Fix errors when use verl to train GLM4.1v model

* Support glm4v load from AutoModelForVision2Seq
* Set glm4v model _checkpoint_conversion_mapping attr from None to {}

Author: kaln27

src/transformers/models/auto/modeling_auto.py

@@ -877,6 +877,7 @@
         ("blip-2", "Blip2ForConditionalGeneration"),
         ("chameleon", "ChameleonForConditionalGeneration"),
         ("git", "GitForCausalLM"),
+        ("glm4v", "Glm4vForConditionalGeneration"),
         ("idefics2", "Idefics2ForConditionalGeneration"),
         ("idefics3", "Idefics3ForConditionalGeneration"),
         ("instructblip", "InstructBlipForConditionalGeneration"),

src/transformers/models/glm4v/modeling_glm4v.py

@@ -949,7 +949,7 @@ def forward(
 @auto_docstring
 class Glm4vModel(Glm4vPreTrainedModel):
     base_model_prefix = ""
-    _checkpoint_conversion_mapping = None
+    _checkpoint_conversion_mapping = {}
     config_class = Glm4vConfig
     _no_split_modules = ["Glm4vTextDecoderLayer", "Glm4vVisionBlock"]
 
@@ -1382,7 +1382,7 @@ class Glm4vCausalLMOutputWithPast(ModelOutput):
 
 
 class Glm4vForConditionalGeneration(Glm4vPreTrainedModel, GenerationMixin):
-    _checkpoint_conversion_mapping = None
+    _checkpoint_conversion_mapping = {}
     _tied_weights_keys = ["lm_head.weight"]
 
     def __init__(self, config):

src/transformers/models/glm4v/modular_glm4v.py

@@ -1001,7 +1001,7 @@ def forward(
 
 
 class Glm4vModel(Qwen2_5_VLModel):
-    _checkpoint_conversion_mapping = None
+    _checkpoint_conversion_mapping = {}
     _no_split_modules = ["Glm4vTextDecoderLayer", "Glm4vVisionBlock"]
 
     def __init__(self, config):
@@ -1356,7 +1356,7 @@ class Glm4vCausalLMOutputWithPast(Qwen2_5_VLCausalLMOutputWithPast):
 
 
 class Glm4vForConditionalGeneration(Qwen2_5_VLForConditionalGeneration):
-    _checkpoint_conversion_mapping = None
+    _checkpoint_conversion_mapping = {}
 
     def forward(
         self,