Added sub heading to ignite.metrics

docs/source/metrics.rst

@@ -4,7 +4,16 @@ ignite.metrics
 Metrics provide a way to compute various quantities of interest in an online
 fashion without having to store the entire output history of a model.
 
-In practice a user needs to attach the metric instance to an engine. The metric
+Attach Engine API
+------------------
+
+The metrics as stated above are computed in a online fashion,which means that the metric instance accumulates some internal counter on 
+each iteration and accuracy value is computed once epoch is ended. Internal counters are reset after every epoch.
+For doing this we use three public APIs ``reset`` , ``update`` and ``compute`` :ref:`read about them here<Reset, Update, Compute API>`. We internally accumulate some counters
+when there is a update call. The metric's value is computed on each compute call and counters are reset on each reset
+call. 
+
+In practice a user needs to attach the metric instance to an engine which attaches the ``reset`` , ``compute`` and ``update`` methods to the events. The metric
 value is then computed using the output of the engine's ``process_function``:
 
 .. code-block:: python
@@ -24,6 +33,7 @@ value is then computed using the output of the engine's ``process_function``:
     print(f"Accuracy: {state.metrics['accuracy']}")
 
 
+
 If the engine's output is not in the format ``(y_pred, y)`` or ``{'y_pred': y_pred, 'y': y, ...}``, the user can
 use the ``output_transform`` argument to transform it: