[Playground] Tune parameters

Topic: tune_params_1
qinfrank · Jan 17, 2023 · a823173 · a823173
1 parent 66b0f03
commit a823173
Show file tree

Hide file tree

Showing 4 changed files with 27 additions and 10 deletions.
diff --git a/riffusion/streamlit/pages/audio_to_audio.py b/riffusion/streamlit/pages/audio_to_audio.py
@@ -52,7 +52,7 @@ def render_audio_to_audio() -> None:
         num_inference_steps = T.cast(
             int,
             st.number_input(
-                "Steps per sample", value=50, help="Number of denoising steps per model run"
+                "Steps per sample", value=25, help="Number of denoising steps per model run"
             ),
         )
 
@@ -115,17 +115,24 @@ def render_audio_to_audio() -> None:
 
     counter = streamlit_util.StreamlitCounter()
 
+    denoising_default = 0.55
     with st.form("audio to audio form"):
         if interpolate:
             left, right = st.columns(2)
 
             with left:
                 st.write("##### Prompt A")
-                prompt_input_a = PromptInput(guidance=guidance, **get_prompt_inputs(key="a"))
+                prompt_input_a = PromptInput(
+                    guidance=guidance,
+                    **get_prompt_inputs(key="a", denoising_default=denoising_default),
+                )
 
             with right:
                 st.write("##### Prompt B")
-                prompt_input_b = PromptInput(guidance=guidance, **get_prompt_inputs(key="b"))
+                prompt_input_b = PromptInput(
+                    guidance=guidance,
+                    **get_prompt_inputs(key="b", denoising_default=denoising_default),
+                )
         elif use_magic_mix:
             prompt = st.text_input("Prompt", key="prompt_a")
 
@@ -150,7 +157,12 @@ def render_audio_to_audio() -> None:
         else:
             prompt_input_a = PromptInput(
                 guidance=guidance,
-                **get_prompt_inputs(key="a", include_negative_prompt=True, cols=True),
+                **get_prompt_inputs(
+                    key="a",
+                    include_negative_prompt=True,
+                    cols=True,
+                    denoising_default=denoising_default,
+                ),
             )
 
         st.form_submit_button("Riff", type="primary", on_click=counter.increment)
@@ -319,7 +331,7 @@ def get_clip_params(advanced: bool = False) -> T.Dict[str, T.Any]:
     p["duration_s"] = cols[1].number_input(
         "Duration [s]",
         min_value=0.0,
-        value=15.0,
+        value=20.0,
     )
 
     if advanced:

diff --git a/riffusion/streamlit/pages/interpolation.py b/riffusion/streamlit/pages/interpolation.py
@@ -115,11 +115,15 @@ def render_interpolation() -> None:
 
         with left:
             st.write("##### Prompt A")
-            prompt_input_a = PromptInput(guidance=guidance, **get_prompt_inputs(key="a"))
+            prompt_input_a = PromptInput(
+                guidance=guidance, **get_prompt_inputs(key="a", denoising_default=0.75)
+            )
 
         with right:
             st.write("##### Prompt B")
-            prompt_input_b = PromptInput(guidance=guidance, **get_prompt_inputs(key="b"))
+            prompt_input_b = PromptInput(
+                guidance=guidance, **get_prompt_inputs(key="b", denoising_default=0.75)
+            )
 
         st.form_submit_button("Generate", type="primary")
 
@@ -201,6 +205,7 @@ def get_prompt_inputs(
     key: str,
     include_negative_prompt: bool = False,
     cols: bool = False,
+    denoising_default: float = 0.5,
 ) -> T.Dict[str, T.Any]:
     """
     Compute prompt inputs from widgets.
@@ -228,7 +233,7 @@ def get_prompt_inputs(
 
     p["denoising"] = right.number_input(
         "Denoising",
-        value=0.75,
+        value=denoising_default,
         key=f"denoising_{key}",
         help="How much to modify the seed image",
     )

diff --git a/riffusion/streamlit/pages/text_to_audio.py b/riffusion/streamlit/pages/text_to_audio.py
@@ -56,7 +56,7 @@ def render_text_to_audio() -> None:
         st.form_submit_button("Riff", type="primary")
 
     with st.sidebar:
-        num_inference_steps = T.cast(int, st.number_input("Inference steps", value=50))
+        num_inference_steps = T.cast(int, st.number_input("Inference steps", value=25))
         width = T.cast(int, st.number_input("Width", value=512))
         guidance = st.number_input(
             "Guidance", value=7.0, help="How much the model listens to the text prompt"

diff --git a/riffusion/streamlit/util.py b/riffusion/streamlit/util.py
@@ -22,12 +22,12 @@
 IMAGE_EXTENSIONS = ["png", "jpg", "jpeg"]
 
 SCHEDULER_OPTIONS = [
+    "DPMSolverMultistepScheduler",
     "PNDMScheduler",
     "DDIMScheduler",
     "LMSDiscreteScheduler",
     "EulerDiscreteScheduler",
     "EulerAncestralDiscreteScheduler",
-    "DPMSolverMultistepScheduler",
 ]