VisualDL

https://visualdl.readthedocs.io/zh/latest/
Admin
VisualDL is a visualization analysis tool for PaddlePaddle that provides various charts to visualize...
Tokens:9,291
Snippets:107
Trust Score:9.6
Update:3 months ago
Show doc for...
Tokens:
Raw
### Clone FastDeploy Examples and Start Server

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_server/README.html

Commands to clone the FastDeploy repository and launch the server for testing.

```bash
git clone https://github.com/PaddlePaddle/FastDeploy.git
cd FastDeploy/examples
visualdl --host 0.0.0.0 --port 8080
```

--------------------------------

### Install VisualDL from source

Source: https://visualdl.readthedocs.io/zh/latest/README.html

Clone the repository and build the wheel file to install VisualDL from source.

```bash
git clone https://github.com/PaddlePaddle/VisualDL.git
cd VisualDL

python setup.py bdist_wheel
pip install --upgrade dist/visualdl-*.whl
```

--------------------------------

### Start the VisualDL server

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_client/README.html

Command to launch the VisualDL server with specified host and port settings.

```bash
visualdl --host 0.0.0.0 --port 8080
```

--------------------------------

### Basic VisualDL Launch Command

Source: https://visualdl.readthedocs.io/zh/latest/README.html

Start the VisualDL panel by pointing it to a specific log directory.

```bash
visualdl --logdir ./log
```

--------------------------------

### Install VisualDL

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_server/README.html

Command to install the VisualDL package within the container environment.

```bash
python -m pip install visualdl
```

--------------------------------

### Install VisualDL via pip

Source: https://visualdl.readthedocs.io/zh/latest/README.html

Install the VisualDL package using the pip command with a specified mirror.

```bash
python -m pip install visualdl -i https://mirror.baidu.com/pypi/simple
```

--------------------------------

### Launch VisualDL Server

Source: https://visualdl.readthedocs.io/zh/latest/components/profiler/README.html

Command to start the VisualDL web interface for visualizing the collected profiling data.

```bash
visualdl --logdir ./profiler_demo --port 8080
```

--------------------------------

### Record Embeddings Example

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Example demonstrating how to initialize a LogWriter and record embeddings using the add_embeddings method.

```python
from visualdl import LogWriter

if __name__ == '__main__':
    hot_vectors = [
        [1.3561076367500755, 1.3116267195134017, 1.6785401875616097],
        [1.1039614644440658, 1.8891609992484688, 1.32030488587171],
        [1.9924524852447711, 1.9358920727142739, 1.2124401279391606],
        [1.4129542689796446, 1.7372166387197474, 1.7317806077076527],
        [1.3913371800587777, 1.4684674577930312, 1.5214136352476377]]

    labels = ["label_1", "label_2", "label_3", "label_4", "label_5"]
    # initialize a recorder
    with LogWriter(logdir="./log/high_dimensional_test/train") as writer:
        # recorde a set of labels and corresponding hot_vectors to the recorder 
        writer.add_embeddings(tag='default',
                              labels=labels,
                              hot_vectors=hot_vectors)
```

--------------------------------

### Image Logging Demo

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Example script demonstrating how to initialize a LogWriter and record cropped images at each training step.

```python
import numpy as np
from PIL import Image
from visualdl import LogWriter

def random_crop(img):
    """get random 100x100 slices of image
    """
    img = Image.open(img)
    w, h = img.size
    random_w = np.random.randint(0, w - 100)
    random_h = np.random.randint(0, h - 100)
    r = img.crop((random_w, random_h, random_w + 100, random_h + 100))
    return np.asarray(r)

if __name__ == '__main__':
    # initialize a recorder
    with LogWriter(logdir="./log/image_test/train") as writer:
        for step in range(6):
            # add image data
            writer.add_image(tag="eye",
                             img=random_crop("../../docs/images/eye.jpg"),
                             step=step)
```

