update

app.py

@@ -22,9 +22,16 @@ from opensora.models.diffusion.latte.modeling_latte import LatteT2V
 from opensora.sample.pipeline_videogen import VideoGenPipeline
 from opensora.serve.gradio_utils import block_css, title_markdown, randomize_seed_fn, set_env, examples, DESCRIPTION
 
-import spaces
+if torch.cuda.is_available():
+    _device = "cuda"
+elif torch.backends.mps.is_available():
+    _device = "mps"
+else:
+    _device = "cpu"
 
-@spaces.GPU(duration=250)
+#import spaces
+
+#@spaces.GPU(duration=250)
 def generate_img(prompt, sample_steps, scale, seed=0, randomize_seed=False, force_images=False):
     seed = int(randomize_seed_fn(seed, randomize_seed))
     set_env(seed)
@@ -42,7 +49,10 @@ def generate_img(prompt, sample_steps, scale, seed=0, randomize_seed=False, forc
                                mask_feature=True,
                                ).video
 
-    torch.cuda.empty_cache()
+    if _device == "cuda":
+        torch.cuda.empty_cache()
+    elif _device == "mps":
+        torch.mps.empty_cache()
     videos = videos[0]
     tmp_save_path = 'tmp.mp4'
     imageio.mimwrite(tmp_save_path, videos, fps=24, quality=9)  # highest quality is 10, lowest is 0
@@ -57,7 +67,8 @@ if __name__ == '__main__':
         'text_encoder_name': 'DeepFloyd/t5-v1_1-xxl',
         'version': '65x512x512'
     })
-    device = torch.device('cuda:0')
+    #device = torch.device('cuda:0')
+    device = torch.device(_device)
 
     # Load model:
     transformer_model = LatteT2V.from_pretrained(args.model_path, subfolder=args.version, torch_dtype=torch.float16, cache_dir='cache_dir').to(device)
@@ -120,4 +131,4 @@ if __name__ == '__main__':
         examples=examples,
         cache_examples=False
     )
-    demo.launch()
+    demo.launch()

requirements.txt

@@ -1,10 +1,10 @@
-torch==2.1.2
-torchvision==0.16.2
+#torch==2.1.2
+#torchvision==0.16.2
 transformers==4.39.1
 accelerate==0.28.0
 albumentations==1.4.0
 av==11.0.0
-decord==0.6.0
+#decord==0.6.0
 einops==0.7.0
 fastapi==0.110.0
 gdown==5.1.0