TypeError: MlClient.put_trained_model() got an unexpected keyword argument 'prefix_strings' when use eland to import model with "--ingest-prefix"

[qubit4zj]

Support prefix_strings for MlClient.put_trained_model()

Use eland to import model intfloat/multilingual-e5-small with command:

docker run -it --rm docker.elastic.co/eland/eland eland_import_hub_model --cloud-id [cloud_id] -u elastic -p [pwd] --hub-model-id intfloat/multilingual-e5-small --task-type text_embedding --ingest-prefix "query: " --search-prefix "query: "
Will get error: TypeError: MlClient.put_trained_model() got an unexpected keyword argument 'prefix_strings'
Logs:
_2024-03-13 06:23:31,065 INFO : Establishing connection to Elasticsearch
2024-03-13 06:23:31,168 INFO : Connected to cluster named 'b4cbfe03eea046129e4ed159d14c302d' (version: 8.12.1)
2024-03-13 06:23:31,169 INFO : Loading HuggingFace transformer tokenizer and model 'intfloat/multilingual-e5-small'
tokenizer_config.json:

100%

STAGE:2024-03-13 06:23:42 1:1 ActivityProfilerController.cpp:294] Completed Stage: Warm Up
STAGE:2024-03-13 06:23:43 1:1 ActivityProfilerController.cpp:300] Completed Stage: Collection
2024-03-13 06:23:48,172 INFO : Creating model with id 'test'
Traceback (most recent call last):
File "/usr/local/bin/eland_import_hub_model", line 8, in
sys.exit(main())
File "/usr/local/lib/python3.10/site-packages/eland/cli/eland_import_hub_model.py", line 295, in main
ptm.put_config(config=config)
File "/usr/local/lib/python3.10/site-packages/eland/ml/pytorch/_pytorch_model.py", line 78, in put_config
self._client.ml.put_trained_model(model_id=self.model_id, **config_map)
File "/usr/local/lib/python3.10/site-packages/elasticsearch/sync/client/utils.py", line 426, in wrapped
return api(*args, **kwargs)
TypeError: MlClient.put_trained_model() got an unexpected keyword argument 'prefix_strings'

Elasticsearch version (bin/elasticsearch --version):

elasticsearch-py version (elasticsearch.__versionstr__): 8.12.0

Please make sure the major version matches the Elasticsearch server you are running.

Description of the problem including expected versus actual behavior:

Please add "prefix_strings" as input parameter to:
https://github.com/elastic/elasticsearch-py/blob/main/elasticsearch/_sync/client/ml.py#L3654

 def put_trained_model(
    self,
    *,
    model_id: str,
    ...
    body: t.Optional[t.Dict[str, t.Any]] = None,
    prefix_strings: t.Optional[t.Dict[str, t.Any]] = None, 
) -> ObjectApiResponse[t.Any]: ...
        if tags is not None:
            __body["tags"] = tags
        if prefix_strings is not None: 
            __body["prefix_strings"] = prefix_strings

Steps to reproduce:

[pquentin]

Thanks @qubit4zj for the issue, and sorry for the double transfer. The Python client is generated from this repository, and prefix_strings is missing from the specification:

elasticsearch-specification/specification/ml/put_trained_model/MlPutTrainedModelRequest.ts

Lines 49 to 105 in 4e97d99

	body: {
	/**
	* The compressed (GZipped and Base64 encoded) inference definition of the
	* model. If compressed_definition is specified, then definition cannot be
	* specified.
	*/
	compressed_definition?: string
	/**
	* The inference definition for the model. If definition is specified, then
	* compressed_definition cannot be specified.
	*/
	definition?: Definition
	/**
	* A human-readable description of the inference trained model.
	*/
	description?: string
	/**
	* The default configuration for inference. This can be either a regression
	* or classification configuration. It must match the underlying
	* definition.trained_model's target_type. For pre-packaged models such as
	* ELSER the config is not required.
	*/
	inference_config?: InferenceConfigCreateContainer
	/**
	* The input field names for the model definition.
	*/
	input?: Input
	/**
	* An object map that contains metadata about the model.
	*/
	metadata?: UserDefinedValue
	/**
	* The model type.
	* @server_default tree_ensemble
	*/
	model_type?: TrainedModelType
	/**
	* The estimated memory usage in bytes to keep the trained model in memory.
	* This property is supported only if defer_definition_decompression is true
	* or the model definition is not supplied.
	*/
	model_size_bytes?: long
	/**
	* The platform architecture (if applicable) of the trained mode. If the model
	* only works on one platform, because it is heavily optimized for a particular
	* processor architecture and OS combination, then this field specifies which.
	* The format of the string must match the platform identifiers used by Elasticsearch,
	* so one of, `linux-x86_64`, `linux-aarch64`, `darwin-x86_64`, `darwin-aarch64`,
	* or `windows-x86_64`. For portable models (those that work independent of processor
	* architecture or OS features), leave this field unset.
	*/
	platform_architecture?: string
	/**
	* An array of tags to organize the model.
	*/
	tags?: string[]
	}

@elastic/ml-core Would one of you be able to fix the definition here? Thank you 🙏

[davidkyle]

Thanks for the bug report @qubit4zj

If you have Elasticsearch version 8.12 then the multilingual-e5-small model can be downloaded and installed via the UI

https://www.elastic.co/guide/en/machine-learning/8.12/ml-nlp-e5.html

[davidkyle]

The prefix_strings option was added to the model config used in the response in #2363 but are missing in the PUT request. I've added that in #2449

[pquentin]

Thanks David for your help!

@qubit4zj This will be in the next version of elasticsearch-py that will be released in a few weeks.

[qubit4zj]

@pquentin @davidkyle Thanks for your update! I'll try the new version once it's ready.

[pquentin]

Now that this is in the main branch of elasticsearch-py, you can also test it using pip install "elasticsearch@git+https://github.com/elastic/elasticsearch-py" if you want.