[Vicuna] Revert the formatting for Brevitas op #1626
Conversation
Abhishek-Varma
force-pushed
the
brevitas_unformat
branch
2 times, most recently
from
2cd1e5e
to
f0f4f83
Compare
-- This commit reverts the formatting for Brevitas op. -- It also excludes vicuna.py script from `black` formatter. Signed-off-by: Abhishek Varma <[email protected]>
Abhishek-Varma
force-pushed
the
brevitas_unformat
branch
from
f0f4f83
to
87c2997
Compare
| rhs_bit_width: int, | ||
| rhs_group_size: int, | ||
| ) -> List[int]: | ||
| def brevitas〇matmul_rhs_group_quant〡shape(lhs: List[int], rhs: List[int], rhs_scale: List[int], rhs_zero_point: List[int], rhs_bit_width: int, rhs_group_size: int) -> List[int]: |
There was a problem hiding this comment.
@stellaraccident do you know how this changes the function signature.
There was a problem hiding this comment.
And @makslevental any thoughts
There was a problem hiding this comment.
I think this is because of the way these functions are verified in Torch-MLIR. They match the functions' first line for verifying the signature (https://github.com/llvm/torch-mlir/blob/main/python/torch_mlir/dialects/torch/importer/jit_ir/build_tools/library_generator.py#L177), and after formatting the function definition goes to multiple lines and as a result, it would fail the verification.
|
@Abhishek-Varma but why revert the other files |
|
Hi @powderluv I wasn't including the other files in this patch initially but the CI showed linting failures in those set of files - opt_params and model_wrappers.py. Also, pyproject was changed to exclude |
|
Opened #1627 to track this issue. so we can unblock the demos |
-- This commit reverts the formatting for Brevitas op.
-- It also excludes vicuna.py script from
blackformatter.Signed-off-by: Abhishek Varma [email protected]