Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -46,17 +46,9 @@ class SamplingOptions:
|
|
| 46 |
guidance: float
|
| 47 |
seed: int | None
|
| 48 |
|
| 49 |
-
@torch.inference_mode()
|
| 50 |
-
def encode(init_image, torch_device, ae):
|
| 51 |
-
init_image = torch.from_numpy(init_image).permute(2, 0, 1).float() / 127.5 - 1
|
| 52 |
-
init_image = init_image.unsqueeze(0)
|
| 53 |
-
init_image = init_image.to(torch_device)
|
| 54 |
-
ae = ae.cuda()
|
| 55 |
-
with torch.no_grad():
|
| 56 |
-
init_image = ae.encode(init_image.to()).to(torch.bfloat16)
|
| 57 |
-
return init_image
|
| 58 |
|
| 59 |
|
|
|
|
| 60 |
class FluxEditor:
|
| 61 |
def __init__(self, args):
|
| 62 |
self.args = args
|
|
@@ -87,6 +79,16 @@ class FluxEditor:
|
|
| 87 |
self.model.cpu()
|
| 88 |
torch.cuda.empty_cache()
|
| 89 |
self.ae.encoder.to(self.device)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
|
| 91 |
@torch.inference_mode()
|
| 92 |
def edit(self, init_image, source_prompt, target_prompt, num_steps, inject_step, guidance, seed):
|
|
@@ -103,7 +105,7 @@ class FluxEditor:
|
|
| 103 |
init_image = init_image[:new_h, :new_w, :]
|
| 104 |
|
| 105 |
width, height = init_image.shape[0], init_image.shape[1]
|
| 106 |
-
init_image = encode(init_image, self.device, self.ae)
|
| 107 |
|
| 108 |
print(init_image.shape)
|
| 109 |
|
|
|
|
| 46 |
guidance: float
|
| 47 |
seed: int | None
|
| 48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
|
| 51 |
+
@spaces.GPU(duration=30)
|
| 52 |
class FluxEditor:
|
| 53 |
def __init__(self, args):
|
| 54 |
self.args = args
|
|
|
|
| 79 |
self.model.cpu()
|
| 80 |
torch.cuda.empty_cache()
|
| 81 |
self.ae.encoder.to(self.device)
|
| 82 |
+
|
| 83 |
+
@torch.inference_mode()
|
| 84 |
+
def encode(init_image, torch_device, ae):
|
| 85 |
+
init_image = torch.from_numpy(init_image).permute(2, 0, 1).float() / 127.5 - 1
|
| 86 |
+
init_image = init_image.unsqueeze(0)
|
| 87 |
+
init_image = init_image.to(torch_device)
|
| 88 |
+
ae = ae.cuda()
|
| 89 |
+
with torch.no_grad():
|
| 90 |
+
init_image = ae.encode(init_image.to()).to(torch.bfloat16)
|
| 91 |
+
return init_image
|
| 92 |
|
| 93 |
@torch.inference_mode()
|
| 94 |
def edit(self, init_image, source_prompt, target_prompt, num_steps, inject_step, guidance, seed):
|
|
|
|
| 105 |
init_image = init_image[:new_h, :new_w, :]
|
| 106 |
|
| 107 |
width, height = init_image.shape[0], init_image.shape[1]
|
| 108 |
+
init_image = self.encode(init_image, self.device, self.ae)
|
| 109 |
|
| 110 |
print(init_image.shape)
|
| 111 |
|