✨ feat: add spotify to find audio features

.gitignore

@@ -0,0 +1,11 @@
+# Environment
+.env
+
+# Flags
+flagged/
+
+# Cache
+.cache/
+.cache
+__pycache__/
+gradio_cached_examples/

app.py

@@ -1,27 +1,33 @@
+import os
 import torch
+import spotipy
 import numpy as np
 import gradio as gr
 import torch.nn as nn
 import torch.nn.functional as F
 
 from transformers import AutoTokenizer, AutoModel
+from spotipy.oauth2 import SpotifyClientCredentials
 
-REPO_NAME = "PunGrumpy/music-genre-classification"
 
-GENRE = {"edm": 0, "r&b": 1, "rap": 2, "rock": 3, "pop": 4}
-AUDIO_FEATURES = {
-    "acousticness": 0,
-    "danceability": 0,
-    "energy": 0,
-    "instrumentalness": 0,
-    "key": 0,
-    "liveness": 0,
-    "loudness": 0,
-    "mode": 0,
-    "speechiness": 0,
-    "tempo": 0,
-    "valence": 0,
-}
+class ConfigApp:
+    REPO_NAME = "PunGrumpy/music-genre-classification"
+    GENRE = {"edm": 0, "r&b": 1, "rap": 2, "rock": 3, "pop": 4}
+    AUDIO_FEATURES = {
+        "acousticness": 0,
+        "danceability": 0,
+        "energy": 0,
+        "instrumentalness": 0,
+        "key": 0,
+        "liveness": 0,
+        "loudness": 0,
+        "mode": 0,
+        "speechiness": 0,
+        "tempo": 0,
+        "valence": 0,
+    }
+    SPOTIFY_CLIENT_ID = os.getenv("SPOTIFY_CLIENT_ID")
+    SPOTIFY_ACCESS_TOKEN = os.getenv("SPOTIFY_ACCESS_TOKEN")
 
 
 class LyricsAudioModelInference:
@@ -30,13 +36,38 @@ class LyricsAudioModelInference:
         self.tokenizer = AutoTokenizer.from_pretrained(model_name)
         self.num_labels = num_labels
         self.classifier = nn.Linear(
-            self.model.config.hidden_size + len(AUDIO_FEATURES), num_labels
+            self.model.config.hidden_size + len(ConfigApp.AUDIO_FEATURES), num_labels
+        )
+        self.sp = spotipy.Spotify(
+            client_credentials_manager=SpotifyClientCredentials(
+                client_id=ConfigApp.SPOTIFY_CLIENT_ID,
+                client_secret=ConfigApp.SPOTIFY_ACCESS_TOKEN,
+            )
         )
 
-    def predict_genre(self, lyrics: str, *audio_features) -> dict:
+    def get_audio_features(self, spotify_track_link: str) -> list:
+        track_id = spotify_track_link.split("/")[-1].split("?")[0]
+        audio_features = self.sp.audio_features(track_id)
+        audio_features = [
+            audio_features[0][feature] for feature in ConfigApp.AUDIO_FEATURES
+        ]
+        return audio_features
+
+    def get_track_info(self, spotify_track_link: str) -> dict:
+        track_id = spotify_track_link.split("/")[-1].split("?")[0]
+        track_info = self.sp.track(track_id)
+        song_name = track_info.get("name", "Unknown")
+        artist_name = ", ".join(
+            [artist["name"] for artist in track_info.get("artists", [])]
+        )
+        print(f"Song Name: {song_name}, Artist Name: {artist_name}")
+        return {"Song Name": song_name, "Artist Name": artist_name}
+
+    def predict_genre(self, lyrics: str, spotify_track_link: str) -> dict:
         input_lyrics = self.tokenizer(
             lyrics, return_tensors="pt", padding=True, truncation=True, max_length=512
         )
+        audio_features = self.get_audio_features(spotify_track_link)
 
         outputs = self.model(**input_lyrics)
         lyrics_embedding = outputs.last_hidden_state.mean(dim=1)
@@ -59,70 +90,64 @@ class LyricsAudioModelInference:
         for i in range(3):
             genre_idx = top3_genres.indices[0][i].item()
             genre_prob = top3_genres.values[0][i].item()
-            genre_label = [key for key, value in GENRE.items() if value == genre_idx][0]
+            genre_label = [
+                key.capitalize()
+                for key, value in ConfigApp.GENRE.items()
+                if value == genre_idx
+            ][0]
             result[genre_label] = genre_prob
 
