biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[GLM-OCR] GLM-OCR with MTP Support (#33005)

Browse Source 
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
2026-01-26 22:24:43 +08:00
committed by GitHub
parent dcd80206b7
commit bb17e8f11c
14 changed files with 873 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/registry.py
View File

@@ -319,8 +319,9 @@ _MULTIMODAL_MODELS = {
),
"GlmAsrForConditionalGeneration": ("glmasr", "GlmAsrForConditionalGeneration"),
"GLM4VForCausalLM": ("glm4v", "GLM4VForCausalLM"),
"Glm4vForConditionalGeneration": ("glm4_1v", "Glm4vForConditionalGeneration"), # noqa: E501
"Glm4vMoeForConditionalGeneration": ("glm4_1v", "Glm4vMoeForConditionalGeneration"), # noqa: E501
"Glm4vForConditionalGeneration": ("glm4_1v", "Glm4vForConditionalGeneration"),
"Glm4vMoeForConditionalGeneration": ("glm4_1v", "Glm4vMoeForConditionalGeneration"),
"GlmOcrForConditionalGeneration": ("glm_ocr", "GlmOcrForConditionalGeneration"), # noqa: E501
"GraniteSpeechForConditionalGeneration": (
"granite_speech",
"GraniteSpeechForConditionalGeneration",
@@ -472,6 +473,7 @@ _SPECULATIVE_DECODING_MODELS = {
"LongCatFlashMTPModel": ("longcat_flash_mtp", "LongCatFlashMTP"),
"Glm4MoeMTPModel": ("glm4_moe_mtp", "Glm4MoeMTP"),
"Glm4MoeLiteMTPModel": ("glm4_moe_lite_mtp", "Glm4MoeLiteMTP"),
"GlmOcrMTPModel": ("glm_ocr_mtp", "GlmOcrMTP"),
"MedusaModel": ("medusa", "Medusa"),
"OpenPanguMTPModel": ("openpangu_mtp", "OpenPanguMTP"),
"Qwen3NextMTP": ("qwen3_next_mtp", "Qwen3NextMTP"),