--------------------------------

### Logging Audio Data Example

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Demonstrates how to use LogWriter to record audio data from a WAV file.

```python
from visualdl import LogWriter
from scipy.io import wavfile

if __name__ == '__main__':
    with LogWriter(logdir="./log/audio_test/train") as writer:
        sample_rate, audio_data = wavfile.read('./test.wav')
        writer.add_audio(tag="audio_tag",
                         audio_array=audio_data,
                         step=0,
                         sample_rate=sample_rate)
```

--------------------------------

### Launch VisualDL Server

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Starts the VisualDL server on the specified port to access components like X2Paddle.

```bash
visualdl --port 8080
```

--------------------------------

### Run FastDeploy Container

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_server/README.html

Command to start a GPU-enabled FastDeploy container with host networking and volume mapping.

```bash
nvidia-docker run -it --net=host --name fd_serving -v `pwd`/:/FastDeploy registry.baidubce.com/paddlepaddle/fastdeploy:1.0.2-gpu-cuda11.4-trt8.4-21.10  bash
```

--------------------------------

### Implement and log a dynamic model

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Example of defining a neural network using paddle.nn.Layer and logging it with LogWriter.

```python
import paddle
import paddle.nn as nn
import paddle.nn.functional as F

from visualdl import LogWriter

class MyNet(nn.Layer):
    def __init__(self):
        super(MyNet, self).__init__()
        self.conv1 = nn.Conv2D(
            in_channels=1, out_channels=20, kernel_size=5, stride=1, padding=2)
        self.max_pool1 = nn.MaxPool2D(kernel_size=2, stride=2)
        self.conv2 = nn.Conv2D(
            in_channels=20,
            out_channels=20,
            kernel_size=5,
            stride=1,
            padding=2)
        self.max_pool2 = nn.MaxPool2D(kernel_size=2, stride=2)
        self.fc = nn.Linear(in_features=980, out_features=10)

    def forward(self, inputs):
        x = self.conv1(inputs)
        x = F.relu(x)
        x = self.max_pool1(x)
        x = self.conv2(x)
        x = F.relu(x)
        x = self.max_pool2(x)
        x = paddle.reshape(x, [x.shape[0], -1])
        x = self.fc(x)
        return x

net = MyNet()
with LogWriter(logdir="./log/graph_test/") as writer:
    writer.add_graph(
        model=net,
        input_spec=[paddle.static.InputSpec([-1, 1, 28, 28], 'float32')],
        verbose=True)
```

--------------------------------

### Record Hyperparameters and Metrics

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Example demonstrating how to log hyperparameter dictionaries and corresponding scalar metrics using LogWriter.

```python
from visualdl import LogWriter

# This demo demonstrates the hyperparameter records of two experiments. Take the first
# experiment data as an example, First, record the data of the hyperparameter `hparams`
# in the `add_hparams` interface. Then specify the name of `metrics` to be recorded later.
# Finally, use `add_scalar` to specifically record the data of `metrics`. Note that the
# `metrics_list` parameter in the `add_hparams` interface needs to include the `tag`
# parameter of the `add_scalar` interface.
if __name__ == '__main__':
    # Record the data of the first experiment
    with LogWriter('./log/hparams_test/train/run1') as writer:
        # Record the value of `hparams` and the name of `metrics`
        writer.add_hparams(hparams_dict={'lr': 0.1, 'bsize': 1, 'opt': 'sgd'},
                           metrics_list=['hparam/accuracy', 'hparam/loss'])
        # Record the metrics values of different steps in an experiment by matching
        # the `tag` in the `add_scalar` interface with `metrics_list` in `add_hparams` interface.
        for i in range(10):
            writer.add_scalar(tag='hparam/accuracy', value=i, step=i)
            writer.add_scalar(tag='hparam/loss', value=2*i, step=i)

    # Record the data of the second experiment
    with LogWriter('./log/hparams_test/train/run2') as writer:
        # Record the value of `hparams` and the name of `metrics`
        writer.add_hparams(hparams_dict={'lr': 0.2, 'bsize': 2, 'opt': 'relu'},
                           metrics_list=['hparam/accuracy', 'hparam/loss'])
        # Record the metrics values of different steps in an experiment by matching
        # the `tag` in the `add_scalar` interface with `metrics_list` in `add_hparams` interface.
        for i in range(10):
            writer.add_scalar(tag='hparam/accuracy', value=1.0/(i+1), step=i)
            writer.add_scalar(tag='hparam/loss', value=5*i, step=i)
```

