dataset artifact if downloaded
- Setup log_dict, initialize bbox_interval
arguments:
opt (namespace) -- commandline arguments for this run
"""
self.log_dict, self.current_epoch = {}, 0
self.bbox_interval = opt.bbox_interval
if isinstance(opt.resume, str):
modeldir, _ = self.download_model_artifact(opt)
if modeldir:
self.weights = Path(modeldir) / "last.pt"
config = self.wandb_run.config
opt.weights, opt.save_period, opt.batch_size, opt.bbox_interval, opt.epochs, opt.hyp = str(
self.weights), config.save_period, config.batch_size, config.bbox_interval, config.epochs, \
config.hyp
data_dict = self.data_dict
if self.val_artifact is None: # If --upload_dataset is set, use the existing artifact, don't download
self.train_artifact_path, self.train_artifact = self.download_dataset_artifact(
data_dict.get('train'), opt.artifact_alias)
self.val_artifact_path, self.val_artifact = self.download_dataset_artifact(
data_dict.get('val'), opt.artifact_alias)
if self.train_artifact_path is not None:
train_path = Path(self.train_artifact_path) / 'data/images/'
data_dict['train'] = str(train_path)
if self.val_artifact_path is not None:
val_path = Path(self.val_artifact_path) / 'data/images/'
data_dict['val'] = str(val_path)
if self.val_artifact is not None:
self.result_artifact = wandb.Artifact(
"run_" + wandb.run.id + "_progress", "evaluation")
columns = ["epoch", "id", "ground truth", "prediction"]
columns.extend(self.data_dict['names'])
self.result_table = wandb.Table(columns)
self.val_table = self.val_artifact.get("val")
if self.val_table_path_map is None:
self.map_val_table_path()
if opt.bbox_interval == -1:
self.bbox_interval = opt.bbox_interval = (
opt.epochs // 10) if opt.epochs > 10 else 1
train_from_artifact = self.train_artifact_path is not None and self.val_artifact_path is not None
# Update the the data_dict to point to local artifacts dir
if train_from_artifact:
self.data_dict = data_dict
def test_artifact_references_internal(
runner,
mocked_run,
mock_server,
internal_sm,
backend_interface,
parse_ctx,
test_settings,
):
with runner.isolated_filesystem():
mock_server.set_context("max_cli_version", "0.11.0")
run = wandb.init(settings=test_settings)
t1 = wandb.Table(columns=[], data=[])
art = wandb.Artifact("A", "dataset")
art.add(t1, "t1")
run.log_artifact(art)
run.finish()
art = wandb.Artifact("A_PENDING", "dataset")
art.add(t1, "t1")
with backend_interface() as interface:
proto_run = interface._make_run(mocked_run)
r = internal_sm.send_run(interface._make_record(run=proto_run))
proto_artifact = interface._make_artifact(art)
proto_artifact.run_id = proto_run.run_id
proto_artifact.project = proto_run.project
proto_artifact.entity = proto_run.entity
proto_artifact.user_created = False
proto_artifact.use_after_commit = False
proto_artifact.finalize = True
for alias in ["latest"]:
proto_artifact.aliases.append(alias)
log_artifact = pb.LogArtifactRequest()
log_artifact.artifact.CopyFrom(proto_artifact)
internal_sm.send_artifact(log_artifact)
def test_artifact_run_lookup_apis():
artifact_1_name = "a1-{}".format(str(time.time()))
artifact_2_name = "a2-{}".format(str(time.time()))
# Initial setup
run_1 = wandb.init()
artifact = wandb.Artifact(artifact_1_name, "test_type")
artifact.add(wandb.Image(np.random.randint(0, 255, (10, 10))), "image")
run_1.log_artifact(artifact)
artifact = wandb.Artifact(artifact_2_name, "test_type")
artifact.add(wandb.Image(np.random.randint(0, 255, (10, 10))), "image")
run_1.log_artifact(artifact)
run_1.finish()
