Python库 | aim-2.0.27rc5.tar.gz

<div align="center"> <img src="https://user-images.githubusercontent.com/13848158/97081166-8f568800-1611-11eb-991c-e9bc1344074e.png" height="95" /> **A super-easy way to record, search and compare AI experiments.** [](https://pypi.org/project/aim-cli/) [](https://pypi.org/project/aim-cli/) [](https://hub.docker.com/r/aimhubio/aim-board) [](https://opensource.org/licenses/Apache-2.0) <a href="https://join.slack.com/t/aimstack/shared_invite/zt-ik4s4zb7-jGg4WIMu4s3NL3YAkOHVXA"> <img src="https://user-images.githubusercontent.com/13848158/97266254-9532b000-1841-11eb-8b06-ed73e99c2e5f.png" height="35" /> </a> --- <h6 style="color: grey"> <a href="http://play.aimstack.io:43900/explore?search=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">PLAY</a> with live demo and check out a short <a href="https://www.youtube.com/watch?v=TeAkyRIMxx4&ab_channel=Aim">INTRO VIDEO</a> </h6> <img src="https://user-images.githubusercontent.com/13848158/98570873-54b94480-22cd-11eb-84d8-c2c13651021d.png" /> --- <h6 style="color: grey">Integrate seamlessly with your favorite tools</h6> <img src="https://user-images.githubusercontent.com/13848158/96861310-f7239c00-1474-11eb-82a4-4fa6eb2c6bb1.jpg" width="100" /> <img src="https://user-images.githubusercontent.com/13848158/97086626-8b3c6180-1635-11eb-9e90-f215b898e298.png" width="100" /> <img src="https://user-images.githubusercontent.com/13848158/96859323-6ba90b80-1472-11eb-9a6e-c60a90f11396.jpg" width="100" /> <img src="https://user-images.githubusercontent.com/13848158/96861315-f854c900-1474-11eb-8e9d-c7a07cda8445.jpg" width="100" /> </div> ## Getting started in three steps > **1. Install Aim in your training environment** ```shell $ pip install aim ``` > **2. Integrate Aim with your code** <details open> <summary> Flexible integration for any Python script </summary> ```python import aim # Save inputs, hparams or any other `key: value` pairs aim.set_params(hyperparam_dict, name='hparams') # Passing name argument is optional ... for step in range(10): # Log metrics to visualize performance aim.track(metric_value, name='metric_name', epoch=epoch_number) ... ``` _See documentation [here](#python-library)._ </details> <details> <summary> PyTorch Lightning integration </summary> ```python from aim.pytorch_lightning import AimLogger ... trainer = pl.Trainer(logger=AimLogger(experiment='experiment_name')) ... ``` _See documentation [here](#pytorch-lightning)._ </details> <details> <summary> Keras & tf.keras integrations </summary> ```python import aim # Save inputs, hparams or any other `key: value` pairs aim.set_params(param_dict, name='params_name') # Passing name argument is optional ... model.fit(x_train, y_train, epochs=epochs, callbacks=[ aim.keras.AimCallback(aim.Session(experiment='experiment_name')) # Use aim.tensorflow.AimCallback in case of tf.keras aim.tensorflow.AimCallback(aim.Session(experiment='experiment_name')) ]) ... ``` _See documentation [here](#tensorflow-and-keras)._ </details> > **3. Run the training like you are used to and start Aim UI** ```shell $ aim up ``` ## Contents - [Aim](#aim) - [Contents](#contents) - [Getting Started In Three Steps](#getting-started-in-three-steps) - [Installation](#installation) - [Concepts](#concepts) - [Where is the Data Stored](#where-is-the-data-stored) - [Python Library](#python-library) - [aim.track()](#track) - [aim.set_params()](#set_params) - [aim.Session()](#session) - [Automatic Tracking](#automatic-tracking) - [TensorFlow and Keras](#tensorflow-and-keras) - [PyTorch Lightning](#pytorch-lightning) - [Searching Experiments](#searching-experiments) - [Search Examples](#search-examples) - [Command Line Interface](#command-line-interface) - [init](#init) - [version](#version) - [experiment](#experiment) - [up](#up) - [down](#down) - [upgrade](#upgrade) - [pull](#pull) - [TensorBoard Experiments](#tensorboard-experiments) - [Contributor Guide](https://github.com/aimhubio/aim/wiki/Contributing) ## Installation To install Aim, you need to have python3 and pip3 installed in your environment 1. Install Aim python package ```shell $ pip install aim ``` In order to start Aim UI you need to have Docker installed. ```shell $ aim up ``` ## Concepts - **Run** - A single training run - **Experiment** - a group of associated training runs ## Where is the Data Stored When the AI training code is instrumented with [Aim Python Library](#python-library) and ran, aim automatically creates a `.aim` directory where the project is located. All the metadata tracked during training via the Python Library is stored in `.aim`. Also see [`aim init`](#init) - an optional and alternative way to initialize aim repository. ## Python Library Use Python Library to instrument your training code to record the experiments. The instrumentation only takes 2 lines: ```py import aim ``` Afterwards, simply use the two following functions to track metrics and any params respectively. ```py ... aim.track(metric_val, name='metric_name', epoch=current_epoch) aim.set_params(hyperparam_dict, name='dict_name') ... ``` ### track aim.**track**_(value, name='metric_name' [, epoch=epoch] [, **context_args]) _{[source](https://github.com/aimhubio/aim/blob/6ef09d8d77c517728978703764fc9ffe323f12b0/aim/sdk/track.py#L6)}_ _Parameters_ - **value** - the metric value of type `int`/`float` to track/log - **name** - the name of the metric of type `str` to track/log (preferred divider: `snake_case`) - **epoch** - an optional value of the epoch being tracked - **context_args** - any set of other parameters passed would be considered as key-value context for metrics _Examples_ ```py aim.track(0.01, name='loss', epoch=43, subset='train', dataset='train_1') aim.track(0.003, name='loss', epoch=43, subset='val', dataset='val_1') ``` Once tracked this way, the following search expressions will be enabled: ```py loss if context.subset in (train, val) # Retrieve all losses in both train and val phase loss if context.subse