Upload Ouro-2.6B_smoothquant_W8A8 with bundled source code

config.json

@@ -5,7 +5,7 @@
   "attention_dropout": 0.0,
   "auto_map": {
     "AutoConfig": "configuration_ouro.OuroConfig",
-    "AutoModelForCausalLM": "qouro_runtime.modeling_qouro::OuroForCausalLMQuantized"
+    "AutoModelForCausalLM": "modeling_qouro.OuroForCausalLMQuantized"
   },
   "bos_token_id": 1,
   "early_exit_threshold": 1.0,

modeling_qouro.py

@@ -0,0 +1 @@
+from qouro_runtime.modeling_qouro import OuroForCausalLMQuantized as OuroForCausalLMQuantized

qouro_runtime/quantization/pipeline.py

@@ -142,8 +142,10 @@ def run_quantization_pipeline(
     quantized_config.quantization = quant_config_dict
     quantized_config.architectures = ["OuroForCausalLMQuantized"]
     quantized_config.auto_map = {
-        "AutoModelForCausalLM": "qouro_runtime.modeling_qouro::OuroForCausalLMQuantized",
-        "AutoConfig": "qouro_runtime.configuration_ouro.OuroConfig",
+        # AutoModel 也采用 module_file.ClassName（根目录存在 modeling_qouro.py 包装文件）
+        "AutoModelForCausalLM": "modeling_qouro.OuroForCausalLMQuantized",
+        # AutoConfig 采用 module_file.ClassName（根目录存在 configuration_ouro.py）
+        "AutoConfig": "configuration_ouro.OuroConfig",
     }
 
     quantized_model = OuroForCausalLMQuantized(quantized_config)