Skip to content

Open-Sora: Democratizing Efficient Video Production for All

Python

debug

Shen-Chenhui8aad3d6de212
diff --git a/opensora/models/vae/vae_3d_v2.py b/opensora/models/vae/vae_3d_v2.py
index 374e877..0be32da 100644
--- a/opensora/models/vae/vae_3d_v2.py
+++ b/opensora/models/vae/vae_3d_v2.py
@@ -915,9 +915,19 @@ class VEALoss(nn.Module):
kl_loss_weight = 0.000001,
vgg=None,
vgg_weights: VGG16_Weights = VGG16_Weights.DEFAULT,
+ device = "cpu",
+ dtype = "bf16"
):
super().__init__()
+ if type(dtype) == str:
+ if dtype == "bf16":
+ dtype = torch.bfloat16
+ elif dtype == "fp16":
+ dtype = torch.float16
+ else:
+ raise NotImplementedError(f'dtype: {dtype}')
+
# KL Loss
self.kl_loss_weight = kl_loss_weight
# Perceptual Loss
@@ -929,7 +939,7 @@ class VEALoss(nn.Module):
weights = vgg_weights
)
vgg.classifier = Sequential(*vgg.classifier[:-2])
- self.vgg = vgg
+ self.vgg = vgg.to(device, dtype)
def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer):
breakpoint() # TODO: scrutinize