From 99e454dff52c32dd2699d836d62b24c5b7552253 Mon Sep 17 00:00:00 2001
From: Yotam-Perlitz <y.perlitz@ibm.com>
Date: Tue, 24 Sep 2024 16:58:15 +0200
Subject: [PATCH] add explanations

Signed-off-by: Yotam-Perlitz <y.perlitz@ibm.com>
---
 examples/run_generic_inference_engine.py | 31 ++++++++++++++++++------
 1 file changed, 24 insertions(+), 7 deletions(-)

diff --git a/examples/run_generic_inference_engine.py b/examples/run_generic_inference_engine.py
index 168902a1b..b234e6467 100644
--- a/examples/run_generic_inference_engine.py
+++ b/examples/run_generic_inference_engine.py
@@ -1,11 +1,21 @@
-from unitxt import get_logger, produce
-from unitxt.inference import GenericInferenceEngine
+from unitxt import get_logger, produce  # Import necessary functions from unitxt
+from unitxt.inference import GenericInferenceEngine  # Import the inference engine class
 
 if __name__ == "__main__":
+    # Create an instance of the GenericInferenceEngine with a default engine.
+    # This means if no engine is specified during inference, it will default to this one.
     generic_engine_with_default = GenericInferenceEngine(
         default="engines.ibm_gen_ai.llama_3_70b_instruct"
     )
+
+    # Define the recipe for data processing and model selection.
+    # - card: Specifies the underlying data (from cards.almost_evil).
+    # - template: Selects the specific template within the card (from templates.qa.open.simple).
+    # - demos_pool_size and num_demos: Control the number of demonstration examples used (set to 0 here).
     recipe = "card=cards.almost_evil,template=templates.qa.open.simple,demos_pool_size=0,num_demos=0"
+
+    # Create a list of instances (data points) for inference.
+    # Each instance has a "question" and its corresponding "answers".
     instances = [
         {
             "question": "How many days there are in a week, answer only with numerals",
@@ -16,19 +26,26 @@
             "answers": ["2"],
         },
     ]
+
+    # Process the instances using the defined recipe.
+    # This likely formats the data according to the chosen card and template.
     dataset = produce(instances, recipe)
 
-    # now, trying without a default, make sure you have something like
-    # export UNITXT_INFERENCE_ENGINE="engines.ibm_gen_ai.llama_3_70b_instruct"
-    # in your ~/.bashrc
+    # Perform inference on the processed dataset using the engine with the default model.
     predictions = generic_engine_with_default.infer(dataset)
-    get_logger().info(predictions)
+    get_logger().info(predictions)  # Log the predictions
 
+    # The following code block demonstrates how to use the GenericInferenceEngine without specifying a
+    # default engine. It expects the engine to be defined in the UNITXT_INFERENCE_ENGINE environment variable.
     try:
+        # Attempt to create an instance without a default engine.
         generic_engine_without_default = GenericInferenceEngine()
+
+        # Perform inference (will use the engine specified in the environment variable).
         predictions = generic_engine_without_default.infer(dataset)
-        get_logger().info(predictions)
+        get_logger().info(predictions)  # Log the predictions
     except:
+        # Handle the case where the environment variable is not set.
         get_logger().error(
             "GenericInferenceEngine could not be initialized without a default since "
             "UNITXT_INFERENCE_ENGINE environmental variable is not set."