Add e5-multilingual to known configurations (#533)

# Description

Adds e5-multilingual to known configurations. PELE tests are passing and
have been added with another PR.

---------

Signed-off-by: Max de Bayser <[email protected]>
Signed-off-by: Christian Kadner <[email protected]>
Co-authored-by: Christian Kadner <[email protected]>

Author: maxdebayser

.github/workflows/test.yml

@@ -119,6 +119,8 @@ jobs:
             uv.lock
             tests/**/*.py
             vllm_spyre/**/*.py
+            vllm_spyre/config/known_model_configs.json
+            vllm_spyre/config/supported_configs.yaml
 
       - name: "Install PyTorch 2.7.1"
         if: steps.changed-src-files.outputs.any_changed == 'true'
@@ -157,6 +159,7 @@ jobs:
 
       - name: "Standardize HF model names for caching"
         id: standardize-names
+        if: steps.changed-src-files.outputs.any_changed == 'true'
         run: |
           # replace '/' characters in HF_MODEL with '--' for GHA cache keys and
           # in model file names in local HF hub cache

docs/user_guide/supported_models.md

@@ -35,13 +35,15 @@ configurations.
 | [BAAI/BGE-Reranker (v2-m3)][]             | 1    | 8192           | 1          |
 | [BAAI/BGE-Reranker (Large)][]             | 1    | 512            | 1          |
 | [BAAI/BGE-Reranker (Large)][]             | 1    | 512            | 64         |
+| [Multilingual-E5-large][]                 | 1    | 512            | 64         |
 
 [Granite-3.3-8b]: https://huggingface.co/ibm-granite/granite-3.3-8b-instruct
 [Granite-3.3-8b (FP8)]: https://huggingface.co/ibm-granite/granite-3.3-8b-instruct
 [Granite-Embedding-125m (English)]: https://huggingface.co/ibm-granite/granite-embedding-125m-english
 [Granite-Embedding-278m (Multilingual)]: https://huggingface.co/ibm-granite/granite-embedding-278m-multilingual
 [BAAI/BGE-Reranker (v2-m3)]: https://huggingface.co/BAAI/bge-reranker-v2-m3
 [BAAI/BGE-Reranker (Large)]: https://huggingface.co/BAAI/bge-reranker-large
+[Multilingual-E5-large]: https://huggingface.co/intfloat/multilingual-e5-large
 
 ## Runtime Validation
 

tests/download_model_configs.py

@@ -1,3 +1,5 @@
+#!/usr/bin/env python3
+
 import os
 from pathlib import Path
 from urllib.request import urlretrieve

tests/fixtures/model_configs/intfloat/multilingual-e5-large/config.json

@@ -0,0 +1,27 @@
+{
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.55.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 250002
+}

tests/utils/test_model_config_validator.py

@@ -212,9 +212,12 @@ def test_find_model_by_config(monkeypatch, caplog):
             assert model_config.model != model_id
 
             models_found = find_known_models_by_model_config(model_config)
-            assert len(models_found) == 1
+            assert len(models_found) == 1, \
+                (f"More than one model found. Need to add more distinguishing"
+                 f" parameters for models `{models_found}` in file"
+                 f" `vllm_spyre/config/known_model_configs.json`")
             assert models_found[0] == model_id
 
             validate(model_config)
-            assert f"Model '{model_config.model}' is not a known model"
+            assert f"Model '{model_config.model}' is not a known" in caplog.text
             assert f"Found model '{model_id}'" in caplog.text

vllm_spyre/config/known_model_configs.json

@@ -1,5 +1,6 @@
 {
   "BAAI/bge-reranker-large": {
+    "architectures": ["XLMRobertaForSequenceClassification"],
     "model_type": "xlm-roberta",
     "max_position_embeddings": 514,
     "num_hidden_layers": 24,
@@ -39,5 +40,11 @@
     "model_type": "roberta",
     "num_hidden_layers": 24,
     "vocab_size": 50265
+  },
+  "intfloat/multilingual-e5-large": {
+    "architectures": ["XLMRobertaModel"],
+    "model_type": "xlm-roberta",
+    "num_hidden_layers": 24,
+    "vocab_size": 250002
   }
-}
+}

vllm_spyre/config/supported_configs.yaml

@@ -31,6 +31,10 @@
   configs: [
     { cb: False, tp_size: 1, warmup_shapes: [[512, 0, 64]] },
   ]
+- model: "intfloat/multilingual-e5-large"
+  configs: [
+    { cb: False, tp_size: 1, warmup_shapes: [[512, 0, 64]] },
+  ]
 - model: "BAAI/bge-reranker-v2-m3"
   configs: [
     { cb: False, tp_size: 1, warmup_shapes: [[8192, 0, 1]] },