GLM-4.5V with new class name at transformers (#22520)

Signed-off-by: zRzRzRzRzRzRzR <2448370773@qq.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
Yuxuan Zhang
2025-08-09 00:50:21 -07:00
committed by GitHub
co-authored by Isotr0py
parent 2be07a0db1
commit a6022e6fbc
5 changed files with 13 additions and 6 deletions
+7 -1
View File
@@ -372,7 +372,13 @@ class Glm4MoeDecoderLayer(nn.Module):
return hidden_states, residual
@support_torch_compile
@support_torch_compile(
dynamic_arg_dims={
"input_ids": 0,
"positions": -1,
"intermediate_tensors": 0,
"inputs_embeds": 0,
})
class Glm4MoeModel(nn.Module):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
+1 -1
View File
@@ -207,7 +207,7 @@ _MULTIMODAL_MODELS = {
"Gemma3ForConditionalGeneration": ("gemma3_mm", "Gemma3ForConditionalGeneration"), # noqa: E501
"GLM4VForCausalLM": ("glm4v", "GLM4VForCausalLM"),
"Glm4vForConditionalGeneration": ("glm4_1v", "Glm4vForConditionalGeneration"), # noqa: E501
"Glm4v_moeForConditionalGeneration": ("glm4_1v", "Glm4vForConditionalGeneration"), # noqa: E501
"Glm4vMoeForConditionalGeneration": ("glm4_1v", "Glm4vForConditionalGeneration"), # noqa: E501
"GraniteSpeechForConditionalGeneration": ("granite_speech", "GraniteSpeechForConditionalGeneration"), # noqa: E501
"H2OVLChatModel": ("h2ovl", "H2OVLChatModel"),
"InternVLChatModel": ("internvl", "InternVLChatModel"),