Upgrade playground app to Streamlit 1.18+

The first change was using the new non-experimental cache decorators,
but then I decided to refactor to get rid of using the streamlit pages
feature and instead have my own dropdown. This allows for more control
to fix a page layout issue that popped up with this version.

requirements.txt

@@ -13,7 +13,7 @@ pysoundfile
 scipy
 soundfile
 sox
-streamlit>=1.10.0
+streamlit>=1.18.0
 torch
 torchaudio
 torchvision

riffusion/streamlit/playground.py

@@ -1,43 +1,29 @@
 import streamlit as st
 
-
-def render_main():
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.title(":guitar: Riffusion Playground")
-
-    left, right = st.columns(2)
-
-    with left:
-        create_link(":pencil2: Text to Audio", "/text_to_audio")
-        st.write("Generate audio clips from text prompts.")
-
-        create_link(":wave: Audio to Audio", "/audio_to_audio")
-        st.write("Upload audio and modify with text prompt (interpolation supported).")
-
-        create_link(":performing_arts: Interpolation", "/interpolation")
-        st.write("Interpolate between prompts in the latent space.")
-
-        create_link(":scissors: Audio Splitter", "/split_audio")
-        st.write("Split audio into stems like vocals, bass, drums, guitar, etc.")
-
-    with right:
-        create_link(":scroll: Text to Audio Batch", "/text_to_audio_batch")
-        st.write("Generate audio in batch from a JSON file of text prompts.")
-
-        create_link(":paperclip: Sample Clips", "/sample_clips")
-        st.write("Export short clips from an audio file.")
-
-        create_link(":musical_keyboard: Image to Audio", "/image_to_audio")
-        st.write("Reconstruct audio from spectrogram images.")
+PAGES = {
+    "🎛️ Home": "tasks.home",
+    "🌊 Text to Audio": "tasks.text_to_audio",
+    "✨ Audio to Audio": "tasks.audio_to_audio",
+    "🎭 Interpolation": "tasks.interpolation",
+    "✂️ Audio Splitter": "tasks.split_audio",
+    "📜 Text to Audio Batch": "tasks.text_to_audio_batch",
+    "📎 Sample Clips": "tasks.sample_clips",
+    "⏈ Spectrogram to Audio": "tasks.image_to_audio",
+}
 
 
-def create_link(name: str, url: str) -> None:
-    st.markdown(
-        f"### <a href='{url}' target='_self' style='text-decoration: none;'>{name}</a>",
-        unsafe_allow_html=True,
+def main() -> None:
+    st.set_page_config(
+        page_title="Riffusion Playground",
+        page_icon="🎸",
+        layout="wide",
     )
 
+    page = st.sidebar.selectbox("Page", list(PAGES.keys()))
+    assert page is not None
+    module = __import__(PAGES[page], fromlist=["render"])
+    module.render()
+
 
 if __name__ == "__main__":
-    render_main()
+    main()

riffusion/streamlit/tasks/audio_to_audio.py

@@ -10,14 +10,12 @@ from PIL import Image
 from riffusion.datatypes import InferenceInput, PromptInput
 from riffusion.spectrogram_params import SpectrogramParams
 from riffusion.streamlit import util as streamlit_util
-from riffusion.streamlit.pages.interpolation import get_prompt_inputs, run_interpolation
+from riffusion.streamlit.tasks.interpolation import get_prompt_inputs, run_interpolation
 from riffusion.util import audio_util
 
 
-def render_audio_to_audio() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":wave: Audio to Audio")
+def render() -> None:
+    st.subheader("✨ Audio to Audio")
     st.write(
         """
     Modify existing audio from a text prompt or interpolate between two.
@@ -408,7 +406,3 @@ def scale_image_to_32_stride(image: Image.Image) -> Image.Image:
     closest_width = int(np.ceil(image.width / 32) * 32)
     closest_height = int(np.ceil(image.height / 32) * 32)
     return image.resize((closest_width, closest_height), Image.BICUBIC)
-
-
-if __name__ == "__main__":
-    render_audio_to_audio()

