Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Indexing with slice and list #643

Closed
t-vi opened this issue Jun 24, 2024 · 4 comments · Fixed by #710
Closed

Indexing with slice and list #643

t-vi opened this issue Jun 24, 2024 · 4 comments · Fixed by #710
Assignees
Labels
high priority nemo Issues needed to support NVIDIA NeMo models. nvfuser operators

Comments

@t-vi
Copy link
Collaborator

t-vi commented Jun 24, 2024

Repro:

def fn(input_ids):
   return input_ids[:, [-1, 0]]

a = torch.randn(5, 5)
jfn = thunder.jit(fn)
jfn(a)

Needed for transformers BERT:

import thunder, torch, transformers

m = transformers.BertForSequenceClassification(transformers.BertConfig())
x = torch.randint(1, 10, (1, 32))
jm = thunder.jit(m)
jm(x)

Might ideally be fixed along with #460 and #187

cc @apaz-cli @tfogal

@t-vi
Copy link
Collaborator Author

t-vi commented Jun 25, 2024

@k223kim wants to work on this. Thank you!

@k223kim
Copy link
Contributor

k223kim commented Jun 27, 2024

Hey Team! I will be working on this issue. Thanks!

@t-vi t-vi assigned k223kim and unassigned t-vi Jun 28, 2024
@tfogal tfogal added nemo Issues needed to support NVIDIA NeMo models. and removed triage review labels Jul 8, 2024
@tfogal
Copy link
Collaborator

tfogal commented Jul 8, 2024

triage review:

  • can we modify the checker to test for this case and not send it to nvfuser until the nvfuser bug gets fixed?
  • this is related to the NeVA model
  • cc @kevinstephano

@jacobhinkle
Copy link
Collaborator

until the nvfuser bug gets fixed?

FYI I'll merge NVIDIA/Fuser#2547 soon which should address that bug.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
high priority nemo Issues needed to support NVIDIA NeMo models. nvfuser operators
Projects
None yet
5 participants