Skip to content

Update code-saving notebook to use run.log_code #636

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We鈥檒l occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
johndmulhausen wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Update code-saving notebook to use run.log_code #636

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 34 additions & 17 deletions colabs/wandb-log/Saving_Code_with_W&B.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@
"\n",
"With Weights & Biases, you **won't need to worry** about that happening again!\n",
"We'll **save the code you ran along with results and hyperparameters**, all in a centralized location with easy comparison and visualization tools.\n",
"Even better, in a Jupyter notebbok this feature **tracks all the cells you executed!**\n",
"Even better, in a Jupyter notebook this feature **tracks all the cells you executed!**\n",
"\n",
"Here is a simple implementation where we simulate logging some metrics and the code that generates them -- the same procedure works for both notebooks and Python scripts.\n",
"\n",
Expand Down Expand Up @@ -110,11 +110,11 @@
"with run:\n",
" for step in range(100):\n",
" # insert training process here\n",
" wandb.log({\n",
" run.log({\n",
" \"acc\": math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_acc\": math.log(0.1 + random.random() + step * 0.01),\n",
" \"loss\": wandb.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_loss\": wandb.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01)})"
" \"loss\": run.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_loss\": run.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01)})"
]
},
{
Expand All @@ -137,10 +137,16 @@
"source": [
"# 馃捑 Logging Metrics and Saving Code\n",
"\n",
"Adding code saving is easy: we just pass the argument\n",
"`save_code=True` to `wandb.init`. That's it!\n",
"To save code, we do two things:\n",
"\n",
"_Hot Tip_: If you don't want to worry about setting this on every project,\n",
"1. Pass `save_code=True` to `wandb.init`. This turns on code saving for the run and, in a notebook, tells W&B to try to capture the notebook and the history of cells you executed in the session.\n",
"2. Call `run.log_code()` before the run finishes. This uploads a snapshot of your code -- the files that match the filter you pass (here, `.py` and `.ipynb` files), plus the notebook copy W&B stages in notebook sessions -- as a versioned code artifact. That artifact is what fills in the `{}` code viewer on the run page.\n",
"\n",
"Calling [`run.log_code`](https://docs.wandb.ai/models/ref/python/experiments/run/) is the most reliable way to make sure your code lands in W&B, whether you run a script, a Jupyter notebook, or Colab. In a script, you can also capture code automatically by passing `save_code=True` together with `settings=wandb.Settings(code_dir=\".\")` to `wandb.init`. See the [code saving docs](https://docs.wandb.ai/models/app/features/panels/code) for details.\n",
"\n",
"This time, we'll also let the run span two cells and close it explicitly with `run.finish()`. Whenever a cell executes while a run is active, W&B stages a copy of the notebook, so `run.log_code` can upload it even on platforms like Colab where the notebook file isn't sitting on disk.\n",
"\n",
"_Hot Tip_: If you don't want to worry about setting `save_code` on every project,\n",
"just change your default on the [settings page](https://wandb.ai/settings), as below:"
]
},
Expand All @@ -159,16 +165,27 @@
"source": [
"run = wandb.init(project=\"code_save\",\n",
" config={\"hyperparameter\": 4},\n",
" save_code=True)\n",
" save_code=True)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"for step in range(100):\n",
" # insert training process here\n",
" run.log({\n",
" \"acc\": math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_acc\": math.log(0.1 + random.random() + step * 0.01),\n",
" \"loss\": run.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_loss\": run.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01)})\n",
"\n",
"with run:\n",
" for step in range(100):\n",
" # insert training process here\n",
" wandb.log({\n",
" \"acc\": math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_acc\": math.log(0.1 + random.random() + step * 0.01),\n",
" \"loss\": wandb.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01),\n",
" \"val_loss\": wandb.config.hyperparameter - math.log(0.1 + random.random() + step * 0.01)})"
"# upload a snapshot of our code to the run as a code artifact\n",
"run.log_code(include_fn=lambda path: path.endswith(\".py\") or path.endswith(\".ipynb\"))\n",
"\n",
"run.finish()"
]
},
{
Expand All @@ -186,7 +203,7 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"If you click it, you'll see the entire history of the notebook session! That includes the code we ran for the section without code logging.\n",
"If you click it, you'll see the code that was saved for this run: the files uploaded by `run.log_code` and, in a notebook, the history of the cells executed in the session -- including the ones we ran for the section without code saving.\n",
"\n",
"W&B also automatically catches the standard out and standard error,\n",
"plus system metrics!\n"
Expand Down
Loading