init commit

README.md

@@ -1,12 +1,13 @@
 ---
-title: Audioldm2 Api
-emoji: 🐨
-colorFrom: purple
-colorTo: pink
+title: AudioLDM2 API
+emoji: 🚀
+colorFrom: yellow
+colorTo: green
 sdk: gradio
-sdk_version: 4.3.0
+sdk_version: 3.41.2
 app_file: app.py
 pinned: false
+license: openrail
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,44 @@
+import gradio as gr
+import torch
+from diffusers import AudioLDM2Pipeline
+
+# make Space compatible with CPU duplicates
+if torch.cuda.is_available():
+    device = "cuda"
+    torch_dtype = torch.float16
+else:
+    device = "cpu"
+    torch_dtype = torch.float32
+
+# load the diffusers pipeline
+repo_id = "cvssp/audioldm2"
+pipe = AudioLDM2Pipeline.from_pretrained(repo_id, torch_dtype=torch_dtype).to(device)
+# pipe.unet = torch.compile(pipe.unet)
+
+# set the generator for reproducibility
+generator = torch.Generator(device)
+
+
+def text2audio(text, negative_prompt, duration, guidance_scale, random_seed, n_candidates):
+    if text is None:
+        raise gr.Error("Please provide a text input.")
+
+    waveforms = pipe(
+        text,
+        audio_length_in_s=duration,
+        guidance_scale=guidance_scale,
+        num_inference_steps=200,
+        negative_prompt=negative_prompt,
+        num_waveforms_per_prompt=n_candidates if n_candidates else 1,
+        generator=generator.manual_seed(int(random_seed)),
+    )["audios"]
+
+    return gr.make_waveform((16000, waveforms[0]), bg_image="bg.png")
+
+gradio_interface = gr.Interface(
+	fn = my_inference_function,
+	inputs = "text",
+	outputs = "audio",
+)
+
+gradio_interface.launch()

requirements.txt

@@ -0,0 +1,3 @@
+diffusers
+torch
+gradio