fix: change class name of GLM-ASR (#15772)

python/sglang/srt/configs/model_config.py

@@ -1047,7 +1047,7 @@ multimodal_model_archs = [
     "Gemma3nForConditionalGeneration",
     "Glm4vForConditionalGeneration",
     "Glm4vMoeForConditionalGeneration",
-    "GlmasrForConditionalGeneration",
+    "GlmAsrForConditionalGeneration",
     "Grok1VForCausalLM",
     "Grok1AForCausalLM",
     "LlavaLlamaForCausalLM",

python/sglang/srt/models/glmasr.py

@@ -22,10 +22,10 @@ from typing import Any, Iterable, List, Optional, Tuple
 
 import torch
 import torch.nn as nn
-from transformers import GlmasrConfig, GlmasrEncoderConfig
+from transformers import GlmAsrConfig, GlmAsrEncoderConfig
 from transformers.models.glmasr.modeling_glmasr import (
-    GlmasrEncoder,
+    GlmAsrEncoder,
-    GlmasrMultiModalProjector,
+    GlmAsrMultiModalProjector,
 )
 
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
@@ -46,7 +46,7 @@ from sglang.srt.utils import add_prefix
 logger = logging.getLogger(__name__)
 
 
-class GlmasrForConditionalGeneration(nn.Module):
+class GlmAsrForConditionalGeneration(nn.Module):
     # BitandBytes specific attributes
     default_bitsandbytes_target_modules = [
         ".gate_proj.",
@@ -68,7 +68,7 @@ class GlmasrForConditionalGeneration(nn.Module):
 
     def __init__(
         self,
-        config: GlmasrConfig,
+        config: GlmAsrConfig,
         quant_config: Optional[QuantizationConfig] = None,
         prefix: str = "",
     ) -> None:
@@ -77,12 +77,12 @@ class GlmasrForConditionalGeneration(nn.Module):
         self.config = config
 
         if getattr(self.config, "audio_config", None) is None:
-            self.config.audio_config = GlmasrEncoderConfig(self.config._name_or_path)
+            self.config.audio_config = GlmAsrEncoderConfig(self.config._name_or_path)
 
-        self.audio_tower = GlmasrEncoder(
+        self.audio_tower = GlmAsrEncoder(
             config.audio_config,
         )
-        self.multi_modal_projector = GlmasrMultiModalProjector(config)
+        self.multi_modal_projector = GlmAsrMultiModalProjector(config)
         self.language_model = LlamaForCausalLM(
             config.text_config, quant_config, prefix=add_prefix("model", prefix)
         )
@@ -168,4 +168,4 @@ class GlmasrForConditionalGeneration(nn.Module):
                 weight_loader(param, loaded_weight)
 
 
-EntryClass = GlmasrForConditionalGeneration
+EntryClass = GlmAsrForConditionalGeneration

python/sglang/srt/multimodal/processors/base_processor.py

@@ -299,7 +299,7 @@ class BaseMultimodalProcessor(ABC):
         if audios:
             if self._processor.__class__.__name__ in {
                 "Gemma3nProcessor",
-                "GlmasrProcessor",
+                "GlmAsrProcessor",
                 "Qwen2AudioProcessor",
                 "Qwen3OmniMoeProcessor",
             }:

python/sglang/srt/multimodal/processors/glmasr.py

@@ -1,14 +1,14 @@
 import re
 
-from sglang.srt.models.glmasr import GlmasrForConditionalGeneration
+from sglang.srt.models.glmasr import GlmAsrForConditionalGeneration
 from sglang.srt.multimodal.processors.base_processor import (
     BaseMultimodalProcessor,
     MultimodalSpecialTokens,
 )
 
 
-class GlmasrProcessor(BaseMultimodalProcessor):
+class GlmAsrProcessor(BaseMultimodalProcessor):
-    models = [GlmasrForConditionalGeneration]
+    models = [GlmAsrForConditionalGeneration]
 
     def __init__(self, hf_config, server_args, _processor, *args, **kwargs):
         super().__init__(hf_config, server_args, _processor, *args, **kwargs)