--------------------------------

### Update VisualDL package

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Ensure the latest version of VisualDL is installed to access service features.

```bash
pip install visualdl --upgrade
```

--------------------------------

### Histogram Usage Example

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

A complete script demonstrating how to use LogWriter to record histogram data during a training loop.

```python
from visualdl import LogWriter
import numpy as np

if __name__ == '__main__':
    values = np.arange(0, 1000)
    with LogWriter(logdir="./log/histogram_test/train") as writer:
        for index in range(1, 101):
            interval_start = 1 + 2 * index / 100.0
            interval_end = 6 - 2 * index / 100.0
            data = np.random.uniform(interval_start, interval_end, size=(10000))
            writer.add_histogram(tag='default tag',
                                 values=data,
                                 step=index,
                                 buckets=10)
```

--------------------------------

### Launch VisualDL via Command Line

Source: https://visualdl.readthedocs.io/zh/latest/README.html

Use these parameters to configure the VisualDL server host, port, and component visibility.

```bash
visualdl --logdir <dir_1, dir_2, ... , dir_n> --host <host> --port <port> --cache-timeout <cache_timeout> --language <language> --public-path <public_path> --api-only --component_tabs <tab_name1, tab_name2, ...>
```

--------------------------------

### FastDeploy Server Launch Parameters

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_server/README.html

Configuration parameters used when launching a FastDeploy server service.

```text
server-name: The service name specified by the user, which is used to name the deployed service for subsequent viewing. If not filled, the default service name is the process number pid of the service process
model-repository: The model repository to be deployed, which is the currently loaded model repository.
backend-config: Specify the backend configuration of service, the format is <backend_name>,<setting>=<value>, where backend_name is the backend name, such as python, tensorrt, etc., followed by the specific configurable items of the backend. The default value is python, shm-default-byte-size=10485760
http-port: Specify the http service port of service, the default is 8000
grpc-port: Specify the grpc service port of service, the default is 8001
metrics-port: Specify the metrics query service port of service, the default is 8002
gpus: Specify the number of gpus that can be used by the deployed service. If card 2 and card 3 are used, it is "2,3". If not filled, all available gpus will be used
```

--------------------------------

### add_image(tag, img, step, walltime=None, dataformats="HWC")

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records an image to the log file for visualization during training.

```APIDOC
## add_image(tag, img, step, walltime=None, dataformats="HWC")

### Description
Records image data to a log file to track changes during training stages.

### Parameters
- **tag** (string) - Required - The name of the image data (e.g., 'train/loss'). Cannot contain '%'.
- **img** (numpy.ndarray) - Required - Image data in ndarray format. Channels can be 1, 3, or 4. Floating point data is clipped to [0, 1).
- **step** (int) - Required - The training step number.
- **walltime** (int) - Optional - The timestamp of the data. Defaults to current time.
- **dataformats** (string) - Optional - Format of the image (e.g., 'NCHW', 'NHWC', 'HWC', 'CHW', 'HW'). Defaults to 'HWC'.
```

--------------------------------

### Collect and Export Profiling Data with PaddlePaddle

Source: https://visualdl.readthedocs.io/zh/latest/components/profiler/README.html