# Create a second version for a1
run_2 = wandb.init()
artifact = wandb.Artifact(artifact_1_name, "test_type")
artifact.add(wandb.Image(np.random.randint(0, 255, (10, 10))), "image")
run_2.log_artifact(artifact)
run_2.finish()
# Use both
run_3 = wandb.init()
a1 = run_3.use_artifact(artifact_1_name + ":latest")
assert _runs_eq(a1.used_by(), [run_3])
assert _run_eq(a1.logged_by(), run_2)
a2 = run_3.use_artifact(artifact_2_name + ":latest")
assert _runs_eq(a2.used_by(), [run_3])
assert _run_eq(a2.logged_by(), run_1)
run_3.finish()
# Use both
run_4 = wandb.init()
a1 = run_4.use_artifact(artifact_1_name + ":latest")
assert _runs_eq(a1.used_by(), [run_3, run_4])
a2 = run_4.use_artifact(artifact_2_name + ":latest")
assert _runs_eq(a2.used_by(), [run_3, run_4])
run_4.finish()
def test_artifact_creation_with_diff_type():
artifact_name = "a1-{}".format(str(time.time()))
# create
with wandb.init() as run:
artifact = wandb.Artifact(artifact_name, "artifact_type_1")
artifact.add(wandb.Image(np.random.randint(0, 255, (10, 10))), "image")
run.log_artifact(artifact)
# update
with wandb.init() as run:
artifact = wandb.Artifact(artifact_name, "artifact_type_1")
artifact.add(wandb.Image(np.random.randint(0, 255, (10, 10))), "image")
run.log_artifact(artifact)
# invalid
with wandb.init() as run:
artifact = wandb.Artifact(artifact_name, "artifact_type_2")
artifact.add(wandb.Image(np.random.randint(0, 255, (10, 10))),
"image_2")
did_err = False
try:
run.log_artifact(artifact)
except ValueError as err:
did_err = True
assert str(
err
) == "Expected artifact type artifact_type_1, got artifact_type_2"
assert did_err
with wandb.init() as run:
artifact = run.use_artifact(artifact_name + ":latest")
# should work
image = artifact.get("image")
assert image is not None
# should not work
image_2 = artifact.get("image_2")
assert image_2 is None
def test_add_http_reference_path(runner):
with runner.isolated_filesystem():
artifact = wandb.Artifact(type="dataset", name="my-arty")
mock_http(artifact, headers={"ETag": '"abc"', "Content-Length": "256",})
artifact.add_reference("http://example.com/file1.txt")
assert artifact.digest == "48237ccc050a88af9dcd869dd5a7e9f4"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["file1.txt"] == {
"digest": "abc",
"ref": "http://example.com/file1.txt",
"size": 256,
"extra": {"etag": '"abc"',},
}
def test_deduplicate_wbimagemask_from_array(runner):
test_folder = os.path.dirname(os.path.realpath(__file__))
im_data_1 = np.random.randint(0, 10, (300, 300))
im_data_2 = np.random.randint(0, 10, (300, 300))
with runner.isolated_filesystem():
artifact = wandb.Artifact(type="dataset", name="artifact")
wb_imagemask_1 = data_types.ImageMask({"mask_data": im_data_1},
key="test")
wb_imagemask_2 = data_types.ImageMask({"mask_data": im_data_2},
key="test2")
artifact.add(wb_imagemask_1, "my-imagemask_1")
artifact.add(wb_imagemask_2, "my-imagemask_2")
assert len(artifact.manifest.entries) == 4
with runner.isolated_filesystem():
artifact = wandb.Artifact(type="dataset", name="artifact")
wb_imagemask_1 = data_types.ImageMask({"mask_data": im_data_1},
key="test")
wb_imagemask_2 = data_types.ImageMask({"mask_data": im_data_1},
key="test2")
artifact.add(wb_imagemask_1, "my-imagemask_1")
artifact.add(wb_imagemask_2, "my-imagemask_2")
assert len(artifact.manifest.entries) == 3
def test_add_gs_reference_object(runner, mocker):
with runner.isolated_filesystem():
artifact = wandb.Artifact(type="dataset", name="my-arty")
mock_gcs(artifact)
artifact.add_reference("gs://my-bucket/my_object.pb")
