Closed t-vi closed 4 days ago
This needs to work for transformers (and thus is high-impact / high prio):
from dataclasses import dataclass import torch import thunder @dataclass class MyContainer: res: torch.Tensor num: int def fn(x): return MyContainer(x, 1) jfn = thunder.jit(fn) x = torch.randn(5, 5) print(jfn(x))
(somewhat related is #461 , but that is to not have infinite recursion, here we want the feature)
cc @apaz-cli
This needs to work for transformers (and thus is high-impact / high prio):
(somewhat related is #461 , but that is to not have infinite recursion, here we want the feature)
cc @apaz-cli