Skip to content

[fix] Update dtype logic for huggingface backend for new containers #3847

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

Merged
merged 1 commit into from
May 15, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
4 changes: 3 additions & 1 deletion src/sagemaker/djl_inference/model.py
Original file line number Diff line number Diff line change
Expand Up @@ -854,11 +854,13 @@ def generate_serving_properties(self, serving_properties=None) -> Dict[str, str]
if self.low_cpu_mem_usage:
serving_properties["option.low_cpu_mem_usage"] = self.low_cpu_mem_usage
# This is a workaround due to a bug in our built in handler for huggingface
# TODO: This needs to be fixed when new dlc is published
# TODO: Remove this logic whenever 0.20.0 image is out of service
if (
serving_properties["option.entryPoint"] == "djl_python.huggingface"
and self.dtype
and self.dtype != "auto"
and self.djl_version
and int(self.djl_version.split(".")[1]) < 21
):
serving_properties["option.dtype"] = "auto"
serving_properties.pop("option.load_in_8bit", None)
Expand Down
2 changes: 1 addition & 1 deletion tests/unit/test_djl_inference.py
Original file line number Diff line number Diff line change
Expand Up @@ -454,7 +454,7 @@ def test_generate_serving_properties_with_valid_configurations(
"option.entryPoint": "djl_python.huggingface",
"option.s3url": VALID_UNCOMPRESSED_MODEL_DATA,
"option.tensor_parallel_degree": 1,
"option.dtype": "auto",
"option.dtype": "fp32",
"option.device_id": 4,
"option.device_map": "balanced",
}
Expand Down