add exaone4 support (#2405)

Author: namgyu-youn

gptqmodel/models/auto.py

@@ -87,6 +87,7 @@
 from .definitions.dream import DreamQModel  # noqa: E402
 from .definitions.ernie4_5 import Ernie4_5QModel  # noqa: E402
 from .definitions.ernie4_5_moe import Ernie4_5_MoeQModel  # noqa: E402
+from .definitions.exaone4 import Exaone4QModel  # noqa: E402
 from .definitions.exaone import ExaOneQModel  # noqa: E402
 from .definitions.falcon_h1 import FalconH1QModel  # noqa: E402
 from .definitions.gemma2 import Gemma2QModel  # noqa: E402
@@ -225,6 +226,7 @@
     "deepseek_v3": DeepSeekV3QModel,
     "dots1": Dots1QModel,
     "exaone": ExaOneQModel,
+    "exaone4": Exaone4QModel,
     "grinmoe": GrinMoeQModel,
     "mllama": MLlamaQModel,
     "marin": Qwen3QModel,

gptqmodel/models/definitions/__init__.py

@@ -21,6 +21,7 @@
 from .dots1 import Dots1QModel
 from .dream import DreamQModel
 from .exaone import ExaOneQModel
+from .exaone4 import Exaone4QModel
 from .ernie4_5 import Ernie4_5QModel
 from .ernie4_5_moe import Ernie4_5_MoeQModel
 from .gemma2 import Gemma2QModel

gptqmodel/models/definitions/exaone4.py

@@ -0,0 +1,32 @@
+from ..base import BaseQModel
+
+
+class Exaone4QModel(BaseQModel):
+
+    pre_lm_head_norm_module = "model.norm"
+
+    # Exaone4 uses GQA (Grouped Query Attention) architecture
+    # o_proj must match v_proj shape for AWQ scaling optimizations
+    awq_scale_optimize_shape_dependent_modules = ["self_attn.o_proj"]
+
+    module_tree = [
+        "model",
+        "layers",
+        "#",
+        {
+            # Skip q_norm and k_norm (too small, RMSNorm layers)
+            "self_attn": (
+                "q_proj:0",
+                "k_proj:0",
+                "v_proj:0",
+                "o_proj:1",
+                "q_norm:!",
+                "k_norm:!"
+            ),
+            # Skip post_attention_layernorm (RMSNorm, too small)
+            "post_attention_layernorm": ("post_attention_layernorm:!",),
+            "mlp": ("gate_proj:0", "up_proj:0", "down_proj:1"),
+            # Skip post_feedforward_layernorm (RMSNorm, too small)
+            "post_feedforward_layernorm": ("post_feedforward_layernorm:!",),
+        }
+    ]