Complete Sentence Transformers integration + patch inference on CPU & Windows (#4)

- Remove reference_compile; complete Sentence Transformers support (5c66f04b1ffb7a7db1a480361c8a574deed2e3b7)
- Notify users that flash_attn is recommended (b94cf64d6cb8ea2e1746f7e6df169b8300804035)

README.md

@@ -6,6 +6,8 @@ base_model:
 - answerdotai/ModernBERT-base
 pipeline_tag: sentence-similarity
 library_name: transformers
+tags:
+- sentence-transformers
 ---
 
 # gte-modernbert-base
@@ -33,10 +35,17 @@ The `gte-modernbert` models demonstrates competitive performance in several text
 
 ## Usage
 
-Use with `Transformers`
+> [!TIP]
+> For `transformers` and `sentence-transformers`, if your GPU supports it, the efficient Flash Attention 2 will be used automatically if you have `flash_attn` installed. It is not mandatory.
+> 
+> ```bash
+> pip install flash_attn
+> ```
+
+Use with `transformers`
 
 ```python
-# Requires transformers>=4.36.0
+# Requires transformers>=4.48.0
 
 import torch.nn.functional as F
 from transformers import AutoModel, AutoTokenizer
@@ -48,9 +57,9 @@ input_texts = [
     "sorting algorithms"
 ]
 
-model_path = 'Alibaba-NLP/gte-modernbert-base'
+model_path = "Alibaba-NLP/gte-modernbert-base"
 tokenizer = AutoTokenizer.from_pretrained(model_path)
-model = AutoModel.from_pretrained(model_path, trust_remote_code=True)
+model = AutoModel.from_pretrained(model_path)
 
 # Tokenize the input texts
 batch_dict = tokenizer(input_texts, max_length=8192, padding=True, truncation=True, return_tensors='pt')
@@ -62,21 +71,31 @@ embeddings = outputs.last_hidden_state[:, 0]
 embeddings = F.normalize(embeddings, p=2, dim=1)
 scores = (embeddings[:1] @ embeddings[1:].T) * 100
 print(scores.tolist())
+# [[42.89073944091797, 71.30911254882812, 33.664554595947266]]
 ```
 
 Use with `sentence-transformers`:
 
 ```python
-# Requires sentence_transformers>=2.7.0
-
+# Requires transformers>=4.48.0
 from sentence_transformers import SentenceTransformer
 from sentence_transformers.util import cos_sim
 
-sentences = ['That is a happy person', 'That is a very happy person']
+input_texts = [
+    "what is the capital of China?",
+    "how to implement quick sort in python?",
+    "Beijing",
+    "sorting algorithms"
+]
+
+model = SentenceTransformer("Alibaba-NLP/gte-modernbert-base")
+embeddings = model.encode(input_texts)
+print(embeddings.shape)
+# (4, 768)
 
-model = SentenceTransformer('Alibaba-NLP/gte-modernbert-base', trust_remote_code=True)
-embeddings = model.encode(sentences)
-print(cos_sim(embeddings[0], embeddings[1]))
+similarities = cos_sim(embeddings[0], embeddings[1:])
+print(similarities)
+# tensor([[0.4289, 0.7131, 0.3366]])
 ```
 
 Use with `transformers.js`:

config.json

@@ -35,7 +35,6 @@
   "num_hidden_layers": 22,
   "pad_token_id": 50283,
   "position_embedding_type": "absolute",
-  "reference_compile": true,
   "sep_token_id": 50282,
   "sparse_pred_ignore_index": -100,
   "sparse_prediction": false,

config_sentence_transformers.json

@@ -0,0 +1,10 @@
+{
+  "__version__": {
+    "sentence_transformers": "2.7.0",
+    "transformers": "4.48.0",
+    "pytorch": "2.5.0+cu121"
+  },
+  "prompts": {},
+  "default_prompt_name": null,
+  "similarity_fn_name": "cosine"
+}

modules.json

@@ -0,0 +1,14 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  }
+]