emb working

config.json

@@ -1,13 +1,12 @@
 {
   "architectures": [
-    "AriaForSequenceEmbedding"
+    "AriaForCausalLM"
   ],
   "bos_token_id": 0,
   "eos_token_id": 1,
   "hidden_size": 1536,
-  "embedding_size": 512,
   "intermediate_size": 6144,
-  "max_seq_len": 2048,
+  "max_seq_len": 8192,
   "model_type": "aria",
   "num_attention_heads": 24,
   "num_hidden_layers": 16,
@@ -18,6 +17,6 @@
   "auto_map": {
     "AutoConfig": "configuration_aria.AriaConfig", 
     "AutoModel": "modeling_aria.AriaModel",
-    "AutoModelForCausalLM": "modeling_aria.AriaForSequenceEmbedding"
+    "AutoModelForCausalLM": "modeling_aria.AriaForCausalLM"
   }
 }

config_emb.json

@@ -0,0 +1,23 @@
+{
+  "architectures": [
+    "AriaForSequenceEmbedding"
+  ],
+  "bos_token_id": 0,
+  "eos_token_id": 1,
+  "hidden_size": 1536,
+  "embedding_size": 512,
+  "intermediate_size": 6144,
+  "max_seq_len": 2048,
+  "model_type": "aria",
+  "num_attention_heads": 24,
+  "num_hidden_layers": 16,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.45.0",
+  "use_cache": true,
+  "vocab_size": 17727,
+  "auto_map": {
+    "AutoConfig": "configuration_aria.AriaConfig", 
+    "AutoModel": "modeling_aria.AriaModel",
+    "AutoModelForCausalLM": "modeling_aria.AriaForSequenceEmbedding"
+  }
+}

modeling_aria.py

@@ -659,9 +659,7 @@ class AriaForSequenceEmbedding(AriaPreTrainedModel):
         _batch_size = input_ids.shape[0]
         eos_mask = input_ids == self.config.eos_token_id
         if not eos_mask.any(dim=1).all():
-            raise ValueError(
-                "Each sequence must contain at least one EOS token"
-            )
+            raise ValueError("Each sequence must contain a EOS token")
         eos_pos = eos_mask.int().argmax(dim=1)
 
         pooled_embedding = embedding[