model = None

app.py

@@ -8,27 +8,27 @@ import random
 import spaces
 import torch
 
-from hyvideo.utils.file_utils import save_videos_grid
-from hyvideo.utils.preprocess_text_encoder_tokenizer_utils import preprocess_text_encoder_tokenizer
-from hyvideo.config import parse_args
-from hyvideo.inference import HunyuanVideoSampler
-from hyvideo.constants import NEGATIVE_PROMPT
+#from hyvideo.utils.file_utils import save_videos_grid
+#from hyvideo.utils.preprocess_text_encoder_tokenizer_utils import preprocess_text_encoder_tokenizer
+#from hyvideo.config import parse_args
+#from hyvideo.inference import HunyuanVideoSampler
+#from hyvideo.constants import NEGATIVE_PROMPT
 
-from huggingface_hub import snapshot_download
+#from huggingface_hub import snapshot_download
 
-if torch.cuda.device_count() > 0:
-    snapshot_download(repo_id="tencent/HunyuanVideo", repo_type="model", local_dir="ckpts", force_download=True)
-    snapshot_download(repo_id="xtuner/llava-llama-3-8b-v1_1-transformers", repo_type="model", local_dir="ckpts/llava-llama-3-8b-v1_1-transformers", force_download=True)
-
-    class Args:
-        def __init__(self, input_dir, output_dir):
-            self.input_dir = input_dir
-            self.output_dir = output_dir
-
-    # Create the object
-    args = Args("ckpts/llava-llama-3-8b-v1_1-transformers", "ckpts/text_encoder")
-    preprocess_text_encoder_tokenizer(args)
-    snapshot_download(repo_id="openai/clip-vit-large-patch14", repo_type="model", local_dir="ckpts/text_encoder_2", force_download=True)
+#if torch.cuda.device_count() > 0:
+#    snapshot_download(repo_id="tencent/HunyuanVideo", repo_type="model", local_dir="ckpts", force_download=True)
+#    snapshot_download(repo_id="xtuner/llava-llama-3-8b-v1_1-transformers", repo_type="model", local_dir="ckpts/llava-llama-3-8b-v1_1-transformers", force_download=True)#
+#
+#    class Args:
+#        def __init__(self, input_dir, output_dir):
+#            self.input_dir = input_dir
+#            self.output_dir = output_dir
+#
+#    # Create the object
+#    args = Args("ckpts/llava-llama-3-8b-v1_1-transformers", "ckpts/text_encoder")
+#    preprocess_text_encoder_tokenizer(args)
+#    snapshot_download(repo_id="openai/clip-vit-large-patch14", repo_type="model", local_dir="ckpts/text_encoder_2", force_download=True)
 
 def initialize_model(model_path):
     print("initialize_model: " + model_path)
@@ -45,7 +45,8 @@ def initialize_model(model_path):
     print("Model initialized: " + model_path)
     return hunyuan_video_sampler
 
-model = initialize_model("ckpts")
+#model = initialize_model("ckpts")
+model = None
 
 def generate_video(
     prompt,