Update demos/musicgen_app.py

demos/musicgen_app.py

@@ -93,15 +93,16 @@ def make_waveform(*args, **kwargs):
         return out
 
 
-def load_model(version="facebook/musicgen-melody"):
+def load_model(version="facebook/musicgen-small"):
     global MODEL
-    print("Loading model", version)
+    print("Loading Musivesal musicgen-small")  # , version
     if MODEL is None or MODEL.name != version:
         # Clear PyTorch CUDA cache and delete model
         del MODEL
         torch.cuda.empty_cache()
         MODEL = None  # in case loading would crash
-        MODEL = MusicGen.get_pretrained(version)
+        MODEL = MusicGen.get_pretrained("/data")
+        print("Custom model loaded.")
 
 
 def load_diffusion():
@@ -290,9 +291,10 @@ def ui_full(launch_kwargs):
         gr.Markdown(
             """
             # MusicGen
-            This is your private demo for [MusicGen](https://github.com/facebookresearch/audiocraft),
+            This is a private demo of [MusicGen](https://github.com/facebookresearch/audiocraft),
             a simple and controllable model for music generation
-            presented at: ["Simple and Controllable Music Generation"](https://huggingface.co/papers/2306.05284)
+            presented at: ["Simple and Controllable Music Generation"](https://huggingface.co/papers/2306.05284).
+            This Space hosts **"facebook/musicgen-small"**. It has been finetuned on a proprietary keyboard dataset from [Musiversal](https://musiversal.com/).
             """
         )
         with gr.Row():
@@ -316,25 +318,25 @@ def ui_full(launch_kwargs):
                     submit = gr.Button("Submit")
                     # Adapted from https://github.com/rkfg/audiocraft/blob/long/app.py, MIT license.
                     _ = gr.Button("Interrupt").click(fn=interrupt, queue=False)
-                with gr.Row():
-                    model = gr.Radio(
-                        [
-                            "facebook/musicgen-melody",
-                            "facebook/musicgen-medium",
-                            "facebook/musicgen-small",
-                            "facebook/musicgen-large",
-                            "facebook/musicgen-melody-large",
-                            "facebook/musicgen-stereo-small",
-                            "facebook/musicgen-stereo-medium",
-                            "facebook/musicgen-stereo-melody",
-                            "facebook/musicgen-stereo-large",
-                            "facebook/musicgen-stereo-melody-large",
-                        ],
-                        label="Model",
-                        value="facebook/musicgen-stereo-melody",
-                        interactive=True,
-                    )
-                    model_path = gr.Text(label="Model Path (custom models)")
+                # with gr.Row():
+                #     model = gr.Radio(
+                #         [
+                #             "facebook/musicgen-melody",
+                #             "facebook/musicgen-medium",
+                #             "facebook/musicgen-small",
+                #             "facebook/musicgen-large",
+                #             "facebook/musicgen-melody-large",
+                #             "facebook/musicgen-stereo-small",
+                #             "facebook/musicgen-stereo-medium",
+                #             "facebook/musicgen-stereo-melody",
+                #             "facebook/musicgen-stereo-large",
+                #             "facebook/musicgen-stereo-melody-large",
+                #         ],
+                #         label="Model",
+                #         value="facebook/musicgen-stereo-melody",
+                #         interactive=True,
+                #     )
+                #     model_path = gr.Text(label="Model Path (custom models)")
                 with gr.Row():
                     decoder = gr.Radio(
                         ["Default", "MultiBand_Diffusion"],
@@ -375,8 +377,8 @@ def ui_full(launch_kwargs):
         ).then(
             predict_full,
             inputs=[
-                model,
-                model_path,
+                # model,
+                # model_path,
                 decoder,
                 text,
                 melody,
@@ -392,49 +394,49 @@ def ui_full(launch_kwargs):
             toggle_audio_src, radio, [melody], queue=False, show_progress=False
         )
 
