hgfh

call_handler.py

@@ -1,14 +1,19 @@
 from handler import EndpointHandler
+import base64
 
 # init handler
 my_handler = EndpointHandler(path=".")
 
 # prepare sample payload
-non_holiday_payload = {"inputs": "I am quite excited how this will turn out", }
-holiday_payload = {"inputs": "Today is a though day"}
+text_payload = {"text": "I am quite excited how this will turn out"}
+audio_payload = {"audio": base64.b64encode(
+    open("/home/arthur/data/musicdb/split_demucses/474499231_456751864 --__-- Breakbot - Programme_mp3_drums/chunk_11.wav", 'rb').read())}
+
 
 # test the handler
-non_holiday_pred=my_handler(non_holiday_payload)
+# text_pred = my_handler(text_payload)
+audio_pred = my_handler(audio_payload)
 
 # show results
-print("holiday_payload", holiday_payload)
+# print("text_pred", text_pred)
+print("audio_pred", audio_pred)

handler.py

@@ -1,6 +1,11 @@
+# import io
 from typing import Dict, List, Any
+
+# import librosa
 from transformers import ClapModel, ClapProcessor
-import gc
+# import gc
+# import base64
+
 
 class EndpointHandler:
     def __init__(self, path=""):
@@ -12,13 +17,20 @@ class EndpointHandler:
         """
          data args:
               inputs (:obj: `str`)
-              date (:obj: `str`)
         Return:
               A :obj:`list` | `dict`: will be serialized and returned
         """
-        print(type(data))
-        query = data['inputs']
-        text_inputs = self.processor(text=query, return_tensors="pt")
-        text_embed = self.model.get_text_features(**text_inputs)[0]
-        gc.collect()
-        return text_embed.detach().numpy()
+        # print(type(data))
+        if 'text' in data:
+            query = data['text']
+            text_inputs = self.processor(text=query, return_tensors="pt")
+            text_embed = self.model.get_text_features(**text_inputs)[0]
+            return text_embed.detach().numpy()
+
+        # if 'audio' in data:
+        #     # Load the audio data into librosa
+        #     audio_buffer = io.BytesIO(base64.b64decode(data['audio']))
+        #     y, sr = librosa.load(audio_buffer, sr=48000)
+        #     inputs = self.processor(audios=y, sampling_rate=sr, return_tensors="pt")
+        #     embedding = self.model.get_audio_features(**inputs)[0]
+        #     gc.collect()