今天试了下 Chat GPT, 惊了, 可以提高生产力, 取代大部分 stack overflow 功能

起因是有个部门分享的 data 是 power BI 数据源 API,我想把这些数据用 python 读取到 pandas 里面,昨天花了很久研究 M query, 不会转换双引号:(, 拿不到 token,刚刚试了下 ChatGPT 给的 code,一次成功.
太厉害了, 以后不用去 stack overflow 了

我: can you convert power bi m code to python?

ChatGPT: Yes, it is possible to convert Power BI M code to Python. However, the conversion process may not be straightforward and will depend on the specific M code that you are trying to convert. Power BI M code is a proprietary language used in the Power BI platform, while Python is a general-purpose programming language that is widely used in data analysis and scientific computing.

If you have specific M code that you would like to convert to Python, I would recommend that you provide the code and I can try to assist you with the conversion. It would also be helpful to know the reason why you want to convert the M code to Python and what you plan to do with the resulting Python code.

我:
let body1 ="{
""username"": " & username &",
""password"": "& password &"

}",

getToken =
Web.Contents(server & tokenAddress, [Content=Text.ToBinary(body1)]),

token1 = "Bearer " & Text.FromBinary(getToken),

ChatGPT: This Power BI M code is defining several variables and making a web request to get a token. The body1 variable is defining the body of the request as a string containing the username and password values. The getToken variable is making a web request to the server & tokenAddress URL with the body1 content. Finally, the token1 variable is constructing a string by concatenating the Bearer string with the response from the getToken web request.

In Python, you can use the requests library to make web requests and parse the response. Here is an example of how the above Power BI M code could be converted to Python:

Copy code
import requests

username = "your_username"
password = "your_password"
server = "http://your_server.com"
tokenAddress = "/api/get_token"

body1 = {
"username": username,
"password": password
}

getToken = requests.post(server + tokenAddress, json=body1)
token1 = "Bearer " + getToken.text
Note that this is just an example and may not work exactly as shown if the actual Power BI M code is different from what you provided. You may need to adjust the code depending on the details of your specific use case.

aigonna

2022-12-08 12:43:41 +08:00

下面是一个使用 PyTorch 实现知识蒸馏的简化代码例子：
# 首先，我们需要导入必要的库
import torch
import torch.nn as nn

# 定义数据集
X = ... # 输入数据
y = ... # 标签

# 定义大型预训练模型
pretrained_model = ...

# 使用大型预训练模型对数据集进行预测
# 这些预测将作为辅助标签
soft_labels = pretrained_model(X)

# 定义较小的模型
small_model = ...

# 定义损失函数
# 这里，我们将使用交叉熵损失函数
# 并将辅助标签作为目标
loss_fn = nn.CrossEntropyLoss(reduction="mean")
loss = loss_fn(y, soft_labels)

# 使用随机梯度下降优化器优化模型参数
optimizer = torch.optim.SGD(small_model.parameters(), lr=0.001)
optimizer.zero_grad()
loss.backward()
optimizer.step()

NLP 也凑合能用

shinession

2022-12-08 13:26:25 +08:00

@amwyyyy 假设鸡有 x 只，兔子有 y 只，鸡和兔子的腿总数量可以用如下方程来表示：

2x + 4y = 38

鸡和兔子的头总数量可以用如下方程来表示：

x + y = 14

现在我们已经有了两个方程，就可以求出 x 和 y 的值。首先，我们可以把第二个方程中的 y 带入第一个方程中，得到：

2x + 4(14 - x) = 38

化简得到：

2x + 56 - 4x = 38

再次化简得到：

-2x = -18

因此，x = 9 。由于 x + y = 14 ，所以 y = 14 - 9 = 5 。因此，鸡有 9 只，兔子有 5 只。

总结一下，当鸡和兔子的头总数为 14 个，腿总数为 38 条时，鸡有 9 只，兔子有 5 只。

试了一下, 没问题