Skip to content

Commit

Permalink
W&B: refactor W&B tables (ultralytics#5737)
Browse files Browse the repository at this point in the history 
* update

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* reformat

* Single-line argparser argument

* Update README.md

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update README.md

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Glenn Jocher <[email protected]>
  • Loading branch information
@AyushExel @pre-commit-ci @glenn-jocher
3 people authored Nov 25, 2021
1 parent e8b4ce9 commit ac3daa6
Show file tree
Hide file tree
Showing 3 changed files with 68 additions and 35 deletions.
2 changes: 1 addition & 1 deletion train.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -475,7 +475,7 @@ def parse_opt(known=False):

# Weights & Biases arguments
parser.add_argument('--entity', default=None, help='W&B: Entity')
parser.add_argument('--upload_dataset', action='store_true', help='W&B: Upload dataset as artifact table')
parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='W&B: Upload data, "val" option')
parser.add_argument('--bbox_interval', type=int, default=-1, help='W&B: Set bounding-box image logging interval')
parser.add_argument('--artifact_alias', type=str, default='latest', help='W&B: Version of dataset artifact to use')

Expand Down
33 changes: 19 additions & 14 deletions utils/loggers/wandb/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
* [About Weights & Biases](#about-weights-&-biases)
* [First-Time Setup](#first-time-setup)
* [Viewing runs](#viewing-runs)
* [Disabling wandb](#disabling-wandb)
* [Advanced Usage: Dataset Versioning and Evaluation](#advanced-usage)
* [Reports: Share your work with the world!](#reports)

Expand Down Expand Up @@ -49,39 +50,44 @@ Run information streams from your environment to the W&B cloud console as you tr
* Environment: OS and Python types, Git repository and state, **training command**

<p align="center"><img width="900" alt="Weights & Biases dashboard" src="https://user-images.githubusercontent.com/26833433/135390767-c28b050f-8455-4004-adb0-3b730386e2b2.png"></p>
</details>

## Disabling wandb
* training after running `wandb disabled` inside that directory creates no wandb run
![Screenshot (84)](https://user-images.githubusercontent.com/15766192/143441777-c780bdd7-7cb4-4404-9559-b4316030a985.png)

</details>
* To enable wandb again, run `wandb online`
![Screenshot (85)](https://user-images.githubusercontent.com/15766192/143441866-7191b2cb-22f0-4e0f-ae64-2dc47dc13078.png)

## Advanced Usage
You can leverage W&B artifacts and Tables integration to easily visualize and manage your datasets, models and training evaluations. Here are some quick examples to get you started.
<details open>
<h3>1. Visualize and Version Datasets</h3>
Log, visualize, dynamically query, and understand your data with <a href='https://docs.wandb.ai/guides/data-vis/tables'>W&B Tables</a>. You can use the following command to log your dataset as a W&B Table. This will generate a <code>{dataset}_wandb.yaml</code> file which can be used to train from dataset artifact.
<details>
<h3> 1: Train and Log Evaluation simultaneousy </h3>
This is an extension of the previous section, but it'll also training after uploading the dataset. <b> This also evaluation Table</b>
Evaluation table compares your predictions and ground truths across the validation set for each epoch. It uses the references to the already uploaded datasets,
so no images will be uploaded from your system more than once.
<details open>
<summary> <b>Usage</b> </summary>
<b>Code</b> <code> $ python utils/logger/wandb/log_dataset.py --project ... --name ... --data .. </code>
<b>Code</b> <code> $ python train.py --upload_data val</code>

![Screenshot (64)](https://user-images.githubusercontent.com/15766192/128486078-d8433890-98a3-4d12-8986-b6c0e3fc64b9.png)
![Screenshot from 2021-11-21 17-40-06](https://user-images.githubusercontent.com/15766192/142761183-c1696d8c-3f38-45ab-991a-bb0dfd98ae7d.png)
</details>

