Closed AndrewYangnb closed 6 months ago
You need to pass a graphcast.CheckPoint
object, not the class definition. In the colab that is the ckpt
variable, but if you're not changing it, you can also just download the file directly. If you want to change it (eg fine tune it), you'll need to construct a new object to pass to checkpoint.dump
.
You need to pass a
graphcast.CheckPoint
object, not the class definition. In the colab that is theckpt
variable, but if you're not changing it, you can also just download the file directly. If you want to change it (eg fine tune it), you'll need to construct a new object to pass tocheckpoint.dump
.
ok, thank you, i have got it.
I have successfully run
graphcast_demo.ipynb
locally, but how do I save the model parameters after the final training using the sample dataset?I try using code:
But it doesn't work.