gogo03

可否将CONV3D,以及GRU量化一下？

1

只看到了卷积的量化。谢谢老大。

there is an error occurs while exec search.py. could anybody help me?

5

1.cp search.py ../search.py 2.in the directory. ..../fast-autoaugment exec the flowing command: python search.py -c confs/wresnet40x2_cifar10_b512.yaml got the error. 46%|████████████████████████████████████████████████████████████████████████▎ | 91/200 [30:06

I found that the initial parameters(kernel_initializer and kernel_regularizer) are more import than octave

6

def _create_normal_residual_block(inputs, ch, N,dropout): # Conv with skip connections x = inputs for i in range(N): # adjust channels if i == 0: skip = Conv2D(ch, 1,kernel_initializer='he_normal', kernel_regularizer=l2(0.0005))(x) skip =...

wide-resnet network structure is not standard or exits errors.

4

I have tried pytorch version wide-resnet, N=4,ch=10,It has 28 Conv2Ds, but yours only have 20 Conv2Ds. and the accuarcy of cifar100 is only up to about 75%.

resnet34 results can be shown: python train.py --seed=24 --scale=5 --optimizer=sgd --fast_auto_augment=True ......... [+] Training step: 62000/64000 Training epoch: 0/351 Elapsed time: 318.71min Learning rate: 0.0011729701340847298 Acc@1 : 88.281% Acc@5 :...

train deeplabv3_mobile时出错。

1

ValueError: Shape must be rank 3 but is rank 4 for '{{node deeplabv3plus/concatenate_1/concat}} = ConcatV2[N=2, T=DT_FLOAT, Tidx=DT_INT32](deeplabv3plus/lambda_3/resize/Squeeze, deeplabv3plus/activation_8/Relu, deeplabv3plus/concatenate_1/concat/axis)' with input shapes: [104,104,2], [?,104,104,48], []. 环境keras-gpu 2.4.3

中文图片输出的jsonl是一些乱码？

11

如题，这个是啥原因呢？{"id": "f02aca94f121a4585e8a5c7d91c9ca8a664640bf", "text": "\u9ad82020\u7ea72022-2023\u5b66\u5e74\u5ea6\u7b2c\u4e00\u5b66\u671f\u67