+        # track_info = self.get_track_info(spotify_track_link)
         return result
 
 
-if __name__ == "__main__":
+with gr.Blocks() as demo:
     iface = gr.Interface(
-        fn=LyricsAudioModelInference(model_name=REPO_NAME).predict_genre,
+        api_name="Music Genre Classifier",
+        fn=LyricsAudioModelInference(model_name=ConfigApp.REPO_NAME).predict_genre,
         inputs=[
             gr.Textbox(
-                lines=20,
+                lines=5,
                 placeholder="Enter lyrics here...",
                 label="Lyrics",
             ),
-            gr.Slider(
-                minimum=0,
-                maximum=1,
-                label="Acousticness",
-                step=0.01,
-            ),
-            gr.Slider(
-                minimum=0,
-                maximum=1,
-                label="Danceability",
-                step=0.01,
+            gr.Textbox(
+                lines=1,
+                placeholder="Enter Spotify Track Link here...",
+                label="Spotify Track Link",
             ),
-            gr.Slider(minimum=0, maximum=1, label="Energy", step=0.01),
-            gr.Slider(
-                minimum=0,
-                maximum=1,
-                label="Instrumentalness",
-                step=0.01,
+        ],
+        outputs=[
+            gr.Label(
+                num_top_classes=3,
+                label="Top 3 Predicted Genres",
+                elem_id="top3-genres",
             ),
-            gr.Slider(minimum=0, maximum=11, label="Key", step=1),
-            gr.Slider(minimum=0, maximum=1, label="Liveness", step=0.01),
-            gr.Slider(minimum=-60, maximum=0, label="Loudness", step=1),
-            gr.Slider(minimum=0, maximum=1, label="Mode", step=1),
-            gr.Slider(minimum=0, maximum=1, label="Speechiness", step=0.01),
-            gr.Slider(minimum=0, maximum=200, label="Tempo", step=1),
-            gr.Slider(minimum=0, maximum=1, label="Valence", step=0.01),
         ],
-        outputs=gr.Label(
-            num_top_classes=3,
-            label="Top 3 Predicted Genres",
-        ),
-        title="Music Genre Classifier",
+        title="🎷 Music Genre Classifier",
         description="This model predicts the genre of a song based on its lyrics and audio features.",
         examples=[
             [
                 "When the sun is rising over streets so barren...",
-                0.7050,
-                0.420,
-                0.247,
-                0.00349,
-                2,
-                0.1270,
-                -13.370,
-                0,
-                0.0360,
-                88.071,
-                0.138,
+                "https://open.spotify.com/track/2rGS4ipEZzldN0EpcfH3PK",
+            ],
+            [
+                "Tastes like strawberries On a summer evenin'...",
+                "https://open.spotify.com/track/6UelLqGlWMcVH1E5c4H7lY",
+            ],
+            [
+                """Da, da, da, da, da
+It's the motherfuckin' D-O-double-G (Snoop Dogg!)
+Da, da, da, da, da
+You know I'm mobbin' with the D.R.E. (Yeah, yeah, yeah)
+You know who's back up in this motherfucker! (What, what, what, what?)
+So blaze the weed up then! (Blaze it up, blaze it up!)
+Blaze that shit up, nigga... yeah 'Sup Snoop?""",
+                "https://open.spotify.com/track/4LwU4Vp6od3Sb08CsP99GC",
             ],
         ],
+        analytics_enabled=True,
     )
 
-    iface.launch(debug=True, show_api=True, share=True, inline=True)
+
+demo.launch(debug=True, show_api=True)

requirements.txt

@@ -1,4 +1,5 @@
 torch==2.2.1
 numpy==1.26.4
 gradio==4.21.0
+spotipy==2.23.0
 transformers==4.38.0