updated app.py

app.py

@@ -119,20 +119,24 @@ DEFAULT_TEXT = "A person is "
 if not os.path.exists("checkpoints/t2m"):
     os.system("bash prepare/download_models_demo.sh")
 
+print(f"Is CUDA available: {torch.cuda.is_available()}")
+print("CUDA is not required; running on CPU.")
+
 ##########################
-######Preparing demo######
+###### Preparing Demo ####
 ##########################
 parser = EvalT2MOptions()
 opt = parser.parse()
-fixseed(opt.seed)
-opt.device = torch.device("cpu" if opt.gpu_id == -1 else "cuda:" + str(opt.gpu_id))
+
+# Force use of CPU
+opt.device = torch.device("cpu")
 dim_pose = 263
 root_dir = pjoin(opt.checkpoints_dir, opt.dataset_name, opt.name)
 model_dir = pjoin(root_dir, 'model')
 model_opt_path = pjoin(root_dir, 'opt.txt')
 model_opt = get_opt(model_opt_path, device=opt.device)
 
-######Loading RVQ######
+###### Loading RVQ ######
 vq_opt_path = pjoin(opt.checkpoints_dir, opt.dataset_name, model_opt.vq_name, 'opt.txt')
 vq_opt = get_opt(vq_opt_path, device=opt.device)
 vq_opt.dim_pose = dim_pose
@@ -142,19 +146,20 @@ model_opt.num_tokens = vq_opt.nb_code
 model_opt.num_quantizers = vq_opt.num_quantizers
 model_opt.code_dim = vq_opt.code_dim
 
-######Loading R-Transformer######
+###### Loading R-Transformer ######
 res_opt_path = pjoin(opt.checkpoints_dir, opt.dataset_name, opt.res_name, 'opt.txt')
 res_opt = get_opt(res_opt_path, device=opt.device)
 res_model = load_res_model(res_opt, vq_opt, opt)
 
 assert res_opt.vq_name == model_opt.vq_name
 
-######Loading M-Transformer######
+###### Loading M-Transformer ######
 t2m_transformer = load_trans_model(model_opt, opt, 'latest.tar')
 
-#####Loading Length Predictor#####
+##### Loading Length Predictor #####
 length_estimator = load_len_estimator(model_opt)
 
+# Set models to evaluation mode and move to CPU
 t2m_transformer.eval()
 vq_model.eval()
 res_model.eval()
@@ -168,6 +173,7 @@ length_estimator.to(opt.device)
 opt.nb_joints = 22
 mean = np.load(pjoin(opt.checkpoints_dir, opt.dataset_name, model_opt.vq_name, 'meta', 'mean.npy'))
 std = np.load(pjoin(opt.checkpoints_dir, opt.dataset_name, model_opt.vq_name, 'meta', 'std.npy'))
+
 def inv_transform(data):
     return data * std + mean