third_codes/eland
1
0
Fork 0
mirror of https://github.com/elastic/eland.git synced 2025-07-11 00:02:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Default truncation to second for text similarity the task type(#713)

Browse Source 
In reranking the first input (the query) is generally shorter. In this case
it makes more sense to truncate the second input (the document text)
This commit is contained in:
David Kyle 2024-08-05 11:47:15 +01:00 committed by GitHub
parent bee6d0e1f7
commit fd8886da6a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
eland/ml/pytorch/transformers.py
View File

@ -770,6 +770,9 @@ class TransformerModel:
tokenization_config.span = 128
tokenization_config.truncate = "none"
if self._task_type == "text_similarity":
tokenization_config.truncate = "second"
if self._traceable_model.classification_labels():
inference_config = TASK_TYPE_TO_INFERENCE_CONFIG[self._task_type](
tokenization=tokenization_config,

3
tests/ml/pytorch/test_pytorch_model_config_pytest.py
View File

@ -217,6 +217,9 @@ class TestModelConfguration:
assert isinstance(config.inference_config.classification_labels, list)
assert len(config.inference_config.classification_labels) > 0
if task_type == "text_similarity":
assert tokenization.truncate == "second"
del tm
def test_model_config_with_prefix_string(self):