From 73ced76fad70dd19914119998856afe9b3c51ddb Mon Sep 17 00:00:00 2001
From: ghylander <74593034+ghylander@users.noreply.github.com>
Date: Tue, 5 Jan 2021 14:03:53 +0100
Subject: [PATCH 1/3] update documentation

Added example about using the built in keras callback and how to declare a new callback to track specific metrics
---
 README.md | 49 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/README.md b/README.md
index c452889..56ad88b 100644
--- a/README.md
+++ b/README.md
@@ -119,6 +119,55 @@ def train_dogs_vs_cats(exp): # Get Experiment object as argument to function.
   model.fit()
   exp.metric(model.accuracy())
 ```
+
+
+### Keras Callbacks
+
+Hyperdash has an included keras callback module:
+```python
+from hyperdash import Experiment
+exp = Experiment ("Keras callback showcase")
+
+callbacks=[exp.callbacks.keras]
+
+history = training_model.fit(train_dataset, 
+                                 epochs=epochs, 
+                                 callbacks=callbacks,
+                                 validation_data=validation_dataset,
+                                 validation_steps=steps
+                                )
+```
+This callback will monitor validation loss (val_loss) and the validation accuracy (val_acc).
+
+If you want to monitor other metrics, you will have to define a callback yourself:
+```python
+from hyperdash import Experiment
+from tensorflow.keras.callbacks import Callback
+
+class Hyperdash(Callback):
+    def __init__(self, entries, exp):
+        super(Hyperdash, self).__init__()
+        self.entries = entries
+        self.exp = exp
+    def on_epoch_end(self, epoch, logs=None):
+        for entrie in self.entries:
+            log = logs.get(entrie)
+            if log is not None:
+                self.exp.metric(entrie, log)
+
+callback_hd = Hyperdash(['accuracy', 'loss', 'val_accuracy', 'val_loss'], exp) # Here you list the metrics you want to track
+
+callbacks = [callback_hd]
+
+history = training_model.fit(train_dataset, 
+                             epochs=epochs, 
+                             callbacks=callbacks,
+                             validation_data=validation_dataset,
+                             validation_steps=steps
+                            )
+```
+
+
 ## API Keys
 ### Storage
 

From b40e498ea7d25ec5c00b9636cecdafc09b3b8068 Mon Sep 17 00:00:00 2001
From: ghylander <74593034+ghylander@users.noreply.github.com>
Date: Tue, 5 Jan 2021 14:17:39 +0100
Subject: [PATCH 2/3] update experiment.py

needs confirmation, seems like the accuracy metric 'accuracy' changed names from 'acc' to 'accuracy'. Hence, when using the built in callback, the val_accuracy will not be tracked
---
 hyperdash/experiment.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/hyperdash/experiment.py b/hyperdash/experiment.py
index 79f022e..2ea6dc8 100644
--- a/hyperdash/experiment.py
+++ b/hyperdash/experiment.py
@@ -217,11 +217,11 @@ def __init__(self, exp):
                     def on_epoch_end(self, epoch, logs=None):
                         if not logs:
                             logs = {}
-                        val_acc = logs.get("val_acc")
+                        val_acc = logs.get("val_accuracy")
                         val_loss = logs.get("val_loss")
 
                         if val_acc is not None:
-                            self._exp.metric("val_acc", val_acc)
+                            self._exp.metric("val_accuracy", val_acc)
                         if val_loss is not None:
                             self._exp.metric("val_loss", val_loss)
                 cb = _KerasCallback(self._exp)

From b74bd243318a7ea7e8b6315f362934e0134b2b34 Mon Sep 17 00:00:00 2001
From: Gustavo Hylander <74593034+ghylander@users.noreply.github.com>
Date: Tue, 5 Oct 2021 10:49:21 +0200
Subject: [PATCH 3/3] Update README.md

---
 README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 56ad88b..32af372 100644
--- a/README.md
+++ b/README.md
@@ -12,7 +12,7 @@ Use Hyperdash if you're looking for cloud-based model monitoring that:
 * Saves your experiment's print output (standard out / error) as a local log file.
 * Notifies you when a long-running experiment has finished.
 
-Hyperdash is compatible with: **Python 2.7-3.6**
+Hyperdash is compatible with: **Python 2.7-3.8**
 
 ## Installation
 *Foreword: We care deeply about making Hyperdash fast and easy to install  on Linux, Mac, and Windows. If you find a snag along the way, please let us know at support@hyperdash.io!*