Demonstrates how to use the paddle.profiler API to capture training metrics within a specific step range and export them to a directory.

```python
import paddle
import paddle.profiler as profiler
import paddle.nn.functional as F
from paddle.vision.transforms import ToTensor
import numpy as np

transform = ToTensor()
cifar10_train = paddle.vision.datasets.Cifar10(mode='train',
                                               transform=transform)
cifar10_test = paddle.vision.datasets.Cifar10(mode='test',
                                              transform=transform)
class MyNet(paddle.nn.Layer):
    def __init__(self, num_classes=1):
        super(MyNet, self).__init__()

        self.conv1 = paddle.nn.Conv2D(in_channels=3, out_channels=32, kernel_size=(3, 3))
        self.pool1 = paddle.nn.MaxPool2D(kernel_size=2, stride=2)

        self.conv2 = paddle.nn.Conv2D(in_channels=32, out_channels=64, kernel_size=(3,3))
        self.pool2 = paddle.nn.MaxPool2D(kernel_size=2, stride=2)

        self.conv3 = paddle.nn.Conv2D(in_channels=64, out_channels=64, kernel_size=(3,3))

        self.flatten = paddle.nn.Flatten()

        self.linear1 = paddle.nn.Linear(in_features=1024, out_features=64)
        self.linear2 = paddle.nn.Linear(in_features=64, out_features=num_classes)

    def forward(self, x):
        x = self.conv1(x)
        x = F.relu(x)
        x = self.pool1(x)

        x = self.conv2(x)
        x = F.relu(x)
        x = self.pool2(x)

        x = self.conv3(x)
        x = F.relu(x)

        x = self.flatten(x)
        x = self.linear1(x)
        x = F.relu(x)
        x = self.linear2(x)
        return x

epoch_num = 10
batch_size = 32
learning_rate = 0.001
val_acc_history = []
val_loss_history = []

def train(model):
    print('start training ... ')
    # turn into training mode
    model.train()

    opt = paddle.optimizer.Adam(learning_rate=learning_rate,
                                parameters=model.parameters())

    train_loader = paddle.io.DataLoader(cifar10_train,
                                        shuffle=True,
                                        batch_size=batch_size,
                                        num_workers=1)

    valid_loader = paddle.io.DataLoader(cifar10_test, batch_size=batch_size)

    # create profiler
    def my_on_trace_ready(prof): # callback function, will be called when profiler stops
      callback = profiler.export_chrome_tracing('./profiler_demo') # export profiling data
      callback(prof)  
      prof.summary(sorted_by=profiler.SortedKeys.GPUTotal) # print summary

    p = profiler.Profiler(scheduler = [3,14], on_trace_ready=my_on_trace_ready, timer_only=False)

    p.start() # start profiler

    for epoch in range(epoch_num):
        for batch_id, data in enumerate(train_loader()):
            x_data = data[0]
            y_data = paddle.to_tensor(data[1])
            y_data = paddle.unsqueeze(y_data, 1)

            logits = model(x_data)
            loss = F.cross_entropy(logits, y_data)

            if batch_id % 1000 == 0:
                print("epoch: {}, batch_id: {}, loss is: {}".format(epoch, batch_id, loss.numpy()))
            loss.backward()
            opt.step()
            opt.clear_grad()

            p.step() # signify profiler into next step
            if batch_id == 19:
              p.stop() # stop profiler
              exit()

        # evaluate model after one epoch
        model.eval()
        accuracies = []
        losses = []
        for batch_id, data in enumerate(valid_loader()):
            x_data = data[0]
            y_data = paddle.to_tensor(data[1])
            y_data = paddle.unsqueeze(y_data, 1)

            logits = model(x_data)
            loss = F.cross_entropy(logits, y_data)
            acc = paddle.metric.accuracy(logits, y_data)
            accuracies.append(acc.numpy())
            losses.append(loss.numpy())

        avg_acc, avg_loss = np.mean(accuracies), np.mean(losses)
        print("[validation] accuracy/loss: {}/{}".format(avg_acc, avg_loss))
        val_acc_history.append(avg_acc)
        val_loss_history.append(avg_loss)
        model.train()

model = MyNet(num_classes=10)
train(model)
```

