alibaba-pai/Wan2.1-Fun-1.3B-InP

1import os2import sys3import time4 5import torch6 7current_file_path = os.path.abspath(__file__)8project_roots = [os.path.dirname(current_file_path), os.path.dirname(os.path.dirname(current_file_path)), os.path.dirname(os.path.dirname(os.path.dirname(current_file_path)))]9for project_root in project_roots:10    sys.path.insert(0, project_root) if project_root not in sys.path else None11 12from cogvideox.api.api import (infer_forward_api,13                               update_diffusion_transformer_api,14                               update_edition_api)15from cogvideox.ui.controller import flow_scheduler_dict16from cogvideox.ui.wan_fun_ui import ui, ui_eas, ui_modelscope17 18if __name__ == "__main__":19    # Choose the ui mode  20    ui_mode = "eas"21    22    # GPU memory mode, which can be choosen in [model_cpu_offload, model_cpu_offload_and_qfloat8, sequential_cpu_offload].23    # model_cpu_offload means that the entire model will be moved to the CPU after use, which can save some GPU memory.24    # 25    # model_cpu_offload_and_qfloat8 indicates that the entire model will be moved to the CPU after use, 26    # and the transformer model has been quantized to float8, which can save more GPU memory. 27    # 28    # sequential_cpu_offload means that each layer of the model will be moved to the CPU after use, 29    # resulting in slower speeds but saving a large amount of GPU memory.30    GPU_memory_mode = "model_cpu_offload"31    # Use torch.float16 if GPU does not support torch.bfloat1632    # ome graphics cards, such as v100, 2080ti, do not support torch.bfloat1633    weight_dtype = torch.bfloat1634    # Config path35    config_path = "config/wan2.1/wan_civitai.yaml"36 37    # Server ip38    server_name = "0.0.0.0"39    server_port = 786040 41    # Params below is used when ui_mode = "modelscope"42    model_name = "models/Diffusion_Transformer/Wan2.1-Fun-1.3B-InP"43    # "Inpaint" or "Control"44    model_type = "Inpaint"45    # Save dir of this model46    savedir_sample = "samples"47 48    if ui_mode == "modelscope":49        demo, controller = ui_modelscope(model_name, model_type, savedir_sample, GPU_memory_mode, flow_scheduler_dict, weight_dtype, config_path)50    elif ui_mode == "eas":51        demo, controller = ui_eas(model_name, flow_scheduler_dict, savedir_sample, config_path)52    else:53        demo, controller = ui(GPU_memory_mode, flow_scheduler_dict, weight_dtype, config_path)54 55    # launch gradio56    app, _, _ = demo.queue(status_update_rate=1).launch(57        server_name=server_name,58        server_port=server_port,59        prevent_thread_lock=True60    )61    62    # launch api63    infer_forward_api(None, app, controller)64    update_diffusion_transformer_api(None, app, controller)65    update_edition_api(None, app, controller)66    67    # not close the python68    while True:69        time.sleep(5)