Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
|
@@ -14,7 +14,7 @@ if DEVICE == "auto":
|
|
| 14 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
| 15 |
print(f"[SYSTEM] | Using {DEVICE} type compute device.")
|
| 16 |
|
| 17 |
-
DEFAULT_INPUT = "Describe in one
|
| 18 |
MAX_FRAMES = 64
|
| 19 |
|
| 20 |
model_name = "openbmb/MiniCPM-o-2_6"
|
|
@@ -85,7 +85,7 @@ def generate(input, instruction=DEFAULT_INPUT, sampling=False, temperature=0.7,
|
|
| 85 |
image = Image.open(input).convert("RGB")
|
| 86 |
content.append(image)
|
| 87 |
elif filetype in ["Video", "GIF"]:
|
| 88 |
-
frames = encode_gif(input) if filetype == "GIF" else encode_video(
|
| 89 |
content.extend(frames)
|
| 90 |
audio = librosa.load(input, sr=16000, mono=True)
|
| 91 |
content.append(audio)
|
|
@@ -95,7 +95,7 @@ def generate(input, instruction=DEFAULT_INPUT, sampling=False, temperature=0.7,
|
|
| 95 |
else:
|
| 96 |
return "Unsupported file type."
|
| 97 |
|
| 98 |
-
filename = os.path.basename(
|
| 99 |
prefix = input_prefixes[filetype].replace("█", filename)
|
| 100 |
content.append(prefix + instruction)
|
| 101 |
inputs_payload = [{"role": "user", "content": content}]
|
|
|
|
| 14 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
| 15 |
print(f"[SYSTEM] | Using {DEVICE} type compute device.")
|
| 16 |
|
| 17 |
+
DEFAULT_INPUT = "Describe in one short sentence."
|
| 18 |
MAX_FRAMES = 64
|
| 19 |
|
| 20 |
model_name = "openbmb/MiniCPM-o-2_6"
|
|
|
|
| 85 |
image = Image.open(input).convert("RGB")
|
| 86 |
content.append(image)
|
| 87 |
elif filetype in ["Video", "GIF"]:
|
| 88 |
+
frames = encode_gif(input) if filetype == "GIF" else encode_video(input)
|
| 89 |
content.extend(frames)
|
| 90 |
audio = librosa.load(input, sr=16000, mono=True)
|
| 91 |
content.append(audio)
|
|
|
|
| 95 |
else:
|
| 96 |
return "Unsupported file type."
|
| 97 |
|
| 98 |
+
filename = os.path.basename(input)
|
| 99 |
prefix = input_prefixes[filetype].replace("█", filename)
|
| 100 |
content.append(prefix + instruction)
|
| 101 |
inputs_payload = [{"role": "user", "content": content}]
|