Pcyslist

i try to apply byte to yolov5 , but i find before pass the results of detections to byte , i need to use the postprocess function , it cost...

could you please explain how the edges_datasets prepared?

1

Thanks for your great jobs. I trained with my customed datasets, but I did't understand the waht shuould I do with edges_datasets. I read your paper carefully, could you please...

why can't use non_max_suppression replace of postprocess ?

5

outputs = self.model(img, False, False) # if self.decoder is not None: # outputs = self.decoder(outputs, dtype=outputs.type()) print('before nms:', outputs.size()) outputs = postprocess(outputs, self.num_classes, self.confthre, self.nmsthre) # outputs = non_max_suppression(outputs, 0.1,...

 Hi yiyang ! In the code you implemented, I found that the one-shot training epoch of NasBench-201 is [600](https://github.com/aoiang/few-shot-NAS/blob/27b9735226f3d0a45dbe7bc3cce7c15db9b544f9/Few-Shot_NasBench201/configs/nas-benchmark/algos/ONE-SHOT-SUPERNET.config#L5), and the few-shot training epoch is [300](https://github.com/aoiang/few-shot-NAS/blob/27b9735226f3d0a45dbe7bc3cce7c15db9b544f9/Few-Shot_NasBench201/configs/nas-benchmark/algos/FEW-SHOT-SUPERNET.config#L5). According to such...

咋样能保证相同的输入，多次生成的内容完全一致呢？

5

我调整了以下参数 tem=0 top_p=1 seed=123 仍然无法实现相同输入的情况下，多次生成内容的一致性。 哪位大佬有办法呢。。。