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

[Model] Placeholder for ResNet 50 (initial planing & stats) #142

Merged
merged 1 commit into from
Aug 21, 2024

Conversation

nvukobratTT
Copy link
Contributor

List of ops that are currently lowered through tt-forge (up to emit to TTIR):

  • add - already supported
  • hstack - potentially not needed, part of conv2d
  • matmul - potentially not needed, part of conv2d
  • narrow - potentially not needed, part of conv2d
  • pad_tile - potentially not needed, part of conv2d
  • reduce_avg - required support on Forge, MLIR has it
  • reduce_max - required support on Forge and MLIR
  • relu - required support on Forge, MLIR has it
  • sparse_matmul - potentially not needed, part of conv2d
  • squeeze - potentially not needed, used after avg poll to reduce dim to the appropriate output
  • transpose - Currently WIP
  • vslice - potentially not needed

Whole model has around 300 ops before MLIR emitting. However, once we remove conv2d decompositions, this number will be drastically reduced.

Most of the ops are coming from conv2d decompositions. Therefore, bringup for them is probably redundant. Some of them are:

  • hstack
  • matmul
  • narrow
  • pad_tile
  • vslice

@nvukobratTT nvukobratTT merged commit 6a0f808 into main Aug 21, 2024
2 checks passed
@nvukobratTT nvukobratTT deleted the nvukobrat/resnet branch September 9, 2024 17:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants