wehub-resource-sync
275 lines
8.4 KiB
Plaintext
275 lines
8.4 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"#| hide\n",
|
|
"#| eval: false\n",
|
|
"! [ -e /content ] && pip install -Uqq fastai # upgrade fastai on colab"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "raw",
|
|
"metadata": {},
|
|
"source": [
|
|
"---\n",
|
|
"skip_exec: true\n",
|
|
"---"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"#| export\n",
|
|
"import tempfile\n",
|
|
"from fastai.basics import *\n",
|
|
"from fastai.learner import Callback"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"#| hide\n",
|
|
"from nbdev.showdoc import *"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"#| default_exp callback.neptune"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"# Neptune.ai\n",
|
|
"\n",
|
|
"> Integration with [neptune.ai](https://www.neptune.ai).\n",
|
|
"\n",
|
|
"> [Track fastai experiments](https://ui.neptune.ai/o/neptune-ai/org/fastai-integration) like in this example project."
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## Registration"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"1. Create account: [neptune.ai/register](https://neptune.ai/register).\n",
|
|
"2. Export API token to the environment variable (more help [here](https://docs.neptune.ai/python-api/tutorials/get-started.html#copy-api-token)). In your terminal run:\n",
|
|
"\n",
|
|
"```\n",
|
|
"export NEPTUNE_API_TOKEN='YOUR_LONG_API_TOKEN'\n",
|
|
"```\n",
|
|
"\n",
|
|
"or append the command above to your `~/.bashrc` or `~/.bash_profile` files (**recommended**). More help is [here](https://docs.neptune.ai/python-api/tutorials/get-started.html#copy-api-token)."
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## Installation"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"1. You need to install neptune-client. In your terminal run:\n",
|
|
"\n",
|
|
"```\n",
|
|
"pip install neptune-client\n",
|
|
"```\n",
|
|
"\n",
|
|
"or (alternative installation using conda). In your terminal run:\n",
|
|
"\n",
|
|
"```\n",
|
|
"conda install neptune-client -c conda-forge\n",
|
|
"```\n",
|
|
"2. Install [psutil](https://psutil.readthedocs.io/en/latest/) to see hardware monitoring charts:\n",
|
|
"\n",
|
|
"```\n",
|
|
"pip install psutil\n",
|
|
"```"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## How to use?"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Key is to call `neptune.init()` before you create `Learner()` and call `neptune_create_experiment()`, before you fit the model.\n",
|
|
"\n",
|
|
"Use `NeptuneCallback` in your `Learner`, like this:\n",
|
|
"\n",
|
|
"```\n",
|
|
"from fastai.callback.neptune import NeptuneCallback\n",
|
|
"\n",
|
|
"neptune.init('USERNAME/PROJECT_NAME') # specify project\n",
|
|
"\n",
|
|
"learn = Learner(dls, model,\n",
|
|
" cbs=NeptuneCallback()\n",
|
|
" )\n",
|
|
"\n",
|
|
"neptune.create_experiment() # start experiment\n",
|
|
"learn.fit_one_cycle(1)\n",
|
|
"```"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"#| export\n",
|
|
"import neptune"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"#| export\n",
|
|
"class NeptuneCallback(Callback):\n",
|
|
" \"Log losses, metrics, model weights, model architecture summary to neptune\"\n",
|
|
" order = Recorder.order+1\n",
|
|
" def __init__(self, log_model_weights=True, keep_experiment_running=False):\n",
|
|
" self.log_model_weights = log_model_weights\n",
|
|
" self.keep_experiment_running = keep_experiment_running\n",
|
|
" self.experiment = None\n",
|
|
"\n",
|
|
" if neptune.project is None:\n",
|
|
" raise ValueError('You did not initialize project in neptune.\\n',\n",
|
|