riffusion/streamlit/tasks/home.py

@@ -0,0 +1,32 @@
+import streamlit as st
+
+
+def render():
+    st.title("✨🎸 Riffusion Playground 🎸✨")
+
+    st.write("Select a task from the sidebar to get started!")
+
+    left, right = st.columns(2)
+
+    with left:
+        st.subheader("🌊 Text to Audio")
+        st.write("Generate audio clips from text prompts.")
+
+        st.subheader("✨ Audio to Audio")
+        st.write("Upload audio and modify with text prompt (interpolation supported).")
+
+        st.subheader("🎭 Interpolation")
+        st.write("Interpolate between prompts in the latent space.")
+
+        st.subheader("✂️ Audio Splitter")
+        st.write("Split audio into stems like vocals, bass, drums, guitar, etc.")
+
+    with right:
+        st.subheader("📜 Text to Audio Batch")
+        st.write("Generate audio in batch from a JSON file of text prompts.")
+
+        st.subheader("📎 Sample Clips")
+        st.write("Export short clips from an audio file.")
+
+        st.subheader("⏈ Spectrogram to Audio")
+        st.write("Reconstruct audio from spectrogram images.")

riffusion/streamlit/tasks/image_to_audio.py

@@ -9,10 +9,8 @@ from riffusion.streamlit import util as streamlit_util
 from riffusion.util.image_util import exif_from_image
 
 
-def render_image_to_audio() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":musical_keyboard: Image to Audio")
+def render() -> None:
+    st.subheader("⏈ Image to Audio")
     st.write(
         """
     Reconstruct audio from spectrogram images.
@@ -77,7 +75,3 @@ def render_image_to_audio() -> None:
         name=Path(image_file.name).stem,
         extension=extension,
     )
-
-
-if __name__ == "__main__":
-    render_image_to_audio()

riffusion/streamlit/tasks/interpolation.py

@@ -13,10 +13,8 @@ from riffusion.spectrogram_params import SpectrogramParams
 from riffusion.streamlit import util as streamlit_util
 
 
-def render_interpolation() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":performing_arts: Interpolation")
+def render() -> None:
+    st.subheader("🎭 Interpolation")
     st.write(
         """
     Interpolate between prompts in the latent space.
@@ -241,7 +239,7 @@ def get_prompt_inputs(
     return p
 
 
-@st.experimental_memo
+@st.cache_data
 def run_interpolation(
     inputs: InferenceInput, init_image: Image.Image, device: str = "cuda", extension: str = "mp3"
 ) -> T.Tuple[Image.Image, io.BytesIO]:
@@ -275,7 +273,3 @@ def run_interpolation(
     )
 
     return image, audio_bytes
-
-
-if __name__ == "__main__":
-    render_interpolation()

riffusion/streamlit/tasks/sample_clips.py

@@ -10,10 +10,8 @@ from riffusion.spectrogram_params import SpectrogramParams
 from riffusion.streamlit import util as streamlit_util
 
 
-def render_sample_clips() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":paperclip: Sample Clips")
+def render() -> None:
+    st.subheader("📎 Sample Clips")
     st.write(
         """
     Export short clips from an audio file.
@@ -125,7 +123,3 @@ def render_sample_clips() -> None:
 
     if save_to_disk:
         st.info(f"Wrote {num_clips} clip(s) to `{str(output_path)}`")
-
-
-if __name__ == "__main__":
-    render_sample_clips()

riffusion/streamlit/tasks/split_audio.py

@@ -9,10 +9,8 @@ from riffusion.streamlit import util as streamlit_util
 from riffusion.util import audio_util
 
 
-def render_split_audio() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":scissors: Audio Splitter")
+def render() -> None:
+    st.subheader("✂️ Audio Splitter")
     st.write(
         """
     Split audio into individual instrument stems.
@@ -99,7 +97,3 @@ def split_audio_cached(
     segment: pydub.AudioSegment, device: str = "cuda"
 ) -> T.Dict[str, pydub.AudioSegment]:
     return split_audio(segment, device=device)
-
-
-if __name__ == "__main__":
-    render_split_audio()

riffusion/streamlit/tasks/text_to_audio.py

@@ -6,10 +6,8 @@ from riffusion.spectrogram_params import SpectrogramParams
 from riffusion.streamlit import util as streamlit_util
 
