Upload src/pipeline.py with huggingface_hub
Browse files- src/pipeline.py +4 -5
src/pipeline.py
CHANGED
|
@@ -47,11 +47,10 @@ def load_pipeline() -> Pipeline:
|
|
| 47 |
local_files_only=True,
|
| 48 |
torch_dtype=torch.bfloat16,
|
| 49 |
).to("cuda")
|
| 50 |
-
|
| 51 |
-
pipeline.transformer.to(memory_format=torch.channels_last)
|
| 52 |
-
pipeline.vae.to(memory_format=torch.channels_last)
|
| 53 |
-
|
| 54 |
-
pipeline.text_encoder = torch.compile(pipeline.text_encoder, fullgraph=True, mode="reduce-overhead")
|
| 55 |
quantize_(pipeline.vae, int8_weight_only())
|
| 56 |
pipeline.vae = torch.compile(pipeline.vae, fullgraph=True, backend="tensorrt")
|
| 57 |
|
|
|
|
| 47 |
local_files_only=True,
|
| 48 |
torch_dtype=torch.bfloat16,
|
| 49 |
).to("cuda")
|
| 50 |
+
pipeline.to(memory_format=torch.channels_last)
|
| 51 |
+
# pipeline.transformer.to(memory_format=torch.channels_last)
|
| 52 |
+
# pipeline.vae.to(memory_format=torch.channels_last)
|
| 53 |
+
|
|
|
|
| 54 |
quantize_(pipeline.vae, int8_weight_only())
|
| 55 |
pipeline.vae = torch.compile(pipeline.vae, fullgraph=True, backend="tensorrt")
|
| 56 |
|