Merge pull request #210655 from MadhuM02/limit_setting_changes

limit_settings_changes

Author: prmerger-automator[bot]

articles/machine-learning/how-to-auto-train-image-models.md

@@ -330,16 +330,25 @@ The primary metric used for model optimization and hyperparameter tuning depends
 * `mean_average_precision` for IMAGE_OBJECT_DETECTION
 * `mean_average_precision` for IMAGE_INSTANCE_SEGMENTATION
 
-### Experiment budget
+### Job Limits
+
+You can control the resources spent on your AutoML Image training job by specifying the `timeout_minutes`, `max_trials` and the `max_concurrent_trials` for the job in limit settings as described in the below example.
+
+Parameter | Detail
+-----|----
+`max_trials` |  Parameter for maximum number of configurations to sweep. Must be an integer between 1 and 1000. When exploring just the default hyperparameters for a given model algorithm, set this parameter to 1. default value is 1.
+`max_concurrent_trials`| Maximum number of runs that can run concurrently. If not specified, all runs launch in parallel. If specified, must be an integer between 1 and 100.  <br><br> **NOTE:** The number of concurrent runs is gated on the resources available in the specified compute target. Ensure that the compute target has the available resources for the desired concurrency. default value is 1.
+`timeout_minutes`| The amount of time in minutes before the experiment terminates. If none specified, default experiment timeout_minutes is seven days (maximum 60 days)
 
-You can optionally specify the maximum time budget for your AutoML Vision training job using the `timeout` parameter in the `limits` - the amount of time in minutes before the experiment terminates. If none specified, default experiment timeout is seven days (maximum 60 days). For example,
 # [Azure CLI](#tab/cli)
 
 [!INCLUDE [cli v2](../../includes/machine-learning-cli-v2.md)]
 
 ```yaml
 limits:
-  timeout: 60
+  timeout_minutes: 60
+  max_trials: 10
+  max_concurrent_trials: 2
 ```
 
 # [Python SDK](#tab/python)
@@ -429,16 +438,9 @@ You can automatically end poorly performing runs with an early termination polic
 
 Learn more about [how to configure the early termination policy for your hyperparameter sweep](how-to-tune-hyperparameters.md#early-termination).
 
-### Resources for the sweep
-
-You can control the resources spent on your hyperparameter sweep by specifying the `max_trials` and the `max_concurrent_trials` for the sweep.
 > [!NOTE]
 > For a complete sweep configuration sample, please refer to this [tutorial](tutorial-auto-train-image-models.md#hyperparameter-sweeping-for-image-tasks).
 
-Parameter | Detail
------|----
-`max_trials` |  Required parameter for maximum number of configurations to sweep. Must be an integer between 1 and 1000. When exploring just the default hyperparameters for a given model algorithm, set this parameter to 1.
-`max_concurrent_trials`| Maximum number of runs that can run concurrently. If not specified, all runs launch in parallel. If specified, must be an integer between 1 and 100.  <br><br> **NOTE:** The number of concurrent runs is gated on the resources available in the specified compute target. Ensure that the compute target has the available resources for the desired concurrency.
 
 You can configure all the sweep related parameters as shown in the example below.
 
@@ -448,9 +450,6 @@ You can configure all the sweep related parameters as shown in the example below
 
 ```yaml
 sweep:
-  limits:
-    max_trials: 10
-    max_concurrent_trials: 2
   sampling_algorithm: random
   early_termination:
     type: bandit

articles/machine-learning/tutorial-auto-train-image-models.md

@@ -322,13 +322,33 @@ In your AutoML job, you can specify the model algorithms by using `model_name` p
 
 In this example, we will train an object detection model with `yolov5` and `fasterrcnn_resnet50_fpn`, both of which are pretrained on COCO, a large-scale object detection, segmentation, and captioning dataset that contains over thousands of labeled images with over 80 label categories.
 
+### Job Limits
+
+You can control the resources spent on your AutoML Image training job by specifying the `timeout_minutes`, `max_trials` and the `max_concurrent_trials` for the job in limit settings. PLease refer to [detailed description on Job Limits parameters](./how-to-auto-train-image-models.md#job-limits).
+# [Azure CLI](#tab/cli)
+
+[!INCLUDE [cli v2](../../includes/machine-learning-cli-v2.md)]
+
+```yaml
+limits:
+  timeout_minutes: 60
+  max_trials: 10
+  max_concurrent_trials: 2
+```
+
+# [Python SDK](#tab/python)
+
+[!Notebook-python[] (~/azureml-examples-main/sdk/jobs/automl-standalone-jobs/automl-image-object-detection-task-fridge-items/automl-image-object-detection-task-fridge-items.ipynb?name=limit-settings)]
+---
+
+
 ### Hyperparameter sweeping for image tasks
 
 You can perform a hyperparameter sweep over a defined search space to find the optimal model.
 
 The following code, defines the search space in preparation for the hyperparameter sweep for each defined algorithm, `yolov5` and `fasterrcnn_resnet50_fpn`.  In the search space, specify the range of values for `learning_rate`, `optimizer`, `lr_scheduler`, etc., for AutoML to choose from as it attempts to generate a model with the optimal primary metric. If hyperparameter values are not specified, then default values are used for each algorithm.
 
-For the tuning settings, use random sampling to pick samples from this parameter space by using the `random` sampling_algorithm. Doing so, tells automated ML to try a total of 10 trials with these different samples, running two trials at a time on our compute target, which was set up using four nodes. The more parameters the search space has, the more trials you need to find optimal models.
+For the tuning settings, use random sampling to pick samples from this parameter space by using the `random` sampling_algorithm. The job limits configured above, tells automated ML to try a total of 10 trials with these different samples, running two trials at a time on our compute target, which was set up using four nodes. The more parameters the search space has, the more trials you need to find optimal models.
 
 The Bandit early termination policy is also used. This policy terminates poor performing configurations; that is, those configurations that are not within 20% slack of the best performing configuration, which significantly saves compute resources.
 
@@ -338,9 +358,6 @@ The Bandit early termination policy is also used. This policy terminates poor pe
 
 ```yaml
 sweep:
-  limits:
-    max_trials: 10
-    max_concurrent_trials: 2
   sampling_algorithm: random
   early_termination:
     type: bandit