 
-def render_text_to_audio() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":pencil2: Text to Audio")
+def render() -> None:
+    st.subheader("🌊 Text to Audio")
     st.write(
         """
     Generate audio from text prompts.
@@ -119,7 +117,3 @@ def render_text_to_audio() -> None:
         )
 
         seed += 1
-
-
-if __name__ == "__main__":
-    render_text_to_audio()

riffusion/streamlit/tasks/text_to_audio_batch.py

@@ -14,7 +14,7 @@ EXAMPLE_INPUT = """
     "seed": 42,
     "num_inference_steps": 50,
     "guidance": 7.0,
-    "width": 512,
+    "width": 512
   },
   "entries": [
     {
@@ -32,10 +32,8 @@ EXAMPLE_INPUT = """
 """
 
 
-def render_text_to_audio_batch() -> None:
-    st.set_page_config(layout="wide", page_icon="🎸")
-
-    st.subheader(":scroll: Text to Audio Batch")
+def render() -> None:
+    st.subheader("📜 Text to Audio Batch")
     st.write(
         """
     Generate audio in batch from a JSON file of text prompts.
@@ -141,7 +139,3 @@ def render_text_to_audio_batch() -> None:
         st.info(f"Output written to {str(output_path)}")
     else:
         st.info("Enter output directory in sidebar to save to disk")
-
-
-if __name__ == "__main__":
-    render_text_to_audio_batch()

riffusion/streamlit/util.py

@@ -33,7 +33,7 @@ SCHEDULER_OPTIONS = [
 ]
 
 
-@st.experimental_singleton
+@st.cache_resource
 def load_riffusion_checkpoint(
     checkpoint: str = DEFAULT_CHECKPOINT,
     no_traced_unet: bool = False,
@@ -49,7 +49,7 @@ def load_riffusion_checkpoint(
     )
 
 
-@st.experimental_singleton
+@st.cache_resource
 def load_stable_diffusion_pipeline(
     checkpoint: str = DEFAULT_CHECKPOINT,
     device: str = "cuda",
@@ -109,7 +109,7 @@ def get_scheduler(scheduler: str, config: T.Any) -> T.Any:
         raise ValueError(f"Unknown scheduler {scheduler}")
 
 
-@st.experimental_singleton
+@st.cache_resource
 def pipeline_lock() -> threading.Lock:
     """
     Singleton lock used to prevent concurrent access to any model pipeline.
@@ -117,7 +117,7 @@ def pipeline_lock() -> threading.Lock:
     return threading.Lock()
 
 
-@st.experimental_singleton
+@st.cache_resource
 def load_stable_diffusion_img2img_pipeline(
     checkpoint: str = DEFAULT_CHECKPOINT,
     device: str = "cuda",
@@ -145,7 +145,7 @@ def load_stable_diffusion_img2img_pipeline(
     return pipeline
 
 
-@st.experimental_memo
+@st.cache_data
 def run_txt2img(
     prompt: str,
     num_inference_steps: int,
@@ -184,7 +184,7 @@ def run_txt2img(
         return output["images"][0]
 
 
-@st.experimental_singleton
+@st.cache_resource
 def spectrogram_image_converter(
     params: SpectrogramParams,
     device: str = "cuda",
@@ -202,7 +202,7 @@ def spectrogram_image_from_audio(
     return converter.spectrogram_image_from_audio(segment)
 
 
-@st.experimental_memo
+@st.cache_data
 def audio_segment_from_spectrogram_image(
     image: Image.Image,
     params: SpectrogramParams,
@@ -212,7 +212,7 @@ def audio_segment_from_spectrogram_image(
     return converter.audio_from_spectrogram_image(image)
 
 
-@st.experimental_memo
+@st.cache_data
 def audio_bytes_from_spectrogram_image(
     image: Image.Image,
     params: SpectrogramParams,
@@ -289,17 +289,17 @@ def select_checkpoint(container: T.Any = st.sidebar) -> str:
     return custom_checkpoint or DEFAULT_CHECKPOINT
 
 
-@st.experimental_memo
+@st.cache_data
 def load_audio_file(audio_file: io.BytesIO) -> pydub.AudioSegment:
     return pydub.AudioSegment.from_file(audio_file)
 
 
-@st.experimental_singleton
+@st.cache_resource
 def get_audio_splitter(device: str = "cuda"):
     return AudioSplitter(device=device)
 
 
-@st.experimental_singleton
+@st.cache_resource
 def load_magic_mix_pipeline(
     checkpoint: str = DEFAULT_CHECKPOINT,
     device: str = "cuda",