Veiiling

Distillation in fp16 mode

1

when student and teacher network traing in fp16, if align_module is none, training is work well. but if i set align_module, the align_module's weight is fp32, but th student's outputs...

[Feature]yolox-kd

1

about yolox, it has some suggestions about kd? I have experimented with some algorithms(reviewkd,dkd), but the accuracy has hardly improved.

In optimze func: ``` def optimize(model, passes=None, fixed_point=False): # type: (ModelProto, Optional[Sequence[Text]], bool) -> ModelProto """Apply the optimization on the serialized ModelProto. Arguments: input (ModelProto): model names (list of string):...

Have any useful kd algo for yolox?

量化损失

1

对于图像识别任务，模型进行uint8量化后，会带来严重的精度损失吗？尤其在使用relu时？

can we use fit_scan.py to register our sace data to flame?

1

Hello. Thank you for your great work first of all. I want to fit my 3D scan data to the flame model. I have marked 51 feature points and scan...

if we trainging in lora mode, how to merge lora adapter