--------------------------------

### add_audio(tag, audio_array, step, sample_rate)

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records audio data to the log writer for visualization in the VisualDL panel.

```APIDOC
## add_audio(tag, audio_array, step, sample_rate)

### Description
Records audio data during the training process to monitor speech recognition or text-to-speech tasks.

### Parameters
- **tag** (string) - Required - The name of the audio record (e.g., 'audio/sample'). Must not contain '%'.
- **audio_array** (numpy.ndarray) - Required - Audio data in ndarray format. Elements must be float values normalized in the range [-1, 1].
- **step** (int) - Required - The current training step.
- **sample_rate** (int) - Optional - The sample rate of the original audio. Defaults to 8000.
```

--------------------------------

### Record Scalar Data with LogWriter

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Initializes a LogWriter to record accuracy and loss metrics during training.

```python
from visualdl import LogWriter

if __name__ == '__main__':
    value = [i/1000.0 for i in range(1000)]
    # initialize a recorder
    with LogWriter(logdir="./log/scalar_test/train") as writer:
        for step in range(1000):
            # add accuracy with tag of 'acc' to the recorder
            writer.add_scalar(tag="acc", step=step, value=value[step])
            # add loss with tag of 'loss' to the recorder
            writer.add_scalar(tag="loss", step=step, value=1/(value[step] + 1))
```

--------------------------------

### Pull FastDeploy Docker Images

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_server/README.html

Commands to download the required FastDeploy Docker images for CPU or GPU environments.

```bash
docker pull registry.baidubce.com/paddlepaddle/fastdeploy:1.0.2-cpu-only-21.10
```

```bash
docker pull registry.baidubce.com/paddlepaddle/fastdeploy:1.0.2-gpu-cuda11.4-trt8.4-21.10
```

--------------------------------

### Upload visualization data

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Upload log directories and model files to generate a shareable visualization URL.

```bash
visualdl service upload --logdir ./log \
                        --model ./__model__
```

--------------------------------

### Audio Logging Interface Definition

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

The signature for the add_audio method used to record audio data.

```python
add_audio(tag, audio_array, step, sample_rate)
```

--------------------------------

### Image Recording Interface

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

The signature for the add_image method used to log image data to the log file.

```python
add_image(tag, img, step, walltime=None, dataformats="HWC")
```

--------------------------------

### Record scalar values with LogWriter

Source: https://visualdl.readthedocs.io/zh/latest/README.html

Initialize a LogWriter and record scalar values to a log file using the add_scalar method.

```python
from visualdl import LogWriter

# create a log file under `./log/scalar_test/train`
with LogWriter(logdir="./log/scalar_test/train") as writer:
    # use `add_scalar` to record scalar values
    writer.add_scalar(tag="acc", step=1, value=0.5678)
    writer.add_scalar(tag="acc", step=2, value=0.6878)
    writer.add_scalar(tag="acc", step=3, value=0.9878)
```

--------------------------------

### Model Configuration Attributes

Source: https://visualdl.readthedocs.io/zh/latest/components/fastdeploy_server/README.html

Definitions for configurable attributes in the model configuration editor, including maxBatchSize, instanceGroup, and optimization.

```text
maxBatchSize: The maximum batch size allowed in inference. Please enter a non-negative integer, such as 16
instanceGroup: Set hardware resources the server uses and how many instances are deployed. It contains three configurable items, count represents the number of deployed instances, please enter a non-negative integer, such as 2. kind indicates the type of device used, and you can choose KIND_CPU to indicate the use of CPU or KIND_GPU to indicate the use of GPU. When KIND_GPU is selected, gpus indicates that the GPU card numbers need to be used, please enter multiple integers separated by commas (,), such as using cards 2 and 3, i.e. 2,3
optimization: Set the configuration of execution acceleration, you can configure the acceleration configuration of cpu and gpu separately. It contains the name of the backend, such as onnxruntime, and parameters related to the backend. The first text is filled with the parameter name, such as cpu_threads, and the second text is filled with the parameter value, such as 4
```

