-
同学你好,运行main_full.py,这句出错,from progress.bar import Bar。出错原因ImportError: No module named progress.bar。。我在网上找了只有progress库,没有progress.bar。我想请问一下你这里怎么解决。谢谢~~。。另外,我是友校的,不知道方不方便加你QQ,我想问问你关于DoReFa-Net的一些问题,我的…
-
请问一下作者,原文中的对gradient的量化是不是没有应用在code中?
-
你好,首先很感谢这份优秀的工作! 我好像在代码中没有看到关于dorefa 的梯度量化的部分,请问是否缺失呢?
-
**Describe the issue**:
I am using level pruning and dorefa quantization. I tried to save the model, using export_model/torch save dict (both seperately) and then load the state_dict, but getting inc…
-
Hello, I have a couple of questions regarding quantizer options for Larq and LCE.
I am designing a BNN using the DoReFa quantizer, however, I noticed a very high number of estimated MACs and Ops wh…
-
Hi~首先非常感谢你的工作。我最近将micronet的dorefa量化部分适用到我的工程中(检测模型),发现几个问题,希望讨论一下:
1. load预训练的浮点模型,使用dorefa的量化inference模式,检测精度直接降为0,求问为什么精度损失会这么大?
2. 采用dorefa的qat模式,正常训练后精度接近于原float模式,但是load预训练模型与不load的结果并没有区别,这是为…
-
# Reference
# Brief
## Pruning
Name | Description
-- | --
AGP Pruner
Slim Pruner
FPGM Pruner
NetAdapt Pruner
SimulatedAnnealing Pruner
ADMM Pruner
AutoCompress Pruner
Prune-EA
#…
-
https://github.com/calipos/dorefa-caffe/blob/42bf6be8a188bb115493bf78cdcfa6a77917555b/src/caffe/util/math_functions.cu#L96
1. 这句话为什么不是0和1之间
2. 源代码没有tanh
我的微信15711325055,方便交流哈
-
Hello, when I loaded the model into it after training it as follows, I got an error.
In this case, how should I deal with it?
training command
` $ python compress_classifier.py --arch simplenet…
-
### Describe the bug
Hi,
I tried to use DoReFa quantiser to train a simple model for CIFAR10, but training failed to converge:
```
10000/1 - 1s 96us/sample - loss: 2.3026 - accuracy: 0.1000
Test …