luyishisi / shisi.eth-in-web3 Goto Github PK
View Code? Open in Web Editor NEW十四君公众号文章的相关代码资源开源
License: BSD 2-Clause "Simplified" License
十四君公众号文章的相关代码资源开源
License: BSD 2-Clause "Simplified" License
learning keras/tensorflow, thinking that your project might be a good starting point for me to learn. However I ran into several issues and I guess maybe your codes were implemented in python 2.7? Can you upgrade it to be compatible to python 3.5?
My cellphone is 13918544873, we may chat.
我在最后的全联通层用的是你注释掉的softmax函数
然后每训练20次输出一次准确率
在训练的时候发现,每训练一次loss就增长很多
但第一次输出acc为0.01
第二次为0.0225
按理说loss增加acc应该会减少才对呀
为什么acc不减反增呢
训练代码在
https://github.com/whousemyDaLaBengBa/DeepNeuralNetwork
输出数据在readme中
以注明出处,如侵权请联系删除
with slim.arg_scope([slim.conv2d, slim.fully_connected],
activation_fn=tf.nn.relu,
biases_initializer=tf.random_normal_initializer,
weights_initializer=tf.random_normal_initializer,
):
conv1 = slim.conv2d(x, 32, [3, 3], 1)
pool1 = slim.max_pool2d(conv1, [2, 2], 2, padding='SAME')
drop1 = slim.dropout(pool1, keep_prob=keep_prob)
conv2 = slim.conv2d(drop1, 64, [3, 3], 1)
pool2 = slim.max_pool2d(conv2, [2, 2], 2, padding='SAME')
drop2 = slim.dropout(pool2, keep_prob=keep_prob)
conv3 = slim.conv2d(drop2, 64, [3, 3], 1)
pool3 = slim.max_pool2d(conv3, [2, 2], 2, padding='SAME')
drop3 = slim.dropout(pool3, keep_prob=keep_prob)
flatten = slim.flatten(drop3)
dense1 = slim.fully_connected(flatten, 1024)
drop4 = slim.dropout(dense1, keep_prob=keep_prob)
out = slim.fully_connected(drop4, MAX_CAPTCHA*CHAR_SET_LEN, activation_fn=None)
return out
贴上代码,其它都一样,但是训练时初始的loss超级大,@luyishisi能帮忙分析一下吗?
能不能保存成.pb并把权重固化进去?
谢谢!
你好,大概训练几条记录才能达到50%呢?
您好!
请问
达到50%成功率需要2000个批次,总计20w张图片。
达到70%成功率需要4000个批次,总计40w张图片。
达到94%成功率需要40000个批次,总计400w张图片。
达到98%成功率需要100000个批次,总计1000w张图片。
这些训练数据的验证码是只包含数字还是包含了数字和字母?
我用您的代码训练,包含数字和字母,跑了200000次,准确率还是0.085
shisi.eth 你好,
看了你的文章,【解读】以太坊上海升级即将激活的四个EIP
在看eip-3651时对比官网:https://eips.ethereum.org/EIPS/eip-3651
我觉得:
这个结论是不是有问题啊?
我看原文中的描述为:这种mismatch,(也就是在coinbase地址冷的时候,access coinbase会消耗更多的gas) 会激励出来ETH之外的其他ERC-20代币进行支付。
如果升级该EIP,就不会导致这种激励了吧?
博客原来共享的98模型还能再共享下吗
你好,一开始我把softmax作为最后的分类层,结果在调learning_rate的时候发现loss要么一开始就在某个值上下浮动,要么就会一直增大,后来用你的sigmoid方式分类,loss到最后会在0.0834上下变化,但是accuracy确实一直很小,而且会有时候增大有时候减小,这大概是什么原因呢?
相对你的代码,我只是添加了scope,然后weights初始化的时候用的是truncated_normal_initializer(stddev=0.01),不知道这样改会有什么影响吗?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.