| from functools import partial |
|
|
| import torch |
| from benchmarking_utils import BenchmarkMixin, BenchmarkScenario, model_init_fn |
|
|
| from diffusers import LTXVideoTransformer3DModel |
| from diffusers.utils.testing_utils import torch_device |
|
|
|
|
| CKPT_ID = "Lightricks/LTX-Video-0.9.7-dev" |
| RESULT_FILENAME = "ltx.csv" |
|
|
|
|
| def get_input_dict(**device_dtype_kwargs): |
| |
| |
| hidden_states = torch.randn(1, 7392, 128, **device_dtype_kwargs) |
| encoder_hidden_states = torch.randn(1, 256, 4096, **device_dtype_kwargs) |
| encoder_attention_mask = torch.ones(1, 256, **device_dtype_kwargs) |
| timestep = torch.tensor([1.0], **device_dtype_kwargs) |
| video_coords = torch.randn(1, 3, 7392, **device_dtype_kwargs) |
|
|
| return { |
| "hidden_states": hidden_states, |
| "encoder_hidden_states": encoder_hidden_states, |
| "encoder_attention_mask": encoder_attention_mask, |
| "timestep": timestep, |
| "video_coords": video_coords, |
| } |
|
|
|
|
| if __name__ == "__main__": |
| scenarios = [ |
| BenchmarkScenario( |
| name=f"{CKPT_ID}-bf16", |
| model_cls=LTXVideoTransformer3DModel, |
| model_init_kwargs={ |
| "pretrained_model_name_or_path": CKPT_ID, |
| "torch_dtype": torch.bfloat16, |
| "subfolder": "transformer", |
| }, |
| get_model_input_dict=partial(get_input_dict, device=torch_device, dtype=torch.bfloat16), |
| model_init_fn=model_init_fn, |
| compile_kwargs={"fullgraph": True}, |
| ), |
| BenchmarkScenario( |
| name=f"{CKPT_ID}-layerwise-upcasting", |
| model_cls=LTXVideoTransformer3DModel, |
| model_init_kwargs={ |
| "pretrained_model_name_or_path": CKPT_ID, |
| "torch_dtype": torch.bfloat16, |
| "subfolder": "transformer", |
| }, |
| get_model_input_dict=partial(get_input_dict, device=torch_device, dtype=torch.bfloat16), |
| model_init_fn=partial(model_init_fn, layerwise_upcasting=True), |
| ), |
| BenchmarkScenario( |
| name=f"{CKPT_ID}-group-offload-leaf", |
| model_cls=LTXVideoTransformer3DModel, |
| model_init_kwargs={ |
| "pretrained_model_name_or_path": CKPT_ID, |
| "torch_dtype": torch.bfloat16, |
| "subfolder": "transformer", |
| }, |
| get_model_input_dict=partial(get_input_dict, device=torch_device, dtype=torch.bfloat16), |
| model_init_fn=partial( |
| model_init_fn, |
| group_offload_kwargs={ |
| "onload_device": torch_device, |
| "offload_device": torch.device("cpu"), |
| "offload_type": "leaf_level", |
| "use_stream": True, |
| "non_blocking": True, |
| }, |
| ), |
| ), |
| ] |
|
|
| runner = BenchmarkMixin() |
| runner.run_bencmarks_and_collate(scenarios, filename=RESULT_FILENAME) |
|
|