Update requirements.txt to be git repo

app.py

@@ -102,7 +102,7 @@ def load_model(version='melody'):
         MODEL = MusicGen.get_pretrained(version, device=device)
 
 
-def _do_predictions(texts, melodies, duration, progress=False, **gen_kwargs):
+def _do_predictions(texts, melodies, duration, progress=False, drums=True, **gen_kwargs):
     MODEL.set_generation_params(duration=duration, **gen_kwargs)
     print("new batch", len(texts), texts, [None if m is None else (m[0], m[1].shape) for m in melodies])
     be = time.time()
@@ -135,14 +135,15 @@ def _do_predictions(texts, melodies, duration, progress=False, **gen_kwargs):
     out_files = []
     for output in outputs:
         # Demucs
-        print("Running demucs")
-        wav = convert_audio(output, MODEL.sample_rate, demucs_model.samplerate, demucs_model.audio_channels)
-        wav = wav.unsqueeze(0)
-        stems = apply_model(demucs_model, wav)
-        stems = stems[:, stem_idx]  # extract stem
-        stems = stems.sum(1)  # merge extracted stems
-        stems = convert_audio(stems, demucs_model.samplerate, MODEL.sample_rate, 1)
-        output = stems[0]
+        if not drums:
+            print("Running demucs")
+            wav = convert_audio(output, MODEL.sample_rate, demucs_model.samplerate, demucs_model.audio_channels)
+            wav = wav.unsqueeze(0)
+            stems = apply_model(demucs_model, wav)
+            stems = stems[:, stem_idx]  # extract stem
+            stems = stems.sum(1)  # merge extracted stems
+            stems = convert_audio(stems, demucs_model.samplerate, MODEL.sample_rate, 1)
+            output = stems[0]
         with NamedTemporaryFile("wb", suffix=".wav", delete=False) as file:
             audio_write(
                 file.name, output, MODEL.sample_rate, strategy="loudness",
@@ -168,7 +169,7 @@ def predict_batched(texts, melodies):
     return [res]
 
 
-def predict_full(model, text, melody, duration, topk, topp, temperature, cfg_coef, progress=gr.Progress()):
+def predict_full(model, text, melody, duration, topk, topp, temperature, cfg_coef, drums=drums, progress=gr.Progress()):
     global INTERRUPTING
     INTERRUPTING = False
     if temperature < 0:
@@ -189,7 +190,7 @@ def predict_full(model, text, melody, duration, topk, topp, temperature, cfg_coe
 
     outs = _do_predictions(
         [text], [melody], duration, progress=True,
-        top_k=topk, top_p=topp, temperature=temperature, cfg_coef=cfg_coef)
+        top_k=topk, top_p=topp, temperature=temperature, drums=drums, cfg_coef=cfg_coef)
 
 
 
@@ -228,10 +229,12 @@ def ui_full(launch_kwargs):
                     topp = gr.Number(label="Top-p", value=0, interactive=True)
                     temperature = gr.Number(label="Temperature", value=1.0, interactive=True)
                     cfg_coef = gr.Number(label="Classifier Free Guidance", value=3.0, interactive=True)
+                with gr.Row():
+                    drums = gr.Checkbox(label="Drums", value=True, interactive=True)
             with gr.Column():
                 output = gr.Video(label="Generated Music")
         submit.click(predict_full,
-                     inputs=[model, text, melody, duration, topk, topp, temperature, cfg_coef],
+                     inputs=[model, text, melody, duration, topk, topp, temperature, cfg_coef, drums],
                      outputs=[output])
         radio.change(toggle_audio_src, radio, [melody], queue=False, show_progress=False)
         gr.Markdown(

requirements.txt

@@ -12,7 +12,7 @@ arrow==1.2.3
 asttokens==2.2.1
 async-timeout==4.0.2
 attrs==23.1.0
-audiocraft==0.0.1
+audiocraft @ git+https://github.com/facebookresearch/audiocraft@0.0.2a2
 audioread==3.0.0
 av==10.0.0
 backcall==0.2.0