--------------------------------

### add_graph(model, input_spec, verbose=False)

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records the model structure for visualization in the VisualDL Graph panel.

```APIDOC
## add_graph(model, input_spec, verbose=False)

### Description
Records the dynamic model structure to be visualized in the VisualDL Graph panel.

### Parameters
- **model** (paddle.nn.Layer) - Required - The dynamic model of paddle.
- **input_spec** (list[paddle.static.InputSpec|Tensor]) - Required - Describes the input of the saved model's forward arguments.
- **verbose** (bool) - Optional - Whether to print graph statistic information in the console.
```

--------------------------------

### LogWriter

Source: https://visualdl.readthedocs.io/zh/latest/README.html

The LogWriter class is the primary interface for initializing a logger to record training information. It supports configuration for log directories, caching behavior, and file naming.

```APIDOC
## class LogWriter(logdir=None, comment='', max_queue=10, flush_secs=120, filename_suffix='', write_to_disk=True, **kwargs)

### Description
Initializes a logger to record training process data into a log file.

### Parameters
- **logdir** (string) - Optional - The path location of the log file. Defaults to `runs/${CURRENT_TIME}`.
- **comment** (string) - Optional - Suffix for the log folder name if logdir is not specified.
- **max_queue** (int) - Optional - Maximum capacity of data before writing to disk. Default is 10.
- **flush_secs** (int) - Optional - Maximum cache time in seconds before writing to disk. Default is 120.
- **filename_suffix** (string) - Optional - Suffix for the default log file name.
- **write_to_disk** (boolean) - Optional - Whether to write data to disk.
- **display_name** (string) - Optional - Custom name for the run.

### Example
```python
from visualdl import LogWriter

with LogWriter(logdir="./log/scalar_test/train") as writer:
    writer.add_scalar(tag="acc", step=1, value=0.5678)
```
```

--------------------------------

### add_histogram(tag, values, step, walltime=None, buckets=10)

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records a histogram of tensor data for a specific training step.

```APIDOC
## add_histogram

### Description
Records the distribution of tensor data (weights, biases, gradients) to be visualized as a histogram in the VisualDL panel.

### Signature
`add_histogram(tag, values, step, walltime=None, buckets=10)`

### Parameters
- **tag** (string) - Required - The name of the data series (e.g., 'train/loss'). Cannot contain '%'.
- **values** (numpy.ndarray or list) - Required - The tensor data to record, expected shape is (N, ).
- **step** (int) - Required - The current training step.
- **walltime** (int) - Optional - The timestamp of the data. Defaults to current time.
- **buckets** (int) - Optional - The number of segments to generate the histogram. Defaults to 10.

### Usage Example
```python
from visualdl import LogWriter
import numpy as np

with LogWriter(logdir="./log") as writer:
    data = np.random.uniform(0, 1, size=(10000))
    writer.add_histogram(tag='layer1/weights', values=data, step=1)
```
```

--------------------------------

### Record Scalar Interface

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

The primary interface for logging scalar data points during training.

```python
add_scalar(tag, value, step, walltime=None)
```

