OpenCastKit icon indicating copy to clipboard operation
OpenCastKit copied to clipboard

The open-source solutions of FourCastNet and GraphCast

Results 24 OpenCastKit issues
Sort by recently updated
recently updated
newest added

In `data_factory/download_era5.py`, it downloads 6-hour total precipitation by specifying 6 hour intevals. But after reviewing the data and reading the [ERA5 cdsapi demo](https://confluence.ecmwf.int/display/CKB/ERA5%3A+How+to+calculate+daily+total+precipitation), It seems that the api only downloads...

请问作者可否说明一下FourCastNet+额外4个通道的含义 @VachelHU

$ python train_fourcastnet.py --pretrain-epochs 10 --fintune-epochs 4 --batch-size 1 报错: 非集群环境 非集群环境 Traceback (most recent call last): File "train_fourcastnet.py", line 207, in hfai.multiprocessing.spawn(main, args=( File "/home/pineapple/mambaforge/envs/OpenCast/lib/python3.8/site-packages/hfai/multiprocessing/spawn.py", line 66, in spawn...

在train_graphcast.py文件的train方法中,有下面两行代码。 ```python param_groups = timm.optim.optim_factory.add_weight_decay(model, args.weight_decay) optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95)) ``` 我查看`timm`的文档,和`timm.optim.optim_factory`的源代码都没有找到`add_weight_decay`方法。代码中`args.weight_decay`参数是一个浮点数`0.05`。 请问这里的作用是不是等效于下面代码?因为我不确定第一行代码中传入`model`是否有其他效果? ```python optimizer = torch.optim.AdamW(lr=args.lr, weight_decay=args.weight_decay, betas=(0.9, 0.95)) ```

Do you plan on releasing model weights / training script for FourCastNet+?

首先感谢作者分享了模型,我在学习运行的过程中遇到了一些问题难以解决,希望能获得帮助。我已经能够成功运行FourcastNet进行推理预测,但是运行Graphcast时遇到一个问题,FourcastNet输入为20个通道,Graphcast输入需要最后一维为49,我能知道x0,x1都为22(20个变量加2个时间特征),y[ :, :, -2]为2,加起来为46。请问我遗漏了什么,是否可以指点一下,非常感谢! ![image](https://github.com/HFAiLab/OpenCastKit/assets/56792866/cef7e3cb-c85e-4878-9a5e-c1d56ed148d7)

请问作者,在您已发布的大模型pt文件中,fourcastnet-plus对比fourcastnet增加的4维变量具体是哪些呢? ![image](https://github.com/HFAiLab/OpenCastKit/assets/35858411/8c7f1f6e-f88c-4938-a771-40780adf31be) ![image](https://github.com/HFAiLab/OpenCastKit/assets/35858411/e2a358e9-a50a-41ee-bcd0-33ed39da4ba6)

感谢博主的模型贡献。我运行时发现pt文件权重和git代码不匹配,删除一些线性层,并将某些DIM从512更改为256,以使其加载。使用加载好的模型推理,第2步的输入由第1步的输入和第1步的预报拼接得到49维,结果在第1步之后预报质量严重下降,请问博主是什么问题呢?

hi, thank for ur share , can u provide the scaler.pkl? i can not download the whole data due to storage limit