I can't find the relevant data, can you teach me the data needed to download the code. Thank you. Best wishes

[Wang-Dlut]

I can't find the relevant data, can you teach me the data needed to download the code.
Thank you. Best wishes

[markub3327]

@Wang-Dlut The data used for training the model was obtained by this Julia script: https://github.com/markub3327/NASA-POWER-BOT. Use can adjust the GPS locations of areas in config/locations.yml. The size of area is defined by arguments width and height, same as the date range of measurements and number of threads to run downloads in parallel fashion.

julia --threads 8 main.jl --start 2010 --end 2015 --width 5 --height 5

Thanks

[Wang-Dlut]

Thank you very much for your strong support！
I just don't quite understand how to operate this website“NASA-POWER-BOT”, hahaha.
I did download the". csv "file to debug the code you shared, and I have sent the code to your email"[email protected][email protected]"
However, the basis for setting the values of 302 and 3 in this code is not well understood."daily_model.build([(None, window_size, num_of_patches, 302), (None, window_size, num_of_patches, 3)])"
Could you please help answer this? Thank you
Wishing you a happy life
Looking forward to your letter

[markub3327]

@Wang-Dlut

daily_model.build([(None, window_size, num_of_patches, 302), (None, window_size, num_of_patches, 3)])

This is needed to initialise the model before load the weights. The model has a 2 input layers one has 302 features and another one is 3 features.

The first one is input to Encoder block and represents weather (pressure, temperature, wind speed, wind direction) and solar irradiance measurements from the past in the region manner.
Regional data contains weather information too.

The second input is to Decoder block and represents the only solar irradiance, and time in two components sine and cosine embedding and its in the point measurement. The Decoder predicts only solar irradiance in the next time and the this input cannot contain the weather information because Decoder must be predict the weather too not only solar irradiance.
Point data does not contain weather information only the location information, time information and irradiance.

[markub3327]

12 * 5 * 5 = 300 + 2 for the sine and cosine time embedding:
df_regional_daily[i][!, "Irradiance$(j)"] = Float32[]
df_regional_daily[i][!, "Temp$(j)"] = Float32[]
df_regional_daily[i][!, "TempMin$(j)"] = Float32[]
df_regional_daily[i][!, "TempMax$(j)"] = Float32[]
df_regional_daily[i][!, "Humidity$(j)"] = Float32[]
df_regional_daily[i][!, "WindX$(j)"] = Float32[]
df_regional_daily[i][!, "WindY$(j)"] = Float32[]
df_regional_daily[i][!, "WindXMin$(j)"] = Float32[]
df_regional_daily[i][!, "WindYMin$(j)"] = Float32[]
df_regional_daily[i][!, "WindXMax$(j)"] = Float32[]
df_regional_daily[i][!, "WindYMax$(j)"] = Float32[]
df_regional_daily[i][!, "Pressure$(j)"] = Float32[]

[Wang-Dlut]

wow! This is amazing!!! HAAAAAA! Thank u very much! Thank you for your strong support！ Best wishes to you！！！

[Wang-Dlut]

I really like and love your this code and ideas, but I still haven't reproduced them yet hahaha。This makes me feel frustrated。。。。。。。。。。。。。。。。。。。。。。

[wujiayi567]

Thank you very much for your strong support！ I just don't quite understand how to operate this website“NASA-POWER-BOT”, hahaha. I did download the". csv "file to debug the code you shared, and I have sent the code to your email"[email protected][email protected]" However, the basis for setting the values of 302 and 3 in this code is not well understood."daily_model.build([(None, window_size, num_of_patches, 302), (None, window_size, num_of_patches, 3)])" Could you please help answer this? Thank you Wishing you a happy life Looking forward to your letter
I encountered a similar problem to yours. When running this line, an error was reported: ValueError: Cannot take the length of shape with unknown rank. Have you solved it?Thank you. Best wishes

[wujiayi567]

Dear author,
I hope you're well. When running "daily_model.build([(None, window_size, num_of_patches, 302), (None, window_size, num_of_patches, 3)])", I got "ValueError: Cannot take the length of shape with unknown rank". I'm puzzled. Could you help me understand the cause and suggest a solution?
I am currently using the data downloaded from the website following the command "julia --threads 8 main.jl --start 2010 --end 2015 --width 5 --height 5" (https://github.com/markub3327/NASA-POWER-BOT).
For the data part in your code, where it is currently written as "df_X = pd.read_csv(r"E:\dni\NASA-POWER-BOT-master\dataset\X_all_daily.csv")" and "df_y_daily = pd.read_csv(r"E:\dni\NASA-POWER-BOT-master\dataset\y_all_daily.csv")", I was wondering if you could provide some guidance on whether any modifications might be needed to better adapt to the data downloaded from the aforementioned website. I'm not entirely sure if the current way of reading the data is the most appropriate for this particular data source.
Thank you.
Best regards

[Wang-Dlut]

TypeError: Unexpected keyword argument passed to optimizer: exclude_from_weight_decay

进程已结束，退出代码为 1

[markub3327]

https://www.tensorflow.org/addons/api_docs/python/tfa/optimizers/AdamW