<h3> 2: Train and Log Evaluation simultaneousy </h3>
This is an extension of the previous section, but it'll also training after uploading the dataset. <b> This also evaluation Table</b>
Evaluation table compares your predictions and ground truths across the validation set for each epoch. It uses the references to the already uploaded datasets,
so no images will be uploaded from your system more than once.
<h3>2. Visualize and Version Datasets</h3>
Log, visualize, dynamically query, and understand your data with <a href='https://docs.wandb.ai/guides/data-vis/tables'>W&B Tables</a>. You can use the following command to log your dataset as a W&B Table. This will generate a <code>{dataset}_wandb.yaml</code> file which can be used to train from dataset artifact.
<details>
<summary> <b>Usage</b> </summary>
<b>Code</b> <code> $ python utils/logger/wandb/log_dataset.py --data .. --upload_data </code>
<b>Code</b> <code> $ python utils/logger/wandb/log_dataset.py --project ... --name ... --data .. </code>

![Screenshot (72)](https://user-images.githubusercontent.com/15766192/128979739-4cf63aeb-a76f-483f-8861-1c0100b938a5.png)
![Screenshot (64)](https://user-images.githubusercontent.com/15766192/128486078-d8433890-98a3-4d12-8986-b6c0e3fc64b9.png)
</details>

<h3> 3: Train using dataset artifact </h3>
When you upload a dataset as described in the first section, you get a new config file with an added `_wandb` to its name. This file contains the information that
can be used to train a model directly from the dataset artifact. <b> This also logs evaluation </b>
<details>
<summary> <b>Usage</b> </summary>
<b>Code</b> <code> $ python utils/logger/wandb/log_dataset.py --data {data}_wandb.yaml </code>
<b>Code</b> <code> $ python train.py --data {data}_wandb.yaml </code>

![Screenshot (72)](https://user-images.githubusercontent.com/15766192/128979739-4cf63aeb-a76f-483f-8861-1c0100b938a5.png)
</details>
Expand Down Expand Up @@ -123,7 +129,6 @@ Any run can be resumed using artifacts if the <code>--resume</code> argument sta

</details>


<h3> Reports </h3>
W&B Reports can be created from your saved runs for sharing online. Once a report is created you will receive a link you can use to publically share your results. Here is an example report created from the COCO128 tutorial trainings of all four YOLOv5 models ([link](https://wandb.ai/glenn-jocher/yolov5_tutorial/reports/YOLOv5-COCO128-Tutorial-Results--VmlldzozMDI5OTY)).

Expand Down
68 changes: 48 additions & 20 deletions utils/loggers/wandb/wandb_utils.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -202,7 +202,6 @@ def check_and_upload_dataset(self, opt):
config_path = self.log_dataset_artifact(opt.data,
opt.single_cls,
'YOLOv5' if opt.project == 'runs/train' else Path(opt.project).stem)
LOGGER.info(f"Created dataset config file {config_path}")
with open(config_path, errors='ignore') as f:
wandb_data_dict = yaml.safe_load(f)
return wandb_data_dict
Expand Down Expand Up @@ -244,7 +243,9 @@ def setup_training(self, opt):

if self.val_artifact is not None:
self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation")
self.result_table = wandb.Table(["epoch", "id", "ground truth", "prediction", "avg_confidence"])
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()
Expand Down Expand Up @@ -331,28 +332,41 @@ def log_dataset_artifact(self, data_file, single_cls, project, overwrite_config=
returns:
the new .yaml file with artifact links. it can be used to start training directly from artifacts
"""
upload_dataset = self.wandb_run.config.upload_dataset
log_val_only = isinstance(upload_dataset, str) and upload_dataset == 'val'
self.data_dict = check_dataset(data_file) # parse and check
data = dict(self.data_dict)
nc, names = (1, ['item']) if single_cls else (int(data['nc']), data['names'])
names = {k: v for k, v in enumerate(names)} # to index dictionary
self.train_artifact = self.create_dataset_table(LoadImagesAndLabels(
data['train'], rect=True, batch_size=1), names, name='train') if data.get('train') else None

