Update federated evaluation documentation (#1624)

* Update federated evaluation documentation and adjust input shape for Keras data loader. Removed input shape specification from the plan.yaml file and updated the feature shape in the KerasHippmapp3rsynth class to [160, 160, 128]. Additionally, modified the README to reflect changes in the plan initialization command by removing the input shape argument.

Signed-off-by: Rahul Garg <rahul.garg@intel.com>

* Update taskrunner tutorial to clarify the command's support for model initializations. Adjusted wording to enhance understanding of the FL experiment plan setup process.

Signed-off-by: Rahul Garg <rahul.garg@intel.com>

---------

Signed-off-by: Rahul Garg <rahul.garg@intel.com>

Author: rahulga1

docs/about/features_index/fed_eval.rst

@@ -108,13 +108,6 @@ The plan initialize command supports the optional `--init_model_path` (shortform
     -c, --cols_config PATH         Path to an authorized collaborator list.  [default: plan/cols.yaml]
     -d, --data_config PATH         The dataset shard configuration file.  [default: plan/data.yaml]
     -a, --aggregator_address TEXT  The FQDN of the federation aggregator
-    -f, --input_shape TEXT         The input spec of the model.
-
-                                    May be provided as a list for single input head: ``--input-shape [3,32,32]``,
-
-                                    or as a dictionary for multihead models (must be passed in quotes):
-
-                                    ``--input-shape "{'input_0': [1, 240, 240, 4],'input_1': [1, 240, 240, 1]}"``.
     -g, --gandlf_config TEXT       GaNDLF Configuration File Path
     -r, --install_reqs BOOLEAN     If set, installs packages listed under 'requirements.txt'.  [default: True]
     -i, --init_model_path PATH     Path to initial model protobuf file.

docs/tutorials/taskrunner.ipynb

@@ -69,7 +69,7 @@
         "### Initialize a Plan\n",
         "\n",
         "This step builds an entire FL experiment plan, along with the initial set of parameters that will be used in the experiment.\n",
-        "We supply `localhost` as the aggregator address, for simulation purposes. The command below supports models of arbitrary input shapes and initializations (e.g. when using pre-trained models).\n",
+        "We supply `localhost` as the aggregator address, for simulation purposes. The command below supports models initializations (e.g. when using pre-trained models).\n",
         "\n",
         "For a complete list of supported options, refer to [`fx plan initialize`](https://openfl.readthedocs.io/en/latest/reference/fx.plan.html#fx-plan-initialize)."
       ]

openfl-workspace/keras/hippmapp3r/plan/plan.yaml

@@ -22,7 +22,6 @@ data_loader :
   template : src.dataloader.KerasHippmapp3rsynth
   settings :
     batch_size : 1
-    input_shape : [160, 160, 128]
 
 task_runner :
   defaults : plan/defaults/task_runner.yaml

openfl-workspace/keras/hippmapp3r/src/dataloader.py

@@ -24,7 +24,7 @@ def __init__(self, data_path=None, batch_size=32, **kwargs):
             **kwargs: Additional arguments, passed to super init and load_mnist_shard
         """
         super().__init__(batch_size, **kwargs)
-        self.feature_shape = [64, 64, 128]  # Hippmapp3r shape for Keras (channels last)
+        self.feature_shape = [160, 160, 128]  # Hippmapp3r shape for Keras (channels last)
 
         # If data_path is None, this is being used for model initialization only
         if data_path is None:

openfl-workspace/torch/mnist/README.md

@@ -168,7 +168,7 @@ To help OpenFL calculate the initial model weights, we need to provide the shape
 
 ```
 mkdir save
-fx plan initialize --input_shape [1,28,28] --aggregator_address localhost
+fx plan initialize --aggregator_address localhost
 ```
 
 We can now perform a test run with the following commands for creating a local PKI setup and starting the aggregator and the collaborators on the same machine: