Update app.py

app.py

@@ -50,51 +50,7 @@ filetypes = {
     "Audio": [".wav", ".mp3", ".flac", ".aac"],
 }
 
-def uniform_sample(idxs, n):
-    gap = len(idxs) / n
-    return [idxs[int(i * gap + gap / 2)] for i in range(n)]
-
-def build_omni_chunks(path, prefix, instruction, sr=AUDIO_SR, seconds_per_unit=1):
-    clip = VideoFileClip(path, audio_fps=sr)
-    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp:
-        clip.audio.write_audiofile(tmp.name, fps=sr, codec="pcm_s16le", verbose=False, logger=None)
-        audio_np, _ = librosa.load(tmp.name, sr=sr, mono=True)
-    total_units = math.ceil(clip.duration / seconds_per_unit)
-    content = []
-    for i in range(total_units):
-        t = min(i * seconds_per_unit, clip.duration - 1e-3)
-        frame = Image.fromarray(clip.get_frame(t).astype("uint8")).convert("RGB")
-        audio_chunk = audio_np[sr * i * seconds_per_unit : sr * (i + 1) * seconds_per_unit]
-        content.extend(["<unit>", frame, audio_chunk])
-    clip.close()
-    os.remove(tmp.name)
-    content.append(prefix + instruction)
-    return content
-
-def build_image_omni(path, prefix, instruction):
-    image = Image.open(path).convert("RGB")
-    return ["<unit>", image, prefix + instruction]
-
-def encode_gif(path):
-    img = Image.open(path)
-    frames = [frame.copy().convert("RGB") for frame in ImageSequence.Iterator(img)]
-    if len(frames) > MAX_FRAMES:
-        frames = uniform_sample(frames, MAX_FRAMES)
-    return frames
-
-def build_gif_omni(path, prefix, instruction):
-    frames = encode_gif(path)
-    content = []
-    for f in frames:
-        content.extend(["<unit>", f])
-    content.append(prefix + instruction)
-    return content
-
-def build_audio_omni(path, prefix, instruction, sr=AUDIO_SR):
-    audio_np, _ = librosa.load(path, sr=sr, mono=True)
-    return ["<unit>", audio_np, prefix + instruction]
-
-ef infer_filetype(ext):
+def infer_filetype(ext):
     return next((k for k, v in filetypes.items() if ext in v), None)