assert artifact.digest == "8aec0d6978da8c2b0bf5662b3fd043a4"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["my_object.pb"] == {
"digest": "1234567890abcde",
"ref": "gs://my-bucket/my_object.pb",
"extra": {"etag": "1234567890abcde", "versionID": "1"},
"size": 10,
}
def test_add_gs_reference_object_with_name(runner, mocker):
with runner.isolated_filesystem():
artifact = wandb.Artifact(type="dataset", name="my-arty")
mock_gcs(artifact)
artifact.add_reference("gs://my-bucket/my_object.pb", name="renamed.pb")
assert artifact.digest == "bd85fe009dc9e408a5ed9b55c95f47b2"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["renamed.pb"] == {
"digest": "1234567890abcde",
"ref": "gs://my-bucket/my_object.pb",
"extra": {"etag": "1234567890abcde", "versionID": "1"},
"size": 10,
}
def _scan_and_log_checkpoints(
self,
checkpoint_callback: "ReferenceType[ModelCheckpoint]") -> None:
# get checkpoints to be saved with associated score
checkpoints = {
checkpoint_callback.last_model_path:
checkpoint_callback.current_score,
checkpoint_callback.best_model_path:
checkpoint_callback.best_model_score,
**checkpoint_callback.best_k_models,
}
checkpoints = sorted((Path(p).stat().st_mtime, p, s)
for p, s in checkpoints.items()
if Path(p).is_file())
checkpoints = [
c for c in checkpoints
if c[1] not in self._logged_model_time.keys()
or self._logged_model_time[c[1]] < c[0]
]
# log iteratively all new checkpoints
for t, p, s in checkpoints:
metadata = ({
"score": s,
"original_filename": Path(p).name,
"ModelCheckpoint": {
k: getattr(checkpoint_callback, k)
for k in [
"monitor",
"mode",
"save_last",
"save_top_k",
"save_weights_only",
"_every_n_train_steps",
"_every_n_val_epochs",
]
# ensure it does not break if `ModelCheckpoint` args change
if hasattr(checkpoint_callback, k)
},
} if _WANDB_GREATER_EQUAL_0_10_22 else None)
artifact = wandb.Artifact(name=f"model-{self.experiment.id}",
type="model",
metadata=metadata)
artifact.add_file(p, name="model.ckpt")
aliases = [
"latest", "best"
] if p == checkpoint_callback.best_model_path else ["latest"]
self.experiment.log_artifact(artifact, aliases=aliases)
# remember logged models - timestamp needed in case filename didn't change (lastkckpt or custom name)
self._logged_model_time[p] = t
def test_artifact_upsert_group_id(runner, live_mock_server, test_settings):
# NOTE: these tests are against a mock server so they are testing the internal flows, but
# not the actual data transfer.
artifact_name = "distributed_artifact_{}".format(round(time.time()))
group_name = "test_group_{}".format(round(np.random.rand()))
artifact_type = "dataset"
# Upsert with a group should succeed
run = wandb.init(group=group_name, settings=test_settings)
artifact = wandb.Artifact(name=artifact_name, type=artifact_type)
image = wandb.Image(np.random.randint(0, 255, (10, 10)))
artifact.add(image, "image_1")
run.upsert_artifact(artifact)
run.finish()
def log_plot_as_wandb_artifact(wand_run,
fig,
fig_name,
temp_dir=Path(
tempfile.NamedTemporaryFile().name)):
temp_dir.mkdir(exist_ok=True)
fig_name = f"{fig_name}.html"
filepath = temp_dir / fig_name
fig.write_html(open(filepath, "w"))
artifact = wandb.Artifact("run_" + wand_run.id + fig_name, type='result')
# Add a file to the artifact's contents
artifact.add_file(filepath)
# Save the artifact version to W&B and mark it as the output of this run
wand_run.log_artifact(artifact)
def test_artifact_finish_no_id(runner, live_mock_server, test_settings):
with runner.isolated_filesystem():
