feat: add GLM-4.7-Flash (glm4_moe_lite) model support

vllm_fl/__init__.py

@@ -96,7 +96,7 @@ def register_model():
     try:
         ModelRegistry.register_model(
             "KimiK25ForConditionalGeneration",
-            "vllm_fl.models.kimi_k25:KimiK25ForConditionalGeneration",
+            "vllm_fl.models.kimi_k25:KimiK25ForConditionalGeneration"
         )
     except Exception as e:
         logger.error(f"Register KimiK25 model error: {str(e)}")
@@ -116,3 +116,16 @@ def register_model():
         )
     except Exception as e:
         logger.error(f"Register GlmMoeDsa model error: {str(e)}")
+
+    # Register GLM-4.7-Flash (Glm4MoeLite) model
+    try:
+        from vllm.transformers_utils.config import _CONFIG_REGISTRY
+        from vllm_fl.configs.glm4_moe_lite import Glm4MoeLiteConfig
+        _CONFIG_REGISTRY["glm4_moe_lite"] = Glm4MoeLiteConfig
+
+        ModelRegistry.register_model(
+            "Glm4MoeLiteForCausalLM",
+            "vllm_fl.models.glm4_moe_lite:Glm4MoeLiteForCausalLM"
+        )
+    except Exception as e:
+        logger.error(f"Register Glm4MoeLite model error: {str(e)}")

vllm_fl/configs/glm4_moe_lite.py

@@ -0,0 +1,56 @@
+# SPDX-License-Identifier: Apache-2.0
+"""GLM-4.7-Flash (Glm4MoeLite) config bridge for vLLM plugin.
+
+transformers 4.57.6 does not recognise model_type ``glm4_moe_lite``.
+This config bridge lets vLLM load the HuggingFace checkpoint without
+upgrading transformers.
+
+GLM-4.7-Flash uses MLA (Multi-head Latent Attention) from DeepSeek V2/V3
+and MoE architecture from GLM-4 MoE.
+"""
+
+from transformers.models.glm4_moe import Glm4MoeConfig
+
+
+class Glm4MoeLiteConfig(Glm4MoeConfig):
+    model_type = "glm4_moe_lite"
+
+    def __init__(
+        self,
+        # MLA (Multi-head Latent Attention) fields
+        qk_nope_head_dim: int = 0,
+        qk_rope_head_dim: int = 0,
+        v_head_dim: int = 0,
+        kv_lora_rank: int = 0,
+        q_lora_rank: int | None = None,
+        # DSA Indexer fields (v32 variant)
+        index_topk: int | None = None,
+        index_n_heads: int | None = None,
+        index_head_dim: int | None = None,
+        indexer_rope_interleave: bool | None = None,
+        # MoE frequency
+        moe_layer_freq: int = 1,
+        # MTP (speculative decoding)
+        num_nextn_predict_layers: int = 0,
+        # Additional fields
+        dtype: str = "bfloat16",
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        # MLA fields
+        self.qk_nope_head_dim = qk_nope_head_dim
+        self.qk_rope_head_dim = qk_rope_head_dim
+        self.v_head_dim = v_head_dim
+        self.kv_lora_rank = kv_lora_rank
+        self.q_lora_rank = q_lora_rank
+        # DSA Indexer fields
+        self.index_topk = index_topk
+        self.index_n_heads = index_n_heads
+        self.index_head_dim = index_head_dim
+        self.indexer_rope_interleave = indexer_rope_interleave
+        # MoE frequency
+        self.moe_layer_freq = moe_layer_freq
+        # MTP fields
+        self.num_nextn_predict_layers = num_nextn_predict_layers
+        # Additional
+        self.dtype = dtype