你好,你用的sdk1.7已经是最新版本了。方便把pt模型文件发给我吗(jiapeiyuan@cambricon.com),我们复现一下,看看问题在哪展开
总结:
1.要实现640*640的模型,还需将 utils/datasets.py中的letterbox接口中的auto修改为False
2.有些老版本的docker在适配自己训练的模型时会有bug(batchsize与corenum仅支持1),比如yellow.hub.cambricon.com/pytorch/pytorch:0.15.0-ubuntu16.04 需要安装cnplugin补丁包后可解决排查backbone的输出有没有差异,是下图这个输出么,下图输出一致corenum 1/4一致
def get_boxes(prediction, batch_size=1, img_size=640): #Returns detections with shape: # (x1, y1, x2, y2, object_conf, class_score, class_pred) reshape_value = torch.reshape(prediction, (-1, 1)) num_boxes_final = reshape_value[0].item() print('num_boxes_final: ',num_boxes_final) all_list = [[] for _ in range(batch_size)] for i in range(int(num_boxes_final)): ...
并且上述代码 print('num_boxes_final: ',num_boxes_final) 框的数量是一致的
请登录后评论