# NOTE: these tests are against a mock server so they are testing the internal flows, but
# not the actual data transfer.
artifact_name = "distributed_artifact_{}".format(round(time.time()))
group_name = "test_group_{}".format(round(np.random.rand()))
artifact_type = "dataset"
# Finish without a distributed_id should fail
run = wandb.init(settings=test_settings)
artifact = wandb.Artifact(artifact_name, type=artifact_type)
with pytest.raises(TypeError):
run.finish_artifact(artifact)
run.finish()
def test_add_reference_local_file_no_checksum(runner):
with runner.isolated_filesystem():
open("file1.txt", "w").write("hello")
size = os.path.getsize("file1.txt")
artifact = wandb.Artifact(type="dataset", name="my-arty")
artifact.add_reference("file://file1.txt", checksum=False)
assert artifact.digest == "415f3bca4b095cbbbbc47e0d44079e05"
manifest = artifact.manifest.to_manifest_json()
assert manifest["contents"]["file1.txt"] == {
"digest": md5_string(str(size)),
"ref": "file://file1.txt",
"size": size,
}
def log_artifact(
self,
tag: str,
artifact: object = None,
path_to_artifact: str = None,
scope: str = None,
# experiment info
run_key: str = None,
global_epoch_step: int = 0,
global_batch_step: int = 0,
global_sample_step: int = 0,
# stage info
stage_key: str = None,
stage_epoch_len: int = 0,
stage_epoch_step: int = 0,
stage_batch_step: int = 0,
stage_sample_step: int = 0,
# loader info
loader_key: str = None,
loader_batch_len: int = 0,
loader_sample_len: int = 0,
loader_batch_step: int = 0,
loader_sample_step: int = 0,
) -> None:
"""Logs artifact (arbitrary file like audio, video, model weights) to the logger."""
if artifact is None and path_to_artifact is None:
ValueError("Both artifact and path_to_artifact cannot be None")
artifact = wandb.Artifact(
name=self.run.id + "_aritfacts",
type="artifact",
metadata={
"stage_key": stage_key,
"loader_key": loader_key,
"scope": scope,
},
)
if artifact:
art_file_dir = os.path.join("wandb", self.run.id, "artifact_dumps")
os.makedirs(art_file_dir, exist_ok=True)
art_file = open(os.path.join(art_file_dir, tag), "wb")
pickle.dump(artifact, art_file)
art_file.close()
artifact.add_file(str(os.path.join(art_file_dir, tag)))
else:
artifact.add_file(path_to_artifact)
self.run.log_artifact(artifact)
def test_image_reference_with_preferred_path():
orig_im_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "..", "assets", "test.png")
orig_im_path_2 = os.path.join(os.path.dirname(os.path.realpath(__file__)), "..", "assets", "test2.png")
desired_artifact_path = "images/sample.png"
with wandb.init(project=WANDB_PROJECT) as run:
artifact = wandb.Artifact("artifact_1", type="test_artifact")
# manually add the image to a desired path
artifact.add_file(orig_im_path, desired_artifact_path)
# create an image that uses this image (it should be smart enough not to add the image twice)
image = wandb.Image(orig_im_path)
image_2 = wandb.Image(orig_im_path_2) # this one does not have the path preadded
# add the image to the table
table = wandb.Table(["image"], data=[[image],[image_2]])
# add the table to the artifact
artifact.add(table, "table")
run.log_artifact(artifact)
_cleanup()
with wandb.init(project=WANDB_PROJECT) as run:
artifact_1 = run.use_artifact("artifact_1:latest")
original_table = artifact_1.get("table")
artifact = wandb.Artifact("artifact_2", type="test_artifact")
# add the image by reference
image = wandb.Image(original_table.data[0][0])
image_2 = wandb.Image(original_table.data[1][0])
# add the image to the table
table = wandb.Table(["image"], data=[[image],[image_2]])
# add the table to the artifact
artifact.add(table, "table")
run.log_artifact(artifact)
_cleanup()
with wandb.init(project=WANDB_PROJECT) as run:
artifact_2 = run.use_artifact("artifact_2:latest")
artifact_2.download()
def create_dataset_table(self, dataset, class_to_id, name='dataset'):
