About issus

Hello: I would like to ask, if I want to convert two tflites into full integers (including input and output with int8), then I know the calibration data of the...

About one model

1

Hi, i found the model convert to tflite have two files,can it was convered to a tflite file?

Hi: I would like to ask what is the appropriate 'count' for feature generation if I use 120 hours of human voice and 80 hours of noise as training data...

About memory

2

Hi, i got a output like this: st->rnn.model->vad_gru_state = 1530144 st->rnn.model->noise_gru_state = 1530256 st->rnn.model->denoise_gru_size = 1530464 Can anyone tell me what this means？ Thanks

使用#define NNOM_USING_CMSIS_NN 的错误。

2

nothing

关于使用DW_Conv2D与Conv2D的移植后使用耗时的问题

1

您好，我在移植后发现在权重很少的情况下，DW_Conv2D的耗时很短但是一旦接上Conv2D后耗时就急剧上升，可能单单DW_Conv2D时是几百微秒，后面接上Conv2D就到了4+毫秒了，权重差异不大都很少，为啥它们会有这么明显的区别？ layer[0] = Input(shape(1, 129, 2), nnom_input_data); layer[1] = model.hook(DW_Conv2D(1, kernel(2, 5), stride(1, 1), dilation(1, 1), PADDING_SAME, &depthwise_conv2d_w, &depthwise_conv2d_b), layer[0]); layer[2] = model.hook(Conv2D(32, kernel(1, 1), stride(1, 1), dilation(1, 1), PADDING_SAME,...

关于输出维度的问题

4

您好： 我想请问下假设我的输出是（1,256,2）也就是512个数。那么输出是什么样的，我看输出好像直接就是512个数了，那么它对于(1,256,2)的组织形式是啥样的，怎么区分或者找出这两个256呢？

about reshape

3

Hi: I want to ask what the nnom can support tf.reshape ?

关于CNN运行速度的问题

1

您好：我在将模型到芯片上运行后测试了下时间，输入为512个数据，cnn大约4k个权重系数 运行时间大约75ms，主频192M。不知道这个时间是否是正常的，如果想要极大缩短这个时间，是否在NNOM有什么建议呢？

当前框架能否支持slice

2

我想请问下如果有“tf_op_layer_strided_slice” 当前框架能否支持