-        gr.Examples(
-            fn=predict_full,
-            examples=[
-                [
-                    "An 80s driving pop song with heavy drums and synth pads in the background",
-                    "./assets/bach.mp3",
-                    "facebook/musicgen-stereo-melody",
-                    "Default",
-                ],
-                [
-                    "A cheerful country song with acoustic guitars",
-                    "./assets/bolero_ravel.mp3",
-                    "facebook/musicgen-stereo-melody",
-                    "Default",
-                ],
-                [
-                    "90s rock song with electric guitar and heavy drums",
-                    None,
-                    "facebook/musicgen-stereo-medium",
-                    "Default",
-                ],
-                [
-                    "a light and cheerly EDM track, with syncopated drums, aery pads, and strong emotions",
-                    "./assets/bach.mp3",
-                    "facebook/musicgen-stereo-melody",
-                    "Default",
-                ],
-                [
-                    "lofi slow bpm electro chill with organic samples",
-                    None,
-                    "facebook/musicgen-stereo-medium",
-                    "Default",
-                ],
-                [
-                    "Punk rock with loud drum and power guitar",
-                    None,
-                    "facebook/musicgen-stereo-medium",
-                    "MultiBand_Diffusion",
-                ],
-            ],
-            inputs=[text, melody, model, decoder],
-            outputs=[output],
-        )
+        # gr.Examples(
+        #     fn=predict_full,
+        #     examples=[
+        #         [
+        #             "An 80s driving pop song with heavy drums and synth pads in the background",
+        #             "./assets/bach.mp3",
+        #             "facebook/musicgen-stereo-melody",
+        #             "Default",
+        #         ],
+        #         [
+        #             "A cheerful country song with acoustic guitars",
+        #             "./assets/bolero_ravel.mp3",
+        #             "facebook/musicgen-stereo-melody",
+        #             "Default",
+        #         ],
+        #         [
+        #             "90s rock song with electric guitar and heavy drums",
+        #             None,
+        #             "facebook/musicgen-stereo-medium",
+        #             "Default",
+        #         ],
+        #         [
+        #             "a light and cheerly EDM track, with syncopated drums, aery pads, and strong emotions",
+        #             "./assets/bach.mp3",
+        #             "facebook/musicgen-stereo-melody",
+        #             "Default",
+        #         ],
+        #         [
+        #             "lofi slow bpm electro chill with organic samples",
+        #             None,
+        #             "facebook/musicgen-stereo-medium",
+        #             "Default",
+        #         ],
+        #         [
+        #             "Punk rock with loud drum and power guitar",
+        #             None,
+        #             "facebook/musicgen-stereo-medium",
+        #             "MultiBand_Diffusion",
+        #         ],
+        #     ],
+        #     inputs=[text, melody, model, decoder],
+        #     outputs=[output],
+        # )
         gr.Markdown(
             """
             ### More details
@@ -459,7 +461,7 @@ def ui_full(launch_kwargs):
             An overlap of 12 seconds is kept with the previously generated chunk, and 18 "new" seconds
             are generated each time.
 
-            We present 10 model variations:
+            There 10 model variations:
             1. facebook/musicgen-melody -- a music generation model capable of generating music condition
                 on text and melody inputs. **Note**, you can also use text only.
             2. facebook/musicgen-small -- a 300M transformer decoder conditioned on text only.
@@ -468,6 +470,8 @@ def ui_full(launch_kwargs):
             5. facebook/musicgen-melody-large -- a 3.3B transformer decoder conditioned on and melody.
             6. facebook/musicgen-stereo-*: same as the previous models but fine tuned to output stereo audio.
 
+            **This is space only provides Musiversal's finetuning of 'facebook/musicgen-small'.**
+
             We also present two way of decoding the audio tokens
             1. Use the default GAN based compression model. It can suffer from artifacts especially
                 for crashes, snares etc.