[Speedster] 优化模型耗时 10 小时,还没有结束
作者: wanglongwork创建于 2023年9月14日更新于 2023年9月14日
class SmilingEncode(nn.Module):
def __init__(self, model):
super().__init__()
self.model = model
def forward(self, xT, cond2):
img = self.model.render_speedster(xT, cond2)
return img
xT_model = SmilingXT(model).to(device).eval()
input_data = [((torch.randn(1, 3, 256, 256), torch.randn(1, 512)), torch.tensor([0])) for _ in range(100)]
xT_optimized_model = optimized_model = optimize_model(
xT_model, input_data=input_data, optimization_time="unconstrained", device=device)
save_model(xT_optimized_model, "speedster/xT_optimized_model")2023-09-13 18:19:45 | INFO | Running Speedster on GPU:1 2023-09-13 18:23:05 | INFO | Benchmark performance of original model 2023-09-13 18:27:29 | INFO | Original model latency: 2.0304924035072327 sec/iter ============= Diagnostic Run torch.onnx.export version 2.0.1+cu117 ============= verbose: False, log level: Level.ERROR ======================= 0 NONE 0 NOTE 0 WARNING 0 ERROR ========================
内容来源: nebuly-ai/optimate