David Kyle

Results 26 issues of David Kyle

Guiding the user through the configuration options and detect misconfigurations early.

topic:NLP

#479 shows that new versions of PyTorch will sometimes be incompatible with older versions. In this case a Torchscript model traced in version 1.12 of PyTorch cannot be evaluated in...

topic:ml

Using the following commands to install eland generates a warning ``` python -m venv venv . venv/bin/activate python -m pip install build wheel python -m build python -m pip install...

enhancement
topic:ml

Closes #669

enhancement

PyTorch 1.13 is not compatible with Python 3.11, #669 has to be implemented before - [ ] Update compatibility statement https://github.com/elastic/eland?tab=readme-ov-file#compatibility - [ ] Add Python 3.11 to the ci...

enhancement

Pytorch was upgraded to 2.1.2 in https://github.com/elastic/ml-cpp/pull/2588 can now be upgraded to the same version. Models traced in Pytorch 2.1.2 may not be compatible with version 1.13.1 of libtorch used...

enhancement
topic:NLP

The tests added in #623 can cause occasional build failures. Sometimes the console log just ends without any errors and sometimes there is a message along the lines of `pytest...

ci
topic:NLP

The best way to check if a model is compatible with Elasticsearch is to run the `eland_import_hub_model.py` script but this requires the user to configure the Elasticsearch connection settings and...

enhancement

Using the [docker](https://www.elastic.co/guide/en/machine-learning/master/ml-nlp-import-model.html#ml-nlp-import-docker) container to run `eland_import_hub_model` and install a large model imposes large memory requirements. For example using the command below to install [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) requires the container to have...

topic:NLP

Several learning to rank tests are failing with the error: ``` Arrays are not almost equal to 2 decimals ``` ``` FAILED tests/ml/test_ml_model_pytest.py::TestMLModel::test_learning_to_rank[rank:ndcg-False] FAILED tests/ml/test_ml_model_pytest.py::TestMLModel::test_learning_to_rank[rank:map-True] FAILED tests/ml/test_ml_model_pytest.py::TestMLModel::test_learning_to_rank[rank:map-False] FAILED tests/ml/test_ml_model_pytest.py::TestMLModel::test_learning_to_rank[rank:pairwise-True] FAILED...

bug