# log train set
if not log_val_only:
self.train_artifact = self.create_dataset_table(LoadImagesAndLabels(
data['train'], rect=True, batch_size=1), names, name='train') if data.get('train') else None
if data.get('train'):
data['train'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'train')

self.val_artifact = self.create_dataset_table(LoadImagesAndLabels(
data['val'], rect=True, batch_size=1), names, name='val') if data.get('val') else None
if data.get('train'):
data['train'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'train')
if data.get('val'):
data['val'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'val')
path = Path(data_file).stem
path = (path if overwrite_config else path + '_wandb') + '.yaml' # updated data.yaml path
data.pop('download', None)
data.pop('path', None)
with open(path, 'w') as f:
yaml.safe_dump(data, f)

path = Path(data_file)
# create a _wandb.yaml file with artifacts links if both train and test set are logged
if not log_val_only:
path = (path.stem if overwrite_config else path.stem + '_wandb') + '.yaml' # updated data.yaml path
path = Path('data') / path
data.pop('download', None)
data.pop('path', None)
with open(path, 'w') as f:
yaml.safe_dump(data, f)
LOGGER.info(f"Created dataset config file {path}")

if self.job_type == 'Training': # builds correct artifact pipeline graph
if not log_val_only:
self.wandb_run.log_artifact(
self.train_artifact) # calling use_artifact downloads the dataset. NOT NEEDED!
self.wandb_run.use_artifact(self.val_artifact)
self.wandb_run.use_artifact(self.train_artifact)
self.val_artifact.wait()
self.val_table = self.val_artifact.get('val')
self.map_val_table_path()
Expand All @@ -371,7 +385,7 @@ def map_val_table_path(self):
for i, data in enumerate(tqdm(self.val_table.data)):
self.val_table_path_map[data[3]] = data[0]

def create_dataset_table(self, dataset: LoadImagesAndLabels, class_to_id: Dict[int,str], name: str = 'dataset'):
def create_dataset_table(self, dataset: LoadImagesAndLabels, class_to_id: Dict[int, str], name: str = 'dataset'):
"""
Create and return W&B artifact containing W&B Table of the dataset.
Expand Down Expand Up @@ -424,23 +438,34 @@ def log_training_progress(self, predn, path, names):
"""
class_set = wandb.Classes([{'id': id, 'name': name} for id, name in names.items()])
box_data = []
total_conf = 0
avg_conf_per_class = [0] * len(self.data_dict['names'])
pred_class_count = {}
for *xyxy, conf, cls in predn.tolist():
if conf >= 0.25:
cls = int(cls)
box_data.append(
{"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]},
"class_id": int(cls),
"class_id": cls,
"box_caption": f"{names[cls]} {conf:.3f}",
"scores": {"class_score": conf},
"domain": "pixel"})
total_conf += conf
avg_conf_per_class[cls] += conf

if cls in pred_class_count:
pred_class_count[cls] += 1
else:
pred_class_count[cls] = 1

for pred_class in pred_class_count.keys():
avg_conf_per_class[pred_class] = avg_conf_per_class[pred_class] / pred_class_count[pred_class]

boxes = {"predictions": {"box_data": box_data, "class_labels": names}} # inference-space
id = self.val_table_path_map[Path(path).name]
self.result_table.add_data(self.current_epoch,
id,
self.val_table.data[id][1],
wandb.Image(self.val_table.data[id][1], boxes=boxes, classes=class_set),
total_conf / max(1, len(box_data))
*avg_conf_per_class
)

def val_one_image(self, pred, predn, path, names, im):
Expand Down Expand Up @@ -490,7 +515,8 @@ def end_epoch(self, best_result=False):
try:
wandb.log(self.log_dict)
except BaseException as e:
LOGGER.info(f"An error occurred in wandb logger. The training will proceed without interruption. More info\n{e}")
LOGGER.info(
f"An error occurred in wandb logger. The training will proceed without interruption. More info\n{e}")
self.wandb_run.finish()
self.wandb_run = None

Expand All @@ -502,7 +528,9 @@ def end_epoch(self, best_result=False):
('best' if best_result else '')])

wandb.log({"evaluation": self.result_table})
self.result_table = wandb.Table(["epoch", "id", "ground truth", "prediction", "avg_confidence"])
columns = ["epoch", "id", "ground truth", "prediction"]
columns.extend(self.data_dict['names'])
self.result_table = wandb.Table(columns)
self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation")

def finish_run(self):
Expand Down

0 comments on commit ac3daa6

Please sign in to comment.