test

VADER-VideoCrafter/lvdm/models/samplers/ddim.py

@@ -153,12 +153,12 @@ class DDIMSampler(object):
         else:
             img = x_T
         
-        print("x_T: ", x_T)
-        print("shape: ", shape)
-        print('random seed debug: ', torch.randn(100, device=device).sum())
-        print("Debug initial noise: ", torch.randn(shape, device=device).sum().item())
-        print("Debug initial noise: ", torch.randn(shape, device=device).sum().item())
-        print("noise device: ", img.device)
+        # print("x_T: ", x_T)
+        # print("shape: ", shape)
+        # print('random seed debug: ', torch.randn(100, device=device).sum())
+        # print("Debug initial noise: ", torch.randn(shape, device=device).sum().item())
+        # print("Debug initial noise: ", torch.randn(shape, device=device).sum().item())
+        # print("noise device: ", img.device)
         
         if timesteps is None:
             timesteps = self.ddpm_num_timesteps if ddim_use_original_steps else self.ddim_timesteps

VADER-VideoCrafter/scripts/main/train_t2v_lora.py

@@ -589,9 +589,9 @@ def run_training(args, model, **kwargs):
         # load the pretrained LoRA model
         peft.set_peft_model_state_dict(peft_model, torch.load(args.lora_ckpt_path))
 
-    print('random seed debug: ', torch.randn(100, device=accelerator.device).sum())
+    # print('random seed debug: ', torch.randn(100, device=accelerator.device).sum())
     print("precision: ", peft_model.dtype)
-    # precision of first_stage_model
+    # # precision of first_stage_model
     print("precision of first_stage_model: ", peft_model.first_stage_model.dtype)
     print("peft_model device: ", peft_model.device)
 
@@ -651,8 +651,7 @@ def run_training(args, model, **kwargs):
                     raise NotImplementedError
 
                 # Inference Step 3.2: inference, batch_samples shape: batch, <samples>, c, t, h, w
-                # no backprop_mode=args.backprop_mode because it is inference process
-                seed_everything_self(args.seed)
+                # no backprop_mode=args.backprop_mode because it is inference process 
                 if isinstance(peft_model, torch.nn.parallel.DistributedDataParallel):
                     batch_samples = batch_ddim_sampling(peft_model.module, cond, noise_shape, args.n_samples, \
                                                         args.ddim_steps, args.ddim_eta, args.unconditional_guidance_scale, None, decode_frame=args.decode_frame, **kwargs)

app.py

@@ -12,12 +12,13 @@ from train_t2v_lora import main_fn, setup_model
 
 examples = [
     ["A fairy tends to enchanted, glowing flowers.", 'huggingface-hps-aesthetic', 8, 400, 384, 512, 12.0, 25, 1.0, 24, 10],
-    ["A cat playing an electric guitar in a loft with industrial-style decor and soft, multicolored lights.", 'huggingface-hps-aesthetic', 8, 206, 384, 512, 12.0, 25, 1.0, 24, 10],
+    ["A cat playing an electric guitar in a loft with industrial-style decor and soft, multicolored lights.", 'huggingface-hps-aesthetic', 8, 208, 384, 512, 12.0, 25, 1.0, 24, 10],
     ["A raccoon playing a guitar under a blossoming cherry tree.", 'huggingface-hps-aesthetic', 8, 204, 384, 512, 12.0, 25, 1.0, 24, 10],
+    ["A raccoon playing an electric bass in a garage band setting.", 'huggingface-hps-aesthetic', 8, 400, 384, 512, 12.0, 25, 1.0, 24, 10],
     ["A mermaid with flowing hair and a shimmering tail discovers a hidden underwater kingdom adorned with coral palaces, glowing pearls, and schools of colorful fish, encountering both wonders and dangers along the way.", 
      "huggingface-pickscore", 16, 205, 384, 512, 12.0, 25, 1.0, 24, 10],
     ["A talking bird with shimmering feathers and a melodious voice leads an adventure to find a legendary treasure, guiding through enchanted forests, ancient ruins, and mystical challenges.",
-     "huggingface-pickscore", 16, 204, 384, 512, 12.0, 25, 1.0, 24, 10]
+     "huggingface-pickscore", 16, 600, 384, 512, 12.0, 25, 1.0, 24, 10]
 ]
 
 model = setup_model()