Skip to content

fix: Repair integer inputs in dynamic shape cases #2876

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.

Sign up for GitHub

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

Jump to bottom
Merged
merged 11 commits into from
Jun 25, 2024
Merged

fix: Repair integer inputs in dynamic shape cases #2876

merged 11 commits into from
Jun 25, 2024

Conversation

gs-olive
Copy link
Collaborator

@gs-olive gs-olive commented Jun 1, 2024

Description

  • Generative inference with HF text generation models such as gpt2 can fail if graph segmentation causes a symbolic integer to be passed from Torch to TRT, since the Torch output is an integer, while TRT expects a tensor
  • Added logic to the modules to address this case
  • Added test cases to validate generation with both Python and C++ runtimes

Type of change

  • Bug fix (non-breaking change which fixes an issue)

Checklist:

  • [ x ] My code follows the style guidelines of this project (You can use the linters)
  • [ x ] I have performed a self-review of my own code
  • [ x ] I have commented my code, particularly in hard-to-understand areas and hacks
  • [ x ] I have made corresponding changes to the documentation
  • [ x ] I have added tests to verify my fix or my feature
  • [ x ] New and existing unit tests pass locally with my changes
  • [ x ] I have added the relevant labels to my PR in so that relevant reviewers are notified

@gs-olive gs-olive requested a review from peri044 June 1, 2024 04:30
@gs-olive gs-olive self-assigned this Jun 1, 2024
@github-actions github-actions bot added component: tests Issues re: Tests component: api [Python] Issues re: Python API component: runtime component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths labels Jun 1, 2024
gs-olive
gs-olive commented Jun 5, 2024
View reviewed changes
@gs-olive
Copy link
Collaborator Author

gs-olive commented Jun 5, 2024

--> Cherry pick to release/2.3

@github-actions github-actions bot requested a review from narendasan June 5, 2024 17:58
@gs-olive gs-olive mentioned this pull request Jun 5, 2024
Merged
@gs-olive gs-olive force-pushed the dynamic_shapes_integer_input_bugfix branch from ef83457 to 5906ab4 Compare June 7, 2024 00:36
@github-actions github-actions bot added documentation Improvements or additions to documentation component: lowering Issues re: The lowering / preprocessing passes component: conversion Issues re: Conversion stage component: core Issues re: The core compiler component: converters Issues re: Specific op converters component: torch_compile labels Jun 7, 2024
@gs-olive gs-olive changed the base branch from dyn_llama_cp to main June 7, 2024 00:38
@gs-olive gs-olive force-pushed the dynamic_shapes_integer_input_bugfix branch from 5906ab4 to 7b0810d Compare June 7, 2024 00:38
@github-actions github-actions bot removed documentation Improvements or additions to documentation component: lowering Issues re: The lowering / preprocessing passes component: conversion Issues re: Conversion stage component: core Issues re: The core compiler component: converters Issues re: Specific op converters component: torch_compile labels Jun 7, 2024
@gs-olive gs-olive force-pushed the dynamic_shapes_integer_input_bugfix branch 3 times, most recently from 046291c to 343d610 Compare June 17, 2024 19:06
@gs-olive gs-olive force-pushed the dynamic_shapes_integer_input_bugfix branch from f960816 to e99514b Compare June 18, 2024 19:08
gs-olive added 10 commits June 24, 2024 15:55
@gs-olive
fix: Repair integer inputs in dynamic shape cases
0ff1572 
- Generative inference with HF text generation models such as gpt2 can
fail if graph segmentation causes a symbolic integer to be passed from
Torch to TRT, since the Torch output is an integer, while TRT expects a
tensor
- Added logic to the modules to address the above case
- Added test cases to validate generation with both Python and C++
runtimes
@gs-olive
bump transformers version
b63c699
@gs-olive
Test upgrade transformers version
cabe918
@gs-olive
Test downgrade transformers version
9a6f676
@gs-olive
Add custom model for dynamic model checking
4fc8ad1
@gs-olive
Shape fix
8f2b7e5
@gs-olive
Shape fix 2
3680653
@gs-olive
Shape fix 3
959bdac
@gs-olive
Shape fix 4
54047df
@gs-olive
Change fallback operator
0d8ae44
@gs-olive gs-olive force-pushed the dynamic_shapes_integer_input_bugfix branch from 89a960e to 0d8ae44 Compare June 24, 2024 22:56
@gs-olive gs-olive merged commit caf3a92 into main Jun 25, 2024
@gs-olive gs-olive deleted the dynamic_shapes_integer_input_bugfix branch June 25, 2024 01:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@peri044 peri044 Awaiting requested review from peri044

@narendasan narendasan Awaiting requested review from narendasan

Assignees

@gs-olive gs-olive

Labels
cla signed component: api [Python] Issues re: Python API component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths component: runtime component: tests Issues re: Tests needs-release-cherrypick
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@gs-olive @facebook-github-bot