# TODO: Explore multiprocessing to slpit this loop parallely| This is essential for speeding up the the logging
artifact = wandb.Artifact(name=name, type="dataset")
img_files = tqdm([dataset.path]) if isinstance(
dataset.path, str) and Path(dataset.path).is_dir() else None
img_files = tqdm(dataset.img_files) if not img_files else img_files
for img_file in img_files:
if Path(img_file).is_dir():
artifact.add_dir(img_file, name='data/images')
labels_path = 'labels'.join(dataset.path.rsplit('images', 1))
artifact.add_dir(labels_path, name='data/labels')
else:
artifact.add_file(img_file,
name='data/images/' + Path(img_file).name)
label_file = Path(img2label_paths([img_file])[0])
artifact.add_file(
str(label_file), name='data/labels/' +
label_file.name) if label_file.exists() else None
table = wandb.Table(columns=["id", "train_image", "Classes", "name"])
class_set = wandb.Classes([{
'id': id,
'name': name
} for id, name in class_to_id.items()])
for si, (img, labels, paths, shapes) in enumerate(tqdm(dataset)):
box_data, img_classes = [], {}
for cls, *xywh in labels[:, 1:].tolist():
cls = int(cls)
box_data.append({
"position": {
"middle": [xywh[0], xywh[1]],
"width": xywh[2],
"height": xywh[3]
},
"class_id": cls,
"box_caption": "%s" % (class_to_id[cls])
})
img_classes[cls] = class_to_id[cls]
boxes = {
"ground_truth": {
"box_data": box_data,
"class_labels": class_to_id
}
} # inference-space
table.add_data(si,
wandb.Image(paths, classes=class_set, boxes=boxes),
list(img_classes.values()),
Path(paths).name)
artifact.add(table, name)
return artifact
def log_dataset_artifact(self, dataset, class_to_id, name='dataset'):
artifact = wandb.Artifact(name=name, type="dataset")
image_path = dataset.path
artifact.add_dir(image_path, name='data/images')
table = wandb.Table(columns=["id", "train_image", "Classes"])
class_set = wandb.Classes([{
'id': id,
'name': name
} for id, name in class_to_id.items()])
for si, (img, labels, paths, shapes) in enumerate(dataset):
height, width = shapes[0]
labels[:, 2:] = (xywh2xyxy(labels[:, 2:].view(-1, 4)))
labels[:, 2:] *= torch.Tensor([width, height, width, height])
box_data = []
img_classes = {}
for cls, *xyxy in labels[:, 1:].tolist():
cls = int(cls)
box_data.append({
"position": {
"minX": xyxy[0],
"minY": xyxy[1],
"maxX": xyxy[2],
"maxY": xyxy[3]
},
"class_id": cls,
"box_caption": "%s" % (class_to_id[cls]),
"scores": {
"acc": 1
},
"domain": "pixel"
})
img_classes[cls] = class_to_id[cls]
boxes = {
"ground_truth": {
"box_data": box_data,
"class_labels": class_to_id
}
} # inference-space
table.add_data(si,
wandb.Image(paths, classes=class_set, boxes=boxes),
json.dumps(img_classes))
artifact.add(table, name)
labels_path = 'labels'.join(image_path.rsplit('images', 1))
zip_path = Path(labels_path).parent / (name + '_labels.zip')
if not zip_path.is_file(): # make_archive won't check if file exists
shutil.make_archive(zip_path.with_suffix(''), 'zip', labels_path)
artifact.add_file(str(zip_path), name='data/labels.zip')
wandb.log_artifact(artifact)
print("Saving data to W&B...")
def log_model(path, name=None, metadata={}):
"Log model file"
if wandb.run is None:
raise ValueError('You must call wandb.init() before log_model()')
path = Path(path)
if not path.is_file():
raise f'path must be a valid file: {path}'
name = ifnone(name, f'run-{wandb.run.id}-model')
_format_metadata(metadata)
artifact_model = wandb.Artifact(name=name,
type='model',
description='trained model',
metadata=metadata)
artifact_model.add_file(str(path.resolve()))
wandb.run.log_artifact(artifact_model)
def test_add_obj_wbimage_no_classes(runner):
test_folder = os.path.dirname(os.path.realpath(__file__))
im_path = os.path.join(test_folder, "..", "assets", "2x2.png")
with runner.isolated_filesystem():
artifact = wandb.Artifact(type="dataset", name="my-arty")
wb_image = wandb.Image(
im_path,
masks={
"ground_truth": {
"path": os.path.join(test_folder, "..", "assets", "2x2.png"),
},
},
)
with pytest.raises(ValueError):
artifact.add(wb_image, "my-image")
