Add cross encoder support

[HenryL27]

Description

Adds support for (huggingface) cross encoders to ml-commons. Uses a new function name (TEXT_SIMILARITY) which takes as input a list of text pairs and spits out 1-dimensional tensors representing the similarity of the items in each pair. E.g.

{
  "query_text": "today is sunny"
  "text_docs": [ 
    "today is sunny",
    "today is july fifth",
    "it is winter"
  ] 
}

yields

{
  "inference_results": [
    {
      "output": [
        {
          "name": "logits",
          "data_type": "FLOAT32",
          "shape": [
            1
          ],
          "data": [
            10.939743
          ],
          "byte_buffer": {
            "array": "MAkvQQ==",
            "order": "LITTLE_ENDIAN"
          }
        }
      ]
    },
    {
      "output": [
        {
          "name": "logits",
          "data_type": "FLOAT32",
          "shape": [
            1
          ],
          "data": [
            -6.067284
          ],
          "byte_buffer": {
            "array": "MSfCwA==",
            "order": "LITTLE_ENDIAN"
          }
        }
      ]
    },
    {
      "output": [
        {
          "name": "logits",
          "data_type": "FLOAT32",
          "shape": [
            1
          ],
          "data": [
            -11.261627
          ],
          "byte_buffer": {
            "array": "oC80wQ==",
            "order": "LITTLE_ENDIAN"
          }
        }
      ]
    }
  ]
}

This was using the model cross-encoder/ms-marco-TinyBERT-L-2-v2 - the config I used to upload it looked like

{
  "name": model_name,
  "version": "1.0.0",
  "description": "Cross Encoder text similarity model",
  "model_format": "TORCH_SCRIPT",
  "function_name": "TEXT_SIMILARITY",
  "model_content_hash_value": hash_value,
  "model_config": {
    "model_type": "bert",
    "embedding_dimension": 1,
    "framework_type": "huggingface_transformers",
    "all_config": cfg.to_json_string(),
  }
}

Issues Resolved

• Cross encoder issue: [FEATURE] Support local cross-encoder model #1589
• Model Types meta issue: [RFC] Support more local model types #1164
• Neural-search cross encoder reranking RFC: opensearch-project/neural-search/485

Check List

• [ x] New functionality includes testing.
  • [ x] All tests pass
• New functionality has been documented.
  • New functionality has javadoc added
• [ x] Commits are signed per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[codecov]

Codecov Report

Attention: 12 lines in your changes are missing coverage. Please review.

Comparison is base (df644ff) 80.83% compared to head (be113ed) 80.98%.
Report is 7 commits behind head on main.

Files	Patch %	Lines
...rch/ml/common/input/nlp/TextSimilarityMLInput.java	86.95%	2 Missing and 4 partials ⚠️
...n/java/org/opensearch/ml/common/input/MLInput.java	66.66%	3 Missing and 2 partials ⚠️
...in/java/org/opensearch/ml/common/FunctionName.java	66.66%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #1615      +/-   ##
============================================
+ Coverage     80.83%   80.98%   +0.15%     
- Complexity     4215     4246      +31     
============================================
  Files           404      408       +4     
  Lines         16977    17122     +145     
  Branches       1818     1835      +17     
============================================
+ Hits          13723    13867     +144     
+ Misses         2539     2534       -5     
- Partials        715      721       +6     

Flag	Coverage Δ
ml-commons	80.98% <91.83%> (+0.15%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[austintlee]

One minor question, but overall looks great!

[dhrubo-os]

Thanks for working on this. Approved.

[austintlee]

LGTM (with one minor question. You can answer and resolve.)

[martin-gaievski]

@HenryL27 can you please share details of meta config for ms-marco-TinyBERT-L-2-v2 model?
I'm using following request but I'm getting errors, probably some param is missing:

POST /_plugins/_ml/models/meta
{
    "name": "ms-marco-TinyBERT-L-2-v2",
    "version": "1.0.0",
    "function_name": "TEXT_SIMILARITY",
    "description": "test model",
    "model_format": "TORCH_SCRIPT",
    "model_group_id": "<MODEL_GROUP_ID>",
    "model_content_hash_value": "90e39a926101d1a4e542aade0794319404689b12acfd5d7e65c03d91c668b5cf",
    "model_config": {
        "model_type": "bert",
        "embedding_dimension": 1,
        "framework_type": "huggingface_transformers",
        "all_config": "{\"total_chunks\":2,\"is_hidden\":false}"
    },
    "url": "https://github.com/opensearch-project/ml-commons/blob/main/ml-algorithms/src/test/resources/org/opensearch/ml/engine/algorithms/text_similarity/TinyBERT-CE-torch_script.zip?raw=true"
}

error response:

        "type": "illegal_argument_exception",
        "reason": "total chunks field is null"