Skip to content

fix(doc_vlm): remove ROCm BF16 _keep_in_fp32_modules workaround in PaddleOCR-VL #5077

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

Open
fchange wants to merge 1 commit into
base: develop
Choose a base branch
from
Open

fix(doc_vlm): remove ROCm BF16 _keep_in_fp32_modules workaround in PaddleOCR-VL #5077

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 5 additions & 3 deletions paddlex/inference/models/doc_vlm/modeling/paddleocr_vl/_paddleocr_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,9 +65,11 @@ class PaddleOCRVLForConditionalGeneration(Ernie4_5PretrainedModel):
_tied_weights_keys = ["lm_head.weight"]
config_class = PaddleOCRVLConfig
_no_split_modules = ["Ernie4_5DecoderLayer", "SiglipEncoderLayer"]
# Keep visual encoder in fp32 for ROCm stability (MIOpen bf16 conv has bugs)
# This also improves precision for vision processing
_keep_in_fp32_modules = ["visual", "mlp_AR"]
# NOTE: Previously _keep_in_fp32_modules = ["visual", "mlp_AR"] was set here
# to work around MIOpen BF16 convolution bugs on ROCm. This is no longer needed
# after Paddle framework adds BF16 conv kernel registration for HIP backend.
# See: https://github.com/PaddlePaddle/Paddle/pull/78587
_keep_in_fp32_modules = None
base_model_prefix = ""

def __init__(self, config):
Expand Down