" 'Please invoke `neptune.init(\"USERNAME/PROJECT_NAME\")` before this callback.')\n",
|
|
"\n",
|
|
" def before_fit(self):\n",
|
|
" try:\n",
|
|
" self.experiment = neptune.get_experiment()\n",
|
|
" except ValueError:\n",
|
|
" print('No active experiment. Please invoke `neptune.create_experiment()` before this callback.')\n",
|
|
"\n",
|
|
" try:\n",
|
|
" self.experiment.set_property('n_epoch', str(self.learn.n_epoch))\n",
|
|
" self.experiment.set_property('model_class', str(type(self.learn.model)))\n",
|
|
" except: print(f'Did not log all properties. Check properties in the {neptune.get_experiment()}.')\n",
|
|
"\n",
|
|
" try:\n",
|
|
" with tempfile.NamedTemporaryFile(mode='w') as f:\n",
|
|
" with open(f.name, 'w') as g: g.write(repr(self.learn.model))\n",
|
|
" self.experiment.log_artifact(f.name, 'model_summary.txt')\n",
|
|
" except: print('Did not log model summary. Check if your model is PyTorch model.')\n",
|
|
"\n",
|
|
" if self.log_model_weights and not hasattr(self.learn, 'save_model'):\n",
|
|
" print('Unable to log model to Neptune.\\n',\n",
|
|
" 'Use \"SaveModelCallback\" to save model checkpoints that will be logged to Neptune.')\n",
|
|
"\n",
|
|
" def after_batch(self):\n",
|
|
" # log loss and opt.hypers\n",
|
|
" if self.learn.training:\n",
|
|
" self.experiment.log_metric('batch__smooth_loss', self.learn.smooth_loss)\n",
|
|
" self.experiment.log_metric('batch__loss', self.learn.loss)\n",
|
|
" self.experiment.log_metric('batch__train_iter', self.learn.train_iter)\n",
|
|
" for i, h in enumerate(self.learn.opt.hypers):\n",
|
|
" for k, v in h.items(): self.experiment.log_metric(f'batch__opt.hypers.{k}', v)\n",
|
|
"\n",
|
|
" def after_epoch(self):\n",
|
|
" # log metrics\n",
|
|
" for n, v in zip(self.learn.recorder.metric_names, self.learn.recorder.log):\n",
|
|
" if n not in ['epoch', 'time']: self.experiment.log_metric(f'epoch__{n}', v)\n",
|
|
" if n == 'time': self.experiment.log_text(f'epoch__{n}', str(v))\n",
|
|
"\n",
|
|
" # log model weights\n",
|
|
" if self.log_model_weights and hasattr(self.learn, 'save_model'):\n",
|
|
" if self.learn.save_model.every_epoch:\n",
|
|
" _file = join_path_file(f'{self.learn.save_model.fname}_{self.learn.save_model.epoch}',\n",
|
|
" self.learn.path / self.learn.model_dir, ext='.pth')\n",
|
|
" else:\n",
|
|
" _file = join_path_file(self.learn.save_model.fname,\n",
|
|
" self.learn.path / self.learn.model_dir, ext='.pth')\n",
|
|
" self.experiment.log_artifact(_file)\n",
|
|
"\n",
|
|
" def after_fit(self):\n",
|
|
" if not self.keep_experiment_running:\n",
|
|
" try: self.experiment.stop()\n",
|
|
" except: print('No neptune experiment to stop.')\n",
|
|
" else:\n",
|
|
" print(f'Your experiment (id: {self.experiment.id}, name: {self.experiment.name}) is left in the running state.\\n',\n",
|
|
" 'You can log more data to it, like this: `neptune.log_metric()`')"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"data": {
|
|
"text/markdown": [
|
|
"---\n",
|
|
"\n",
|
|
"### NeptuneCallback\n",
|
|
"\n",
|
|
"> NeptuneCallback (log_model_weights=True, keep_experiment_running=False)\n",
|
|
"\n",
|
|
"Log losses, metrics, model weights, model architecture summary to neptune"
|
|
],
|
|
"text/plain": [
|
|
"<nbdev.showdoc.BasicMarkdownRenderer>"
|
|
]
|
|
},
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"output_type": "execute_result"
|
|
}
|
|
],
|
|
"source": [
|
|
"show_doc(NeptuneCallback)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": []
|
|
}
|
|
],
|
|
"metadata": {
|
|
"kernelspec": {
|
|
"display_name": "python3",
|
|
"language": "python",
|
|
"name": "python3"
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 4
|
|
}
|