update readme, add error messages, time logging

README.md

@@ -8,8 +8,21 @@ sdk_version: 4.7.1
 app_file: app.py
 pinned: false
 license: mit
-models: ["onlycaps/pace_model_weights", "Salesforce/blip-image-captioning-large", "facebook/musicgen-small"]
-tags: ["image2music-generation", "image-captioning"]
+models:
+- onlycaps/pace_model_weights
+- Salesforce/blip-image-captioning-large
+- facebook/musicgen-small
+tags:
+- "image2music-generation"
+- "image-captioning"
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Audio Palette
+
+### Usage
+
+Since this space is running on CPU, it is not possible to generate music in a reasonable time.
+
+To address this, we have provided a [python notebook](./notebooks/AudioPalette.ipynb) that handles the music generation part which can be run locally (if you have GPU) or elsewhere.
+
+This uses fastAPI to accept api requests and ngrok to expose the server. The same ngrok link needs to be pasted in the input box. (Make sure to include the trailing `/`).

lib/sentiment_analyser.py

@@ -1,7 +1,7 @@
 import os
 import string
 from collections import Counter
-from datetime import datetime
+from datetime import datetime, timezone, timedelta
 from pathlib import Path
 
 import nltk
@@ -13,9 +13,11 @@ from nltk.tokenize import word_tokenize
 from utils import *
 
 datetime_format = "%d/%m/%Y %H:%M:%S"
-
+ist_offset = timedelta(hours=5, minutes=30)
 def now():
-    return datetime.now().strftime(datetime_format)
+    utc_time = datetime.now(timezone.utc)
+    ist_time = utc_time.astimezone(timezone(ist_offset))
+    return ist_time.strftime(datetime_format)
 
 class SentimentAnalyser:
     def __init__(self):

utils/audio_palette.py

@@ -1,15 +1,19 @@
 import typing
-from datetime import datetime
+from datetime import datetime, timezone, timedelta
 
 import PIL
 from PIL import Image
 from moviepy.editor import *
+from gradio import Error
 
 from lib import *
 
 datetime_format = "%d/%m/%Y %H:%M:%S"
+ist_offset = timedelta(hours=5, minutes=30)
 def now():
-    return datetime.now().strftime(datetime_format)
+    utc_time = datetime.now(timezone.utc)
+    ist_time = utc_time.astimezone(timezone(ist_offset))
+    return ist_time.strftime(datetime_format)
 
 class AudioPalette:
     def __init__(self, pace_model_weights_path, resnet50_tf_model_weights_path, height, width, channels):
@@ -36,12 +40,22 @@ class AudioPalette:
         
         return prompt
     
-    def generate_single(self, input_image: PIL.Image.Image, instrument: typing.Union[str, None], ngrok_endpoint: str):
+    def generate_single(self, input_image: PIL.Image.Image, instrument: typing.Union[str, None], ngrok_endpoint: typing.Union[str, None]):
+        if not ngrok_endpoint:
+            print(f"[{now()}] ngrok endpoint missing")
+            raise Error("ngrok endpoint missing")
+        print(f"[{now()}] {ngrok_endpoint}")
+
         pace = self.pace_model.predict(input_image)
         print(f"[{now()}]", pace)
         print(f"[{now()}] Pace Prediction Done")
 
-        generated_text = self.image_captioning.query(input_image)[0].get("generated_text")
+        try:
+            generated_text = self.image_captioning.query(input_image)[0].get("generated_text")
+        except Exception as e:
+            print(f"[{now()}] image captioning error")
+            raise Error(repr(e))
+
         print(f"[{now()}]", generated_text)
         print(f"[{now()}] Captioning Done")
 
@@ -71,7 +85,12 @@ class AudioPalette:
         concat_clip.write_videofile(file_name, fps=24)
         return file_name
     
-    def generate_multiple(self, file_paths: typing.List[str], instrument: typing.Union[str, None], ngrok_endpoint: str):
+    def generate_multiple(self, file_paths: typing.List[str], instrument: typing.Union[str, None], ngrok_endpoint: typing.Union[str, None]):
+        if not ngrok_endpoint:
+            print(f"[{now()}] ngrok endpoint missing")
+            raise Error("ngrok endpoint missing")
+        print(f"[{now()}] {ngrok_endpoint}")
+
         images = [Image.open(image_path) for image_path in file_paths]
         pace = []
         generated_text = []
@@ -86,9 +105,14 @@ class AudioPalette:
         print(f"[{now()}] Pace Prediction Done")
         
         # Generating the caption for all the images
-        for image in images:
-            caption = self.image_captioning.query(image)[0].get("generated_text")
-            generated_text.append(caption)
+        try:
+            for image in images:
+                caption = self.image_captioning.query(image)[0].get("generated_text")
+                generated_text.append(caption)
+        except Exception as e:
+            print(f"[{now()}] image captioning error")
+            raise Error(repr(e))
+
         print(f"[{now()}]", generated_text)
         print(f"[{now()}] Captioning Done")
 

utils/gradio_helper.py

@@ -21,7 +21,7 @@ def single_image_interface(model: AudioPalette):
             gr.Textbox(
                 lines=1,
                 placeholder="ngrok endpoint",
-                label="colab endpoint",
+                label="ngrok endpoint",
                 show_label=True,
                 container=True,
                 type="text",
@@ -95,7 +95,7 @@ def multi_image_interface(model: AudioPalette):
             gr.Textbox(
                 lines=1,
                 placeholder="ngrok endpoint",
-                label="colab endpoint",
+                label="ngrok endpoint",
                 show_label=True,
                 container=True,
                 type="text",