add tutorial

app.py

@@ -55,8 +55,7 @@ def get_args(parser: argparse.ArgumentParser):
         default=1.2,
         help="the temperature of the sampling operation",
     )
-    parser.add_argument("-seed", type=int, default=None,
-                        help="seed for randomstate")
+    parser.add_argument("-seed", type=int, default=None, help="seed for randomstate")
     parser.add_argument(
         "-show_control_code",
         type=bool,
@@ -150,10 +149,9 @@ def generate_music(
 
         else:
             prefix = patchilizer.decode(input_patches[0])
-            remaining_tokens = prompt[len(prefix):]
+            remaining_tokens = prompt[len(prefix) :]
             tokens = torch.tensor(
-                [patchilizer.bos_token_id] +
-                [ord(c) for c in remaining_tokens],
+                [patchilizer.bos_token_id] + [ord(c) for c in remaining_tokens],
                 device=DEVICE,
             )
 
@@ -224,8 +222,7 @@ def generate_music(
         tunes = tunes.replace(f"\nK:{K_val}\n", f"\nK:{K_val.split('m')[0]}\n")
 
     elif (mode == "minor") and (not "m" in K_val):
-        tunes = tunes.replace(
-            f"\nK:{K_val}\n", f"\nK:{K_val.replace('dor', '')}min\n")
+        tunes = tunes.replace(f"\nK:{K_val}\n", f"\nK:{K_val.replace('dor', '')}min\n")
 
     print("Generation time: {:.2f} seconds".format(time.time() - start_time))
     timestamp = time.strftime("%a_%d_%b_%Y_%H_%M_%S", time.localtime())
@@ -402,7 +399,8 @@ if __name__ == "__main__":
 
     with gr.Blocks() as demo:
         gr.Markdown(
-            "## The current CPU-based version on HuggingFace has slow inference, you can access the GPU-based mirror on [ModelScope](https://www.modelscope.cn/studios/monetjoe/EMusicGen)")
+            "## The current CPU-based version on HuggingFace has slow inference, you can access the GPU-based mirror on [ModelScope](https://www.modelscope.cn/studios/monetjoe/EMusicGen)"
+        )
         with gr.Row():
             with gr.Column():
                 dataset_option = gr.Dropdown(
@@ -411,7 +409,7 @@ if __name__ == "__main__":
                     value="Rough4Q",
                 )
                 gr.Markdown(
-                    "# Generate by emotion condition<br><img width='100%' src='https://www.modelscope.cn/studio/monetjoe/EMusicGen/resolve/master/4q.jpg'>"
+                    "# Generate by emotion condition<br><img width='100%' src='https://www.modelscope.cn/studio/monetjoe/EMusicGen/resolve/master/src/4q.jpg'>"
                 )
                 valence_radio = gr.Radio(
                     ["Low", "High"],
@@ -488,6 +486,11 @@ url       = {https://github.com/monetjoe/EMusicGen}
                 )
 
             with gr.Column():
+                gr.Markdown(
+                    """## Tutorial
+                    <video src='https://www.modelscope.cn/studio/monetjoe/EMusicGen/resolve/master/src/tutorial.mp4' controls></video>
+                    """
+                )
                 wav_audio = gr.Audio(label="Audio", type="filepath")
                 midi_file = gr.File(label="Download MIDI")
                 pdf_file = gr.File(label="Download PDF score")