--------------------------------

### Compare Multiple Experiments with Scalar

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records data from multiple experiments using the same tags to enable side-by-side comparison in the UI.

```python
from visualdl import LogWriter

if __name__ == '__main__':
    value = [i/1000.0 for i in range(1000)]
    # Step 1: Create a parent folder: log and a child folder: scalar_test
    with LogWriter(logdir="./log/scalar_test") as writer:
        for step in range(1000):
            # Step 2: Add data with tag train/acc to the recorder
            writer.add_scalar(tag="train/acc", step=step, value=value[step])
            # Step 2: Add data with tag train/loss to the recorder
            writer.add_scalar(tag="train/loss", step=step, value=1/(value[step] + 1))
    # Step 1: Create a second child folder: scalar_test2    
    value = [i/500.0 for i in range(1000)]
    with LogWriter(logdir="./log/scalar_test2") as writer:
        for step in range(1000):
            # Step 2: Add the accuracy data of scalar_test2 under the same name `train/acc`
            writer.add_scalar(tag="train/acc", step=step, value=value[step])
            # Step 2: Add the loss data of scalar_test2 under the same name as `train/loss`
            writer.add_scalar(tag="train/loss", step=step, value=1/(value[step] + 1))
```

--------------------------------

### Define add_graph interface

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

The interface signature for adding a model graph to the VisualDL log.

```python
add_graph(model, input_spec, verbose=False):
```

--------------------------------

### add_hparams(hparam_dict, metric_list, walltime=None)

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records hyperparameter data and specifies the names of metrics to be tracked for a specific experiment.

```APIDOC
## add_hparams

### Description
Records the hyperparameter configuration and the list of metrics that will be tracked for the experiment. Metrics specified here must match the tags used in subsequent add_scalar calls.

### Parameters
- **hparam_dict** (dict) - Required - A dictionary containing the names and values of the hyperparameters.
- **metric_list** (list) - Required - A list of metric names (strings) that correspond to the 'tag' parameter in the add_scalar interface.
- **walltime** (int) - Optional - The timestamp for the data record. Defaults to the current timestamp.
```

--------------------------------

### Histogram Record Interface

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

The function signature for adding histogram data to the log.

```python
add_histogram(tag, values, step, walltime=None, buckets=10)
```

--------------------------------

### LogWriter class definition

Source: https://visualdl.readthedocs.io/zh/latest/README.html

The constructor signature for the LogWriter class used to initialize the logging backend.

```python
class LogWriter(logdir=None,
                comment='',
                max_queue=10,
                flush_secs=120,
                filename_suffix='',
                write_to_disk=True,
                **kwargs)
```

--------------------------------

### Define add_hparams interface

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Signature for the add_hparams method used to initialize hyperparameter tracking.

```python
add_hparams(hparam_dict, metric_list, walltime=None):
```

--------------------------------

### add_scalar(tag, value, step, walltime=None)

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records a scalar value to be displayed in a line chart.

```APIDOC
## add_scalar(tag, value, step, walltime=None)

### Description
Records a scalar value for tracking training parameters. The data is visualized as a line chart.

### Parameters
- **tag** (string) - Required - The name of the scalar data (e.g., 'train/loss'). Cannot contain '%'.
- **value** (float) - Required - The data value to record.
- **step** (int) - Required - The training step index.
- **walltime** (int) - Optional - The timestamp of the data. Defaults to the current time.
```

--------------------------------

### add_embeddings(tag, labels, hot_vectors, labels_meta=None, walltime=None)

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

Records high-dimensional data (hot_vectors) associated with specific labels to the log writer for dimensionality reduction analysis.

```APIDOC
## add_embeddings

### Description
Records a set of labels and corresponding hot_vectors to the recorder for high-dimensional data analysis.

### Parameters
- **tag** (string) - Required - The name of the high dimensional data (cannot contain '%').
- **labels** (numpy.array or list) - Required - The label of hot_vectors. Shape should be (N,) or (M, N).
- **hot_vectors** (numpy.array or list) - Required - Features corresponding to the tag and label.
- **labels_meta** (numpy.array or list) - Optional - Metadata corresponding to labels. Defaults to '__metadata__'.
- **walltime** (int) - Optional - Timestamp of the data. Defaults to current time.
```

--------------------------------

### Record High Dimensional Data Interface

Source: https://visualdl.readthedocs.io/zh/latest/components/README.html

The signature for the add_embeddings method used to record high-dimensional data.

```python
add_embeddings(tag, labels, hot_vectors, walltime=None)
```
Tokens:
Raw