问题描述
我有一个自定义构建的张量流图,该图实现了我自己实现的MobileNetV2-SSDLite.在PC上运行正常.
I have a custom built tensorflow graph implementing MobileNetV2-SSDLite which I implemented myself. It is working fine on the PC.
但是,当我将模型转换为TFLite(全部浮动,没有量化)时,模型权重会急剧变化.
However, when I convert the model to TFLite (all float, no quantization), the model weights are changed drastically.
举个例子,最初是一个过滤器- 0.13172674179077148, 2.3185202252437188e-32, -0.003990101162344217
To give an example, a filter which was initially - 0.13172674179077148, 2.3185202252437188e-32, -0.003990101162344217
成为-4.165565013885498,-2.3981268405914307,-1.1919032335281372
becomes-4.165565013885498,-2.3981268405914307,-1.1919032335281372
较大的权重值完全摆脱了我在设备上的推断.需要帮忙! :(
The large weight values are completely throwing off my on-device inferences. Need help! :(
推荐答案
您使用什么命令将其转换为tflite?例如,您使用的是toco,如果使用的是什么参数?虽然我没有查看过滤器,但这是我的默认说明用于微调MobileNetV2-SSD和SSDLite图,并且模型运行良好.
What command are you using to convert to tflite? For instance are you using toco, and if so what parameters are you using? While I haven't been looking at the filters, here are my default instructions for finetuning a MobileNetV2-SSD and SSDLite graphs and the model has been performing well.
这篇关于TFLite Conversion更改模型权重的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持!