Upload 2 files
Browse files- txt2vid_1.py +37 -0
- txt2vid_2.py +35 -0
txt2vid_1.py
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
!pip install diffusers --upgrade
|
2 |
+
|
3 |
+
!pip install invisible_watermark transformers accelerate safetensors
|
4 |
+
|
5 |
+
from kaggle_secrets import UserSecretsClient
|
6 |
+
user_secrets = UserSecretsClient()
|
7 |
+
secret_value_0 = user_secrets.get_secret("huggingface")
|
8 |
+
|
9 |
+
import torch
|
10 |
+
from diffusers import DiffusionPipeline
|
11 |
+
from diffusers.utils import export_to_video
|
12 |
+
import shutil # Import shutil for cross-filesystem moves
|
13 |
+
|
14 |
+
# Load the pipeline
|
15 |
+
pipe = DiffusionPipeline.from_pretrained("damo-vilab/text-to-video-ms-1.7b", torch_dtype=torch.float16, variant="fp16")
|
16 |
+
pipe = pipe.to("cuda")
|
17 |
+
|
18 |
+
# Generate video frames
|
19 |
+
prompt = "Spiderman is surfing"
|
20 |
+
video_frames = pipe(prompt).frames[0]
|
21 |
+
|
22 |
+
# Export video to a temporary path
|
23 |
+
video_path = export_to_video(video_frames) # This returns the path to the saved video
|
24 |
+
|
25 |
+
# Define the desired output path
|
26 |
+
output_path = "/kaggle/working/spiderman_surfing.mp4"
|
27 |
+
|
28 |
+
# Move the video to the desired output path using shutil.move
|
29 |
+
shutil.move(video_path, output_path)
|
30 |
+
print(f"Video saved at: {output_path}")
|
31 |
+
|
32 |
+
# Utility function to convert bytes to gigabytes
|
33 |
+
def bytes_to_giga_bytes(bytes):
|
34 |
+
return bytes / 1024 / 1024 / 1024
|
35 |
+
|
36 |
+
# Print memory usage
|
37 |
+
print(f"Max memory allocated: {bytes_to_giga_bytes(torch.cuda.max_memory_allocated())} GB")
|
txt2vid_2.py
ADDED
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from diffusers import DiffusionPipeline
|
3 |
+
from diffusers.utils import export_to_video
|
4 |
+
import shutil # Import shutil for cross-filesystem moves
|
5 |
+
|
6 |
+
# Clear GPU memory
|
7 |
+
torch.cuda.empty_cache()
|
8 |
+
|
9 |
+
# Set memory allocation configuration
|
10 |
+
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True
|
11 |
+
|
12 |
+
# Load the pipeline
|
13 |
+
pipe = DiffusionPipeline.from_pretrained("damo-vilab/text-to-video-ms-1.7b", torch_dtype=torch.float16, variant="fp16")
|
14 |
+
pipe = pipe.to("cuda")
|
15 |
+
|
16 |
+
# Generate video frames
|
17 |
+
prompt = "Spiderman is surfing"
|
18 |
+
video_frames = pipe(prompt, num_frames=75).frames[0]
|
19 |
+
|
20 |
+
# Export video to a temporary path
|
21 |
+
video_path = export_to_video(video_frames) # This returns the path to the saved video
|
22 |
+
|
23 |
+
# Define the desired output path
|
24 |
+
output_path = "/kaggle/working/spiderman_surfing.mp4"
|
25 |
+
|
26 |
+
# Move the video to the desired output path using shutil.move
|
27 |
+
shutil.move(video_path, output_path)
|
28 |
+
print(f"Video saved at: {output_path}")
|
29 |
+
|
30 |
+
# Utility function to convert bytes to gigabytes
|
31 |
+
def bytes_to_giga_bytes(bytes):
|
32 |
+
return bytes / 1024 / 1024 / 1024
|
33 |
+
|
34 |
+
# Print memory usage
|
35 |
+
print(f"Max memory allocated: {bytes_to_giga_bytes(torch.cuda.max_memory_allocated())} GB")
|