Muhammad Taqi Raza
commited on
Commit
·
4ada65a
1
Parent(s):
3c780d6
push optimized changes
Browse files
controlnet_pipeline.py
CHANGED
@@ -161,6 +161,7 @@ class ControlnetCogVideoXImageToVideoPCDPipeline(DiffusionPipeline, CogVideoXLor
|
|
161 |
tokenizer=tokenizer, text_encoder=text_encoder, vae=vae, transformer=transformer, controlnet=controlnet, scheduler=scheduler
|
162 |
)
|
163 |
self.vae_scale_factor_spatial = (
|
|
|
164 |
2 ** (len(self.vae.config.block_out_channels) - 1) if hasattr(self, "vae") and self.vae is not None else 8
|
165 |
)
|
166 |
self.vae_scale_factor_temporal = (
|
@@ -317,6 +318,8 @@ class ControlnetCogVideoXImageToVideoPCDPipeline(DiffusionPipeline, CogVideoXLor
|
|
317 |
)
|
318 |
|
319 |
num_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
|
|
|
|
|
320 |
shape = (
|
321 |
batch_size,
|
322 |
num_frames,
|
|
|
161 |
tokenizer=tokenizer, text_encoder=text_encoder, vae=vae, transformer=transformer, controlnet=controlnet, scheduler=scheduler
|
162 |
)
|
163 |
self.vae_scale_factor_spatial = (
|
164 |
+
# 6
|
165 |
2 ** (len(self.vae.config.block_out_channels) - 1) if hasattr(self, "vae") and self.vae is not None else 8
|
166 |
)
|
167 |
self.vae_scale_factor_temporal = (
|
|
|
318 |
)
|
319 |
|
320 |
num_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
|
321 |
+
# 36//
|
322 |
+
|
323 |
shape = (
|
324 |
batch_size,
|
325 |
num_frames,
|
inference/cli_demo_camera_i2v_pcd.py
CHANGED
@@ -254,7 +254,7 @@ def generate_video(
|
|
254 |
vae=vae,
|
255 |
controlnet=controlnet,
|
256 |
scheduler=scheduler,
|
257 |
-
)
|
258 |
|
259 |
# If you're using with lora, add this code
|
260 |
if lora_path:
|
@@ -275,6 +275,7 @@ def generate_video(
|
|
275 |
|
276 |
# pipe.to("cuda")
|
277 |
pipe = pipe.to(dtype=dtype)
|
|
|
278 |
# pipe.enable_sequential_cpu_offload()
|
279 |
if pipe_cpu_offload:
|
280 |
pipe.enable_model_cpu_offload()
|
@@ -427,7 +428,7 @@ if __name__ == "__main__":
|
|
427 |
parser.add_argument("--num_frames", type=int, default=49)
|
428 |
parser.add_argument("--start_camera_idx", type=int, default=0)
|
429 |
parser.add_argument("--end_camera_idx", type=int, default=1)
|
430 |
-
parser.add_argument("--controlnet_transformer_num_attn_heads", type=int, default=
|
431 |
parser.add_argument("--controlnet_transformer_attention_head_dim", type=int, default=64)
|
432 |
parser.add_argument("--controlnet_transformer_out_proj_dim_factor", type=int, default=64)
|
433 |
parser.add_argument("--controlnet_transformer_out_proj_dim_zero_init", action="store_true", default=True, help=("Init project zero."),
|
|
|
254 |
vae=vae,
|
255 |
controlnet=controlnet,
|
256 |
scheduler=scheduler,
|
257 |
+
)
|
258 |
|
259 |
# If you're using with lora, add this code
|
260 |
if lora_path:
|
|
|
275 |
|
276 |
# pipe.to("cuda")
|
277 |
pipe = pipe.to(dtype=dtype)
|
278 |
+
pipe = pipe.to("cuda")
|
279 |
# pipe.enable_sequential_cpu_offload()
|
280 |
if pipe_cpu_offload:
|
281 |
pipe.enable_model_cpu_offload()
|
|
|
428 |
parser.add_argument("--num_frames", type=int, default=49)
|
429 |
parser.add_argument("--start_camera_idx", type=int, default=0)
|
430 |
parser.add_argument("--end_camera_idx", type=int, default=1)
|
431 |
+
parser.add_argument("--controlnet_transformer_num_attn_heads", type=int, default=2)
|
432 |
parser.add_argument("--controlnet_transformer_attention_head_dim", type=int, default=64)
|
433 |
parser.add_argument("--controlnet_transformer_out_proj_dim_factor", type=int, default=64)
|
434 |
parser.add_argument("--controlnet_transformer_out_proj_dim_zero_init", action="store_true", default=True, help=("Init project zero."),
|