CODEWITHSUNDEEP
pythonwandb
Charlie Parker
Charlie Parker

Reputation: 5201

How does one save a plot in wandb with wandb.log?

I'm trying to save a plot with wandb.log. Their docs say to do:

    wandb.log({"chart": plt})

but this fails for me.

I get two errors, 1st error (when I do NOT do plt.show() before trying to do wand.log):

Traceback (most recent call last):
  File "/Applications/PyCharm.app/Contents/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_exec2.py", line 3, in Exec
    exec(exp, global_vars, local_vars)
  File "<input>", line 1, in <module>
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 256, in wrapper
    return func(self, *args, **kwargs)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 222, in wrapper
    return func(self, *args, **kwargs)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 1548, in log
    self._log(data=data, step=step, commit=commit)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 1339, in _log
    self._partial_history_callback(data, step, commit)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 1228, in _partial_history_callback
    self._backend.interface.publish_partial_history(
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/interface/interface.py", line 541, in publish_partial_history
    data = history_dict_to_json(run, data, step=user_step, ignore_copy_err=True)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/data_types/utils.py", line 54, in history_dict_to_json
    payload[key] = val_to_json(
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/data_types/utils.py", line 82, in val_to_json
    val = Plotly.make_plot_media(val)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/data_types/plotly.py", line 48, in make_plot_media
    val = util.matplotlib_to_plotly(val)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/util.py", line 560, in matplotlib_to_plotly
    return tools.mpl_to_plotly(obj)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/tools.py", line 112, in mpl_to_plotly
    matplotlylib.Exporter(renderer).run(fig)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/exporter.py", line 53, in run
    self.crawl_fig(fig)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/exporter.py", line 124, in crawl_fig
    self.crawl_ax(ax)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/exporter.py", line 146, in crawl_ax
    self.draw_collection(ax, collection)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/exporter.py", line 289, in draw_collection
    offset_order = offset_dict[collection.get_offset_position()]
AttributeError: 'LineCollection' object has no attribute 'get_offset_position'

I get two errors, 2nd error (when I DO plt.show() before trying to do wand.log):

Traceback (most recent call last):
  File "/Applications/PyCharm.app/Contents/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_exec2.py", line 3, in Exec
    exec(exp, global_vars, local_vars)
  File "<input>", line 1, in <module>
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 256, in wrapper
    return func(self, *args, **kwargs)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 222, in wrapper
    return func(self, *args, **kwargs)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 1548, in log
    self._log(data=data, step=step, commit=commit)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 1339, in _log
    self._partial_history_callback(data, step, commit)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/wandb_run.py", line 1228, in _partial_history_callback
    self._backend.interface.publish_partial_history(
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/interface/interface.py", line 541, in publish_partial_history
    data = history_dict_to_json(run, data, step=user_step, ignore_copy_err=True)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/data_types/utils.py", line 54, in history_dict_to_json
    payload[key] = val_to_json(
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/data_types/utils.py", line 82, in val_to_json
    val = Plotly.make_plot_media(val)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/sdk/data_types/plotly.py", line 48, in make_plot_media
    val = util.matplotlib_to_plotly(val)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/wandb/util.py", line 560, in matplotlib_to_plotly
    return tools.mpl_to_plotly(obj)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/tools.py", line 112, in mpl_to_plotly
    matplotlylib.Exporter(renderer).run(fig)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/exporter.py", line 53, in run
    self.crawl_fig(fig)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/exporter.py", line 122, in crawl_fig
    with self.renderer.draw_figure(fig=fig, props=utils.get_figure_properties(fig)):
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/contextlib.py", line 119, in __enter__
    return next(self.gen)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mplexporter/renderers/base.py", line 45, in draw_figure
    self.open_figure(fig=fig, props=props)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/renderer.py", line 90, in open_figure
    self.mpl_x_bounds, self.mpl_y_bounds = mpltools.get_axes_bounds(fig)
  File "/Users/brandomiranda/opt/anaconda3/envs/meta_learning/lib/python3.9/site-packages/plotly/matplotlylib/mpltools.py", line 265, in get_axes_bounds
    x_min, y_min, x_max, y_max = min(x_min), min(y_min), max(x_max), max(y_max)
ValueError: min() arg is an empty sequence

Note that their trivial example DOES work:

import matplotlib.pyplot as plt

plt.plot([1, 2, 3, 4])
plt.ylabel("some interesting numbers")
wandb.log({"chart": plt})

for me.

cross posted: https://community.wandb.ai/t/how-does-one-save-a-plot-in-wandb-with-wandb-log/2373

Upvotes: 4

Views: 7233

Answers (2)

Charlie Parker
Charlie Parker

Reputation: 5201

Use:

    wandb.log({"plot": wandb.Image(fig)})

– Rylan Schaeffer

Full working example:


import numpy as np
import matplotlib.pyplot as plt
import wandb

def generate_data() -> tuple[np.ndarray, np.ndarray]:
    # Initialize constants and data arrays
    C = np.array(
        [[7e9, 2e12], [13e9, 2e12], [34e9, 2e12], [70e9, 2e12]]
    )  # [m, Din] = [4, 2]
    m = C.shape[0]  # 4
    K = 1  # Number of columns for L_target
    
    # Create target loss values and add noise
    L_target = np.array([[1.85], [1.66], [1.55], [1.5]])
    L_target = np.repeat(L_target, K, axis=1)  # [m, K]
    assert m == L_target.shape[0], f"{m=} {L_target.shape[0]=}"
    noise = np.random.normal(loc=0.0, scale=0.01, size=(m, K))
    L_target += noise  # [m, K]
    
    return C, L_target

def main():
    # Initialize Weights & Biases
    wandb.init(project='your_project_name')

    # Generate data
    C, L_target = generate_data()

    # Flatten the arrays for plotting
    compute_values = C[:, 0]  # Taking the first column for compute values
    loss_values = L_target.flatten()  # Flatten the loss values to 1D array

    # Create the plot
    fig, ax = plt.subplots()
    ax.plot(compute_values, loss_values, linestyle='-', marker='x', label='Random Scaling Law')
    ax.set_xscale("log")
    ax.set_yscale("linear")
    ax.set_xlabel("Log Compute (log C), C=6ND")
    ax.set_ylabel("Loss")
    ax.set_title("Scaling Law Prediction (Log-Log Plot)")
    ax.grid(True)
    ax.legend()

    # Log the plot as an image to wandb
    wandb.log({"plot": wandb.Image(fig)})

    # Display the plot
    plt.show()

if __name__ == "__main__":
    main()

Upvotes: 0

APapaya
APapaya

Reputation: 9

Check out the Log Plots portion of this Notebook:

https://colab.research.google.com/github/wandb/examples/blob/master/colabs/wandb-log/Log_(Almost)_Anything_with_W%26B_Media.ipynb#scrollTo=Log_plots

After importing your wandb and logging in / authenticating, try executing the Log Plots cell and then clicking on the link to your visualize-predictions project. (This project gets automatically created when you call wandb.init with the project parameter equal to "visualize-predictions" in Line 5 of the code snippet pasted below.)

The code in the Log_plots section is pasted below for posterity / so that the search engines can find it too:

import matplotlib.pyplot as plt
%matplotlib inline

# Initialize a new run
wandb.init(project="visualize-predictions", name="plots")

# Make the plot
fibonacci = [0, 1, 1, 2, 3, 5, 8, 13, 21, 34]
fig, ax = plt.subplots()
ax.plot(fibonacci)
ax.set_ylabel("some interesting numbers")

# Log the plot
wandb.log({"plot": fig})
wandb.finish()
fig